Causes and consequences of hysteresis: aggregate demand, productivity, and employment

Causes and consequences of hysteresis: aggregate demand, productivity, and employment Abstract In this work we develop an agent-based model where hysteresis in major macroeconomic variables (e.g., gross domestic product, productivity, unemployment) emerges out of the decentralized interactions of heterogeneous firms and workers. Building upon the “Schumpeter meeting Keynes” family of models (cf. in particular Dosi et al. (2016b, 2017c)), we specify an endogenous process of accumulation of workers’ skills and a state-dependent process of firms entry. Indeed, hysteresis is ubiquitous. However, this is not due to market imperfections, but rather to the very functioning of decentralized economies characterized by coordination externalities and dynamic increasing returns. So, contrary to the insider–outsider hypothesis (Blanchard and Summers, 1986), the model does not support the findings that rigid industrial relations may foster hysteretic behavior in aggregate unemployment. On the contrary, this contribution provides evidence that during severe downturns, and thus declining aggregate demand, phenomena like decreasing investment and innovation rates, skills deterioration, and declining entry dynamics are better candidates to explain long-run unemployment spells and reduced output growth. In that, more rigid labor markets may well dampen hysteretic dynamics by sustaining aggregate demand, thus making the economy more resilient. 1. Introduction In this work, we explore the extent to which the labor-augmented “Schumpeter meeting Keynes” (K + S) model is able to display the endogenous emergence of hysteresis out of the interaction of heterogeneous firms and workers. The article focuses on both the causes and the consequences of the hysteretical properties of macroeconomic time series, including gross domestic product (GDP), productivity, and unemployment. Further, refining upon Dosi et al. (2016b, 2017c), we introduce an endogenous process of accumulation of workers’ skills and a state-dependent process of firms entry, studying their hysteretic effects. As we shall briefly discuss below, there are different notions of hysteresis. Basically, they boil down to three interpretations of the phenomenon (more in Piscitelli et al., 2000, Hallett and Piscitelli, 2002, Amable et al., 2004). The first is formulated in terms of the persistence in the deviations from some equilibrium path; the second is defined as a random-walk dynamics in equilibrium itself; the third, we believe a more genuine one, is in terms of the heterogeneous and nonlinear responses of a system characterized by multiple equilibria or path-dependent trajectories. Even if Piscitelli et al. (2000: 59–60) define the former two as bastard usages of the notion of hysteresis, they have been so far the most common ones in economics. In this work we shall adopt the third notion which encompasses the phenomena of remanence, super-hysteresis, persistence, nonlinearity, and path dependency. Nonetheless, in an archetypical example, Blanchard and Summers (1986) used the second of the foregoing interpretations in an attempt to explain the structural unemployment in the late 1980s in many European countries, at around 10% and quite far from the predicted 2–3% equilibrium level: The recent European experience has led to the development of alternative theories of unemployment embodying the idea that the equilibrium unemployment rate depends on the history of the actual unemployment rate. Such theories may be labelled hysteresis theories after the term in the physical sciences referring to situations where equilibrium is path-dependent (Blanchard and Summers, 1987: 1). Two alternative hypotheses were proposed by these authors to explain the emergence of hysteresis, a first one resting on the membership channel according to which only insider workers are able to exert pressure in the wage setting process, and a second one based on the duration channel because the long-term unemployed workers are less relevant in the wage determination process. In the latter case, unemployment duration can (i) induce a process of worker skills deterioration, implying that the long-term unemployed worker experiences a fall in its productivity; and, (ii) trigger search discouragement in unemployed people, less re-employable, and so less prone to search in the labor market. Together with the supply-side channels emphasized since the 80s, after the current crisis, some acknowledgements have gone to aggregate demand shocks conceived as potential sources of hysteresis. Therefore, the notion of hysteresis has been extended from unemployment to permanent output loss. Blanchard et al. (2015) revisit hysteresis as the permanent effect exerted by crises on the the levels of output relative to the pre-crises ones. That work provides evidence of a persistent output gap in 69% of the cases, among 22 countries in the period 1960–2010, where in 47% of them the recession was followed by an increasing output gap, meaning that recessionary periods affected not only the levels but also the subsequent growth rates, an effect named by Ball (2014) as super-hysteresis. In fact, Ball (2014) reports that over 23 countries in the period 2007–2014, most of them have been hit by severe recessions, and some of them, like Greece, faced up to 30% losses in potential output. […] in most countries the loss of potential output is almost as large as the shortfall of actual output from its pre-crisis trend. This finding implies that hysteresis effects have been very strong during the Great Recession. Second, in the countries hit hardest by the recession, the growth rate of potential output is significantly lower today than it was before 2008. This growth slowdown means that the level of potential output is likely to fall even farther below its pre-crisis trend in the years to come (Ball, 2014: 2). The empirical detection of hysteresis, of course, goes together with the analysis of its determinants. Agent-based models are particularly suitable to the task as one knows by construction the micro data-generating process and thus can explore the possible hysteretic features of aggregate variables as emergent properties of the evolutionary dynamics.1 The model, built upon the “Schumpeter meeting Keynes” family of models (Dosi et al., 2010, Napoletano et al., 2012, Dosi et al., 2013, 2015, 2017a and Dosi et al., 2017c), as we shall see, is able to generically yield hysteresis in the macro variables under scrutiny both inter-regimes and intra-regimes, i.e., across institutional setups governing the labor markets. Indeed, hysteresis is ubiquitous. According to our analysis, hysteresis is not due to market imperfections but rather to the very functioning of decentralized economies characterized by coordination externalities and dynamic increasing returns. Contrary to what suggested by, e.g., Blanchard and Summers (1986), our model does not support the hypothesis that rigid industrial relations, via the insider–outsider channel, are the driving source of hysteresis in aggregate unemployment. On the contrary, more in line with Ball et al. (2014), our results indicate that during severe downturns and thus declining aggregate demand, phenomena like decreasing investment and innovation rates, skills deterioration, and declining entry dynamics are better candidates to explain long-term unemployment spells and reduced output growth. In such a framework, more rigid labor markets, by supporting aggregate demand, do not foster hysteresis but rather may well dampen it, thus making the economy more resilient. The article is organized as follows. Section 2 discusses the nature and the sources of hysteresis. In Section 3, we present the model structure. The empirical regularities matched by the K + S model are discussed in Section 4. In Section 5, we study the emergence and the causes of hysteresis. Finally, Section 6 concludes. 2. The nature and determinants of hysteresis In this section, we provide a brief exploration on the sources and potential channels which might induce hysteretic behaviors in the macroeconomic variables. 2.1 The nature of hysteresis Hysteresis, a concept adopted from the natural sciences but with similar instances in economics, is a nonlinear mechanism, often implying multiple (alternative) time trajectories and equilibria. In a very broad perspective, a dynamical system can be considered hysteretical when the time trajectories of some or all of its variables do exhibit path dependency, in turn also implying non-ergodicity. The very notion of multiple paths for the development of both socioeconomic and natural complex systems ultimately rests on the idea that history is an essential part of the interpretation of many dynamic phenomena. The property that history matters is also intimately related to that of time irreversibility, that is, a situation where it is not possible, even theoretically, to “reverse the arrow of time” and still expects to recover invariant properties of the system under investigation. Reviewing the literature on complex systems is beyond the scope of this article. Suffice to recall the distinction between nonlinear deterministic systems and stochastic ones, both however displaying forms of path dependency. Concerning the former, instantiations are bifurcation, chaotic, and catastrophe dynamics (see Lorenz, 1993). With respect to the latter, Generalized Polya Urns are a well-known example. Both families of processes are often characterized by the presence of tipping points whereby a tip is a threshold point (variable or parameter) which, when reached, might induce irreversible changes on the evolution of the state-space (see Lamberson and Page, 2012 for a detailed discussion in social sciences). On empirical grounds, in tackling path-dependent phenomena in the social sciences (but also in, e.g., biology), an intrinsic difficulty rests in the fact that frequently only one of the many possible realizations of the system, dependent on its initial state, is empirically observed. In that, how much is history-dependence shaped by initial conditions or conversely how does it relate to irreversible effects of some particular unfolding events (e.g., crises or regime changes)? Related, how do the set of all possible evolutionary paths are shaped and constrained by the structure inherited from the past?2 In economics—at least in the dominant theory as distinct from e.g. economic history—the very notion of hysteresis has only been acknowledged with some skepticism and often in the most restrictive interpretations. In the 1980s and 1990s, a stream of literature has faced head-on the challenge of nonlinearity of growth processes and thus the multiplicity of alternative paths and the related hysteretic properties (insightful examples are the contributions in Anderson et al., 1988, Day and Chen, 1993 and Rosser, 2013).3 However, such a stream of investigation was progressively marginalized, possibly due to its “revolutionary” theoretical implications, particularly in terms of equilibria existence, selection, and the associated welfare theorems. Fundamentally, any form of innovation/knowledge accumulation/learning is associated with dynamic increasing returns and thus nonlinearities (Arrow (1996) witnesses from the General Equilibrium side). Illustrative applications of path-dependent stochastic systems to technology diffusion are in David (1985), Arthur (1989), and Dosi and Kaniovski (1994). Finally, an analysis of tipping points in hybrid agent-based models (ABMs) have been performed in Gualdi et al. (2015). A usual “safer” path has been that of formalizing the phenomenon based on linear stochastic models with close-to-unit-root auto-regressive processes. In this perspective, Blanchard and Summers (1986) identify hysteresis in the unemployment series whenever the coefficient of persistence ρ in the equation Ut=ρUt−1+αt+ϵt+θϵt−1 was estimated to be greater or equal to 1. Whether a (close to) unit-root process is an adequate sign of hysteresis has been strongly debated. In general, this modeling approach is based on a somewhat naive epistemology—like “Which processes should present unit-roots? The natural rate of unemployment, the inflation target, or the wage setting curve?”—, but without jeopardizing the underlying unique equilibrium assumption. So, for example, Galí (2015) explores, without conclusive results, three alternative sources to a unit-root process of the European unemployment rate, testing whether it lies (i) in the natural rate of unemployment (Utn=Ut−1n+ϵt), (ii) in the central bank inflation target (πt*=πt−1*+ϵt*), or (iii) in the insider–outsider hypothesis (à la Blanchard–Summers) via alternative specifications for the New Keynesian Wage Phillips Curve. The obvious dissatisfaction with the unit-root process approach is currently yielding a revival of the detection of nonlinearities in empirical macroeconomics. For example, Beaudry et al. (2016), while examining empirical time series like unemployment and working hours, do find evidence of recurrent cyclical patterns, not detectable when estimating auto-regressive linear stochastic models. However, the critique to the unit-root process approach is deeper and concerns its very underlying theory: as suggested by Piscitelli et al. (2000), Hallett and Piscitelli (2002), Amable et al. (2004), and Bassi and Lang (2016), genuine models of hysteresis should embed a nonlinear structure—or at least do not discard nonlinearity in advance. According to Piscitelli et al. (2000), three features characterize hysteretic processes, namely, nonlinearity, selectivity, and remanence. Being this memory process nonlinear, reversing a shock may not drive the system to recover its starting point. Moreover, selectivity means that not all shocks affect the system in the same way in different circumstances. Finally, remanence entails that temporary or non-recurrent shocks may lead to permanent new system states. Widespread origins of hysteresis in the socioeconomic domain are, first, feedback mechanisms related to coordination externalities, and, second, amplification processes stemming from some form of increasing returns.4 In particular, it is frequently associated with (i) positive feedbacks between levels of aggregate activities and innovative search, and (ii) powerful interactions between aggregate demand and diffusion of innovations. Whenever one abandons the unfortunate idea that the macroeconomic system is held up to some mysteriously stable and unique equilibrium path, it could well be, for example, that negative demand shocks exert persistent effects because less aggregate demand entails less innovative search, which in turn yields less innovation stemming from technological shocks: [During recessionary phases], typically firms also reduce their expenditures in R&D and productivity-enhancing expenditures. The reduction in output reduces opportunities to “learn by doing.” Thus, the attempt to pare all unnecessary expenditures may have a concomitant effect on long-run productivity growth. In this view, the loss from a recession may be more than just the large, but temporary, costs of idle and wasted resources: the growth path of the economy may be permanently lowered (Stiglitz, 1994: 122). Despite the 2008 crisis, many economists continue to believe in some version of the model underlying the Example A in Figure 1: the economy is bound to “spring back,” with no permanent loss to its long-run equilibrium rate of growth. The econometric side of this view is the Frisch-like idea of the economy as a “pendulum,” responding to exogenous shocks.5 In this perspective, it seems almost a “miracle” that in the empirical literature one recently finds impulse response functions with multipliers significantly greater than 1. This, we suggest, is a witness of the depth of the current crisis (see Blanchard and Leigh, 2013). Figure 1. View largeDownload slide Effects of recessions: (A) short-run (no hysteresis), (B) long-run (hysteresis), (C) permanent/divergent (super-hysteresis). Source: (A and B): Stiglitz, 1994, p. 123. Figure 1. View largeDownload slide Effects of recessions: (A) short-run (no hysteresis), (B) long-run (hysteresis), (C) permanent/divergent (super-hysteresis). Source: (A and B): Stiglitz, 1994, p. 123. However, a small but significant minority of the profession has been forced by the evidence to accept Case B in Figure 1: recession-induced output losses are permanent because even if the system goes back to the pre-crisis rate of growth, then that is associated with an absolute level gap growing exponentially over time. Moreover, as discussed in Stiglitz (1994), imperfect capital markets and credit rationing may well exacerbate the effects of recessions, hampering the recovery of the growth rate even further. Beyond that, recurrent negative demand shocks, such as those deriving from austerity or labor market flexibilization policies, might yield reduced long-term rates of growth: this is what is shown in Dosi et al. (2016a) and Dosi et al. (2017c). In the latter scenario, as in the Example C in Figure 1, the pre- and post-crisis growth trajectories diverge, implying a reduced long-run rate of the output growth. 2.2 Innovation, diffusion, and investment At the empirical level, a first microeconomic channel6 which might induce hysteresis is the lower innovation rate associated with a reduction in the aggregate demand, which turns out in a decline in the productivity growth. Indeed, R&D expenditures are pro-cyclical. Moreover, the diffusion of new technologies and the adoption of capital-embodied, best-practice techniques slow down during crises. Reifschneider et al. (2015) document a drop in the yearly rate of growth of R&D expenditure in the United States from 3.6% during the pre-crisis period (1990–2007), on average, to 1.6% after 2007. Not only the propensity to innovate but also the process of adoption and diffusion of innovation is slowed down by the contraction of aggregate demand. Both phenomena have been emphasized long ago by Freeman et al. (1982) in their search for the patterns and determinants of long-term fluctuations in growth and employment, and, more recently, theoretically investigated in Dosi et al. (2016a, 2017b). Together with the slower rates of innovation, a process of destruction of the installed productive capacity, due to the lack of sales prospects, seems markedly happening in the post-2008. Indeed, even non-Keynesian commentators have identified the current economic crisis as one stemming from the lack of aggregate demand. As the interest rate reached its zero lower bound without fostering any surge in the investment rate, only accelerator-type investment processes seem able to explain the deteriorating dynamics of the productive capacity. Consistently with the accelerator hypothesis, Kothari et al. (2014) report fresh evidence that investments are ultimately affected by the dynamics of sales, rather than by the interest rate. Overall, lower innovation, diffusion and investment rates seem very plausible candidates to explain the current slowdown in productivity. In turn, the fundamental point is that such changes may well bear a long-term impact, that is, hysteretic effects, on the future dynamics of productivity, GDP, and employment. 2.3 Entry dynamics The second microeconomic channel is the declining entry rate of firms in the market, which has been recently investigated especially in the United States (see Gourio et al., 2016). Entry rates have declined since 2006 by about 27%, a widespread phenomenon across all sectors of the economy. This has been accompanied by steady exit rates and, consequently, also shrinking net entry rates. One direct effect of less entry is the reduced creation of new job opportunities. Decker et al. (2016) document a long-term pattern in the declining business dynamism which the authors attribute, mainly, to the contracting share of young firms. In a similar vein, Siemer (2014) introduces the hypothesis of a missing generation of entrants after the 2008 crisis as a result of the tightening financial constraints, primarily affecting young firms. According to his estimates, the more finance-dependent entrant firms reduced their rates of job creation between 4.8 and 10.5 percentage points relative to the less finance-constrained incumbents. In fact, constrained access to credit may represent an important barrier to entry, together with the usual setup costs, particularly during crises and the associated tight finance availability. Conversely, periods of easy access to debt may induce a higher entry rate (see among others Kerr and Nanda, 2009 and Bertrand et al., 2007). All in all, both in bad and good times, the entry dynamics seems to be a potentially relevant source of hysteresis. 2.4 Skills deterioration A third microeconomic channel which might trigger hysteresis is the workers’ skills deterioration process. Once the economy enters a long recessionary phase, firms tend to fire workers. During severe recessions, like the 2008 crisis, unemployment, which under milder downturns could be in principle temporary and cyclical, turns out to be persistent, implying that many workers experience long unemployment spells. Unemployed workers, of course, stop learning-by-doing, lose contact with the new practices and techniques introduced by firms and gradually deteriorate their existing skills. As the economy recovers and the unemployed are finally hired, their productivity is lower than incumbent workers, reducing the overall productivity. Looking at the recent figures, Reifschneider et al. (2015) document that the share of workers who have been unemployed for more than 26 weeks peaked at 45% in 2011, and it was still about 30% in 2013. On a similar vein, Jaimovich and Siu (2012) analyze the speed of economic recovery during different economic recessions (1970, 1975, 1982, 1991, 2001, 2009) in the United States. Their findings suggest that while in the first three recessions aggregate employment begun to expand within 6 months of the trough of the downturn, during the last three crises employment continued to contract for about 20 months before turning around. Therefore, at the end of 2013 employment had not returned to the pre-crisis level. In turn, Abraham et al. (2016) studying the effect of long-term unemployment on employment probability, and earnings find evidence that long unemployment duration is negatively associated with both job-finding rates and earning opportunities, while Ghayad (2013), based on a résumé review study, reports that employers have a strong rejection for long-term unemployed applicants, even in case of equivalent or superior résumé qualification. Hence, the effects of long unemployment episodes upon skills and job-finding probabilities are yet another important candidate to be a source of macroeconomic hysteresis. 3. The model We build a general disequilibrium, stock-and-flow consistent (see Table B2 in Appendix B), agent-based model, populated by heterogeneous firms and workers who behave according to bounded-rational rules. More specifically, we extend the K + S model (Dosi et al., 2010) with explicitly decentralized interactions among firms and workers in the labor market (Dosi et al., 2016b, 2017c), further adding an endogenous process of workers’ skills accumulation and variable number of firms in each market. The two-sector economy in the model is composed of three populations of heterogeneous agents, Ft1 capital-good firms, Ft2 consumption-good firms, LS consumers/workers, plus a bank and the Government.7 The basic structure of the model is depicted in Figure 2. Capital-good firms invest in R&D and produce heterogeneous machine-tools whose productivity stochastically evolves over time. Consumption-good firms combine machines bought from capital-good firms and labor to produce a homogeneous product for consumers. There is a minimal financial system represented by a single bank that provides credit to firms to finance production and investment plans. Credit is allocated to each firm according to their own demand, which is constrained by their past performance, according to a loan-to-sales cap rule applied by the bank. Conversely, credit supply is completely elastic, adapting to the approved credit demand. Workers submit job applications to a small random subset of firms. Firms hire according to their individual adaptive demand expectations. The government levies taxes on firms’ profits, pays unemployment benefits, and sets minimum wages, according to the policy setting, absorbing excess profits and losses from the bank and keeping a relatively balanced budget in the long run. Figure 2. View largeDownload slide The model structure. Boxes in bold style represent heterogeneous agents’ populations. Figure 2. View largeDownload slide The model structure. Boxes in bold style represent heterogeneous agents’ populations. In the following, we first summarize the functioning of the capital- and the consumption-good sectors of our economy, with a focus on the entry process, and then present the labor market dynamics, detailing the skills accumulation and deterioration mechanisms. Finally, we describe the two alternative policy regime settings under which the model has been explored. In Appendix A, we briefly present the firms’, the workers’, and the Government behavioral rules (for details, see also Dosi et al., 2010 and Dosi et al., 2017c). The model main variables, its configuration, and the parameter setup are presented in Appendix B. 3.1 The capital- and consumption-good sectors The capital-good industry is the locus where innovation is endogenously generated in the model. Capital-good firms develop new machine-embodied techniques or imitate the ones of their competitors to produce and sell more productive and cheaper machinery. On demand, they supply machine-tools to consumption-good firms, producing with labor as the only input.8 The capital-good market is characterized by imperfect information and Schumpeterian competition driven by technological innovation. Machine-tool firms signal the price and productivity of their machines to the current customers as well to a subset of potential new ones, and invest a fraction of past revenues in R&D aimed at searching for new machines or copy existing ones. Prices are set using a fixed markup over (labor) costs of production. Consumption-good firms produce a homogeneous good employing capital (composed by different “vintages” of machines) and labor under constant returns to scale. Desired production is determined according to adaptive (myopic) demand expectations. Given the actual inventories, if the current capital stock is not sufficient to produce the desired output, firms order new machines to expand their installed capacity, paying in advance—drawing on their retained past profits or, up to some limit, on bank credit. Moreover, they replace old machines according to a payback period rule. As new machines embed state-of-the-art technologies, the labor productivity of consumption-good firms increases over time according to the mix of vintages of machines in their capital stocks. Consumption-good firms choose in every period their capital-good supplier comparing the price and the productivity of the machines they are aware of. Firms then fix their prices applying a variable markup rule on their production costs, trying to balance profit margins and market shares. More specifically, firms increase their markup and price whenever their market share is expanding and vice versa. Imperfect information is also the normal state of the consumption-good market, so consumers do not instantaneously switch to the most competitive (cheaper) producer. Market shares evolve according to a (quasi) replicator dynamics: more competitive firms expand, while firms with relatively lower competitiveness levels shrink, or exit the market.9 3.2 The entry and exit processes We expanded the earlier K + S model to account for a variable number of firms in both the consumption- and the capital-good sectors ( Ft1, Ft2). In this new version, entry and exit are now independent processes. As before, firms leave the market whenever their market shares get close to 0 or their net assets turn negative (bankruptcy). However, we now define the number of entrants by means of the random variables bt1 and bt2: btz=Ft−1z[(1−o)MAtz+oπtz]  (lower bounded to 0), (1) where z∈{1,2} denotes the sector (capital- or consumption-good, respectively), Ft−1z is the existing number of incumbent firms, MAtz the “financial attractiveness” of the industry, 1≤o≤1 is a mix balance parameter, and πtz is a random draw from a uniform distribution on the fixed support [x¯2,x¯2]. The number of entrants stochastically depends on the number of incumbents (recalling a spin-off process of the former from the latter) with the financial conditions influencing the decision of would be entrants. The industry-specific “financial attractiveness” MAtz in period t is defined as: MAtz=MCtz−MCt−1z  (bounded to [x¯2,x¯2]). (2) MCtz is calculated on firms’ balance sheets as the (log) ratio between the aggregate stocks of liquid assets NWy,t (bank deposits) and bank debt Deby,t: MCtz= log  (∑yNWy,t−1)− log  (∑yDeby,t−1), (3) in each sector, y∈{i,j}, accordingly. So, MCtz measures the sectoral liquidity-to-debt ratio and thus the tightness of the credit market, and MAtz is a proxy to its dynamics. Correspondingly, negative (positive) values of MAtz represent leveraged (deleveraged) markets, meaning that debt is growing faster (slower) than the accumulation of cash equivalents. This means that whenever the overall liquidity-to-debt ratio is shrinking would-be firms are more inclined to enter, and vice versa. The adopted formulation for the entry process tries to model some well-known facts in the industrial dynamics and business cycle literature: (i) the number of entrants is roughly proportional to the number of incumbent firms (Geroski, 1991, 1995), (ii) entry is affected by the easiness of access to credit (Kerr and Nanda, 2009; Bertrand et al., 2007), and (iii) the process is pro-cyclical (Gomis and Khatiwada, 2017; Lee and Mukoyama, 2015). 3.3 The labor market and skills dynamics The labor market in the model implements a fully decentralized search and hiring process between workers and firms (more on that in Dosi et al., 2016b, 2017c). The aggregate supply of labor LS is fixed and all workers are available to be hired in any period. Moreover, also the labor market is characterized by imperfect information. When unemployed, workers submit a certain number of job applications to firms. Employed workers may apply or not for better positions, according to the institutional setup (see Section 3.5 below). Larger firms, in terms of market share, have a proportionally higher probability of receiving job applications, which are organized in separated, firm-specific queues. Firms decide about their individual labor demand based on the received orders (capital-good sector), the expected demand (consumption-good sector), and the expected labor productivity levels. Considering the number and the productivity of the already employed workers, firms decide to (i) hire new workers, (ii) fire part of the existing ones, or (iii) keep the existing labor force. Each hiring firm defines a unique wage offer for the applicant workers, based on its internal conditions and the received applications. Workers select the best offer they get from the firms to which they submitted applications, if any. If already employed (depending on the institutional regime), they quit the current job if a better wage offer is received. There is no second round of bargaining between workers and firms in the same period, and, so, firms have no guarantee of fulfilling all the open positions (no market clearing). Moreover, there is no firing or hiring transaction costs. We extended the K + S model to account for the process of workers’ skills accumulation and deterioration. Such a process is driven by the worker-specific job tenures, assuming learning-by-doing mechanism when employed and a gradual deterioration of skills while unemployed, and assuming firms keep introducing new techniques all the time, deprecating the skills of unemployed workers. The skill level sℓ,t>0 of each worker ℓ evolves over time as a multiplicative process: sℓ,t={(1+τ)sℓ,t−1 if employed in t−111+τsℓ,t−1 if unemployed in t−1, (4) with the learning rate τ≥0 a parameter. As a consequence, when worker ℓ is employed her skills improve over time, as she becomes more experienced in her task. Conversely, unemployed workers lose skills. In particular, when a worker is hired, she may immediately acquire the minimum level of skills already present in the firm (the existing worker with the lowest skills), if above her present level. Also, workers have a fixed working life. After a fixed number of periods Tr∈ℕ* in the labor market, workers retire and are replaced by younger ones,10 whose skills are equivalent to the current minimum level in the incumbent firms. Workers’ skills define their individual (potential) productivity Aℓ,t: Aℓ,t=sℓ,ts¯tAiτ,  s¯t=1LS∑ℓsℓ,t, (5) where s¯t is the average worker skills level, and Aiτ is the warranted productivity of the machinery vintage the worker operates. The ratio sℓ,t/s¯t, or the worker normalized productivity, represents her ability to produce more (if sℓ,t>s¯t) or less (otherwise) when using a certain machine technology, in relation to the warranted vintage productivity. Note that the sectoral aggregation over the firm-level effective productivities Aj,t is a truly emergent properties of the model, resulting, simultaneously, from the technical innovation dynamics (mainly, the introduction of new vintages Aiτ), the worker skills accumulation/deterioration process, and the effective demand, which guides firms when deciding Qj,td, the capital stock dynamics, and the employed machine mix (see Appendix A for more details). The influence of the workers’ skills upon production reflects a learning by tenure/doing mechanism well established in the literature at least since the seminal contribution of Arrow (1962). On the empirical side, for the links between job tenure, capability accumulation, and firm productivity, see Zhou et al. (2011) and Lucidi and Kleinknecht (2009), among others. 3.4 Time line of events In each simulation time step, which can be taken to roughly represent a quarter, firms and workers behavioral rules are applied according to the following time line: machines ordered in the previous period (if any) are delivered; capital-good firms perform R&D and signal their machines to consumption-good firms; consumption-good firms decide on how much to produce, invest, and hire/fire; to fulfill production and investment plans, firms allocate cash flows, and (if needed) borrow from bank; firms send/receive machine-tool orders for the next period (if applicable); firms open job queues and job seekers send applications to them (“queue”); wages are set (indexation or bargaining) and job vacancies are partly or totally filled; workers (employed and unemployed) update their skills; government collects taxes and pays unemployment subsidies; consumption-good market opens and the market shares evolve according to competitiveness; firms in both sectors compute their profits, pay wages, and repay debt; exit takes place, firms with near-zero market share, or negative net assets are eschewed from the market; prospective entrants decide to enter according to the markets conditions; and aggregate variables are computed and the cycle restarts. 3.5 Alternative labor market policy regimes We employ the model described above to study two alternative policy regimes, which we call Fordist (our baseline) and Competitive.11 The policy regimes are telegraphically sketched in Table 1 (see Table B3 in Appendix B for the regime-specific parameter values). Table 1. Main characteristics of tested policy regimes Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Table 1. Main characteristics of tested policy regimes Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Under the Fordist regime, wages are insensitive to the labor market conditions and indexed on a convex combination between economy-wide and firm-level productivity growth. There is a sort of covenant between firms and workers concerning “long term” employment: firms fire only when their profits become negative, while workers are loyal to employers and do not seek for alternative jobs. When hiring/firing, firms aim to keep the more skilled worker. Labor market institutions contemplate a minimum wage fully indexed to the aggregate economy productivity and unemployment benefits financed by taxes on profits. Conversely, in the Competitive regime, flexible wages respond to unemployment in a decentralized labor market dynamics, and are set by means of an asymmetric bargaining process where firms have the last say. Employed workers search for better paid jobs with some positive probability and firms freely adjust (fire) their excess workforce according to their planned production. Hiring/firing workers by firms are based on a trade-off between skills and wages, using a simple payback comparison rule. The Competitive regime is also characterized by different labor institutions: minimum wage is only partially indexed to productivity, and unemployment benefits—and the associated taxes on profits—are relatively lower. The simulation exercises in Section 5 are built so that there is a regime transition at a certain time step, capturing a set of labor market “structural reforms.” This institutional shock is aimed at spurring flexibility on the relations among agents in the labor market and implies that the social compromise embodied in the Fordist regime is replaced by the Competitive one. 4. Empirical validation The K + S model is able to generate endogenous growth and business cycles, emergent crises, and to reproduce a rich set of macro (e.g., relative volatility, co-movements, etc.) and micro (firm size distributions, firm productivity dynamics, etc.) stylized facts (see Dosi et al., 2010, 2013, 2015, 2017b). The detailed list of empirical regularities matched by the model is reported in Table 2. In addition, the labor-enhanced version of the model (Dosi et al., 2016b, 2017c), which explicitly accounts for microeconomic firm–worker interactions, has already proved to be able to robustly reproduce most of the labor market macro empirical regularities (cf. the bottom part of Table 2).12 Table 2. Stylized facts matched by the K + S model at different aggregation levels Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function Table 2. Stylized facts matched by the K + S model at different aggregation levels Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function The extensions to the K + S model proposed here add some new empirical regularities matched by the model. First, the new labor force learning dynamics produces fat-tailed worker-level skill distributions and firm-level productivity ones, consistent with the empirical evidence suggesting the presence of both firm- and worker-specific ample heterogeneity. Second, the more realistic entry dynamics increases the number of variables that match the cross-correlation/lag structures among aggregate macro indicators. Table 3 shows the time series correlation structure of firms net entry (entries minus exits) in the market, total bank debt, and average liquidity-to-sales ratios with respect to the GDP. These and the other macro variables correlation structures (not presented) seem reasonably in line with empirical evidence, as detailed in Dosi et al. (2010). Table 3. Correlation structure with respect to GDP on selected variables Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Note: All results significant at 5% level. MC standard errors in parentheses. Non-rate series are Baxter–King bandpass-filtered (6, 32, 12). Table 3. Correlation structure with respect to GDP on selected variables Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Note: All results significant at 5% level. MC standard errors in parentheses. Non-rate series are Baxter–King bandpass-filtered (6, 32, 12). 5. At the roots of hysteresis Let us study the emergence of hysteresis in our model, addressing its possible causes and discussing its consequences for the economic dynamics. We will first study inter-regime long-run hysteresis (cf. Figure 1). We will then analyze the emergence of intra-regime (transient) hysteresis (Section 5.2). 5.1 Regime change: super-hysteresis We begin with the long-run dynamics of the model, when affected by an institutional shock, namely, the introduction of “structural reforms” aimed at increasing the flexibility of the labor market, leaving however untouched the technological fundamentals. In our policy typology, the reforms are supposed to move the labor market regime from a Fordist to a Competitve setup (see Section 3.5 above). In that, we are also implicitly testing the insider–outsider hypothesis of hysteresis proposed by Blanchard and Summers (1987). In our model, the transition from a Fordist toward a Competitive type of labor relations captures the structural reforms, aimed at achieving both numerical (easier firing) and wage flexibility (wages more respondent to unemployment), as illustrated in Table 1.13 The normative implication of such hypothesis is the advocacy of a more flexible labor market, where unions have lower bargaining power in the wage formation process, with the aim of making wages more respondent to unemployment conditions. In Figure 3, we report the time series of the main macroeconomic variables in the two regimes.14 The institutional shock occurs at time t = 100 (the vertical dotted line). The widening GDP gap between the two regimes, as presented in Figure 3A, shows how the structural reforms determine super-hysteresis (i.e., a permanently lower growth rate of the GDP), whereby the effects propagate in the very long run (see also Dosi et al., 2017c, 2016b). The actual level of the long-run capacity utilization increases from the 85% to 90% after the introduction of the Competitive regime (cf. Figure 3B), hinting at a process of underinvestment due to the steeper fluctuations in investments opportunities for firms. In the Competitive regime, as a result of the more flexible wage dynamics, increased GDP volatility, and their effect on the aggregate demand, firms reduce their average expansionary investments, which depend on the difference between (demand-led) desired and installed production capacity (see Equation (19) in Appendix A), pushing down the number of machines ordered from the capital-good sector. Therefore, firms decrease the gap between the effective production and the potential capacity, leading to a cyclical surge in the capacity utilization which tends to yield self-rationing. Capital accumulation is slower when structural reforms are in place: the long-run growth rate falls from 1.55% to 1.44% per period. Figure 3C shows the dynamics of unemployment and vacancy rates, which are negatively correlated, consistent with a Beveridge Curve, while unemployment is significantly higher in the Competitive regime. The negative effects of structural reforms spill over the long-run: the number of successful innovations in the capital-good sector takes a lower trajectory (Figure 3D), and the average level of workers skills is significantly reduced (Figure 3E). Finally, the trend of the net entry15 of firms in the market is more turbulent after the reforms, also as a consequence of a higher level of volatility in credit conditions (Figure 3F).16 Figure 3. View largeDownload slide Macroeconomic dynamics in alternate policy regimes. Lines represent 50 MC runs averages (Fordist: black; Competitive: blue/gray). Figure 3. View largeDownload slide Macroeconomic dynamics in alternate policy regimes. Lines represent 50 MC runs averages (Fordist: black; Competitive: blue/gray). The different performance of the two regimes is quantitatively summarized in Table 4, which presents the averages, the ratios between selected variables of the two setups, and the P-values for a t test comparing the averages. The results confirm, at a 5% significance level, that after the introduction of structural reforms the short- and long-run performance of the economy significantly worsens. Note that as the technological configuration of the model is invariant between the two regime specifications, the differences in terms of productivity, innovation, and imitation rates are entirely caused by the institutional shock.17 Table 4. Comparison between policy regimes Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Note: Averages for 50 MC runs in period [200,400] (excluding warm-up). P-value for a two-means t-test, H0: no difference between regimes. Table 4. Comparison between policy regimes Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Note: Averages for 50 MC runs in period [200,400] (excluding warm-up). P-value for a two-means t-test, H0: no difference between regimes. What are the drivers of the soaring super-hysteresis in the model? The huge surge in unemployment reflects the widening gap between the long-run dynamics of real wages in the two regimes,18 which, in turn, leads to the emergence of Keynesian unemployment due to the contraction of aggregate demand, the slowdown in skills accumulation and actual productivity growth. Figure 4 shows the box-plot comparison between the Monte Carlo simulation runs for the two regimes, for the long-term consequences in terms of the innovation and imitation rates, productivity growth, job tenure, workers’ skills, and net entry of firms (see Section 2). The results in the first row of plots (Figure 4A–C) indicate a reduction in the innovation and imitation rates in the majority of the simulation runs—the latter variables are calculated as the rate of successful innovators and imitators in the capital-good sector—and, as a consequence, in the productivity growth rate. This is an indirect outcome of the fall in the aggregate demand, which yields lower R&D expenditure by firms.19 In the same direction, the results in the second row of Figure 4 show the quite significant fall on the average tenure period (Plot d) and the ensuing slower pace of the workers skills accumulation (Plot e), which, in turn, also has a direct and negative effect on the growth of productivity. Finally, the dynamics of net entry (number of entrants minus the exiting firms) is presented in Plot f.20 In the Competitive regime the financial cycle is amplified due to the increased volatility, exacerbating the entry dynamics: in good times there are more entrants in Competitive than in the Fordist regime, which exhibits a stabler financial cycle, while the opposite occurs in bad times. Both emergent phenomena, i.e. the more pronounced leverage cycle and the tighter availability of credit, have been empirically documented by Ng and Wright (2013) from the last three recessions (1990, 2001, and 2007). Figure 4. View largeDownload slide Performance comparison between policy regimes (Fordist: left; Competitive: right). Summary statistics for 50 MC runs in period [200,400] (excluding warm-up). Bar: median; box: second to third quartile; whiskers: maximum–minimum; dots: outliers. Figure 4. View largeDownload slide Performance comparison between policy regimes (Fordist: left; Competitive: right). Summary statistics for 50 MC runs in period [200,400] (excluding warm-up). Bar: median; box: second to third quartile; whiskers: maximum–minimum; dots: outliers. The transmission channels in the model operate through both numerical and wage flexibility. First, higher numerical flexibility, where workers are more freely fired, determines a sharp drop in workers job tenure and, indirectly, has a negative effect on skills accumulation and, consequently, on productivity. Not only the firing rule but also the firing order criteria affect the dynamics of productivity growth. In the Fordist regime, firms first hire (fire) workers with higher (lower) skills.21 Conversely, in the Competitive case, firms use the skills-to-wage “payback” ratio as a decision guide to preferentially hire (fire) workers with superior (inferior) short-term “returns.” Such a behavior has a negative impact on the aggregate skill level of the incumbent workers over time. On the other hand, higher wage flexibility, by limiting the wage indexation upon the productivity gains, causes a straightforward drop in the aggregate demand via the reduced consumption of workers. In turn, the shrinking sales opportunities drive a fall in investment and labor demand, which induces more unemployment, characterizing a typical Keynesian feedback-amplified downturn. Moreover, the slower economy also impacts upon the entry/exit and the innovation/imitation rates, via the overall cut in total R&D expenditure and the higher volatility in the number of operating firms. In fact, Table 3 shows the significant level of correlation between the business cycle and the net entry of firms in the market. The severe effects of super-hysteresis are particularly well illustrated by the probability distributions for the time unemployed workers need to find a new job, presented in Figure 5.22 As shown by the huge increase in the distribution support, long-term unemployment is by far higher in the Competitive case.23 Figure 5. View largeDownload slide Actual probability distribution vs normal fit of worker unemployment time. Each t corresponds to a quarter. Figure 5. View largeDownload slide Actual probability distribution vs normal fit of worker unemployment time. Each t corresponds to a quarter. To sum up, our experiments generically yield super-hysteresis stemming from an institutional shock. Indeed, institutions are a “carrier of history” (David, 1994) also here. However, contrary to the insider–outsider hypothesis (Blanchard and Summers, 1987), “pro-market” institutions bear a negative hysteretic effect. The model suggests that structural reforms aimed at increasing the flexibility in the labor market may well spur even more hysteresis instead of reducing it. Granted that, in the next section, we focus on intra-regime hysteresis phenomena. 5.2 Detecting intra-regime hysteresis Assessing the emergence of intra-regime hysteresis is not a trivial task, as there is no conclusive test or even widely accepted criteria for this. However, there are several properties and techniques which do help uncover particular aspects of hysteresis. In the following, we present a set of analytical methods, summarized in Table 5, which provide evidence of the presence of hysteretical properties in the K + S model. In line with the literature, we study whether the time series generated by the model present evidence of (i) remanence, (ii) persistency, (iii) nonlinearity, (iv) path dependency, and (v) super-hysteresis. Needless to say, these properties are to some degree overlapping. Table 5. Selected tests to evaluate hysteretic properties in times series Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Table 5. Selected tests to evaluate hysteretic properties in times series Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Figure 6 illustrates the number of periods (gray area) necessary to put the economy back to the pre-crisis growth trend (dashed line) in typical simulation runs.24 The analysis is inspired by previous work (Blanchard et al., 2015) and simply performs an extrapolation of the long-run GDP trend to detect the recovery from crises under the possible presence of hysteresis. The results show the coexistence of shorter business cycle downturns with longer, hysteretical crises, requiring significant more times for the economy to recover. Note also the presence of super-hysteresis, revealed by the different slopes of the peak-to-peak GDP trends (dashed lines). Figure 6. View largeDownload slide GDP recovery after crises. Typical simulation runs. Dashed line: pre-crisis trends; gray boxes: recovery periods. Figure 6. View largeDownload slide GDP recovery after crises. Typical simulation runs. Dashed line: pre-crisis trends; gray boxes: recovery periods. Table 6 reports the average recovery duration for both the GDP and the mean unemployment time (the average period a worker takes to find a new job). While the duration of the GDP trend recovery is similar among regimes (around 16 quarters), the unemployment time takes almost five times more to return to its pre-crisis level in the Competitive case. To better assess the severity of the crises, we also track the peak GDP trend deviation during the recovery period (the farther the GDP gets from the pre-crisis trend) and the accumulated GDP losses in comparison to the trend (the crisis “cost”). The model robustly shows how Competitive regime crises are about twice deeper than in the Fordist scenario. The accumulated GDP loss comparison leads to a similar conclusion. Table 6. Comparison between policy regimes: crises, GDP and unemployment time recovery Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Note: Averages for 50 MC runs in period [200,400] (excluding warm-up), MC standard errors in parentheses. Table 6. Comparison between policy regimes: crises, GDP and unemployment time recovery Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Note: Averages for 50 MC runs in period [200,400] (excluding warm-up), MC standard errors in parentheses. In Table 7, we report a set of statistical tests to detect unit-roots vs. stationarity (Augmented Dickey–Fuller/ADF, Phillips–Perron/PP, and Kwiatkowski–Phillips–Schmidt–Shin/KPSS tests), i.i.d. vs. nonlinear processes (Brock–Dechert–Scheinkman/BDS test), and ergodicity (Kolmogorov–Smirnov/KS and Wald–Wolfowitz/WW tests). Except for the WW case, the tests are applied for individual Monte Carlo simulation runs (or multiple run-pair combinations, in the case of KS) and, so, the results present the frequency of the rejection of the null hypothesis for the set of 50 runs at the usual 5% significance level (see Table 7 for the definition of H0 in each case). Table 7. Comparison between policy regimes: statistical tests for detecting hysteresis Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Note: Frequencies of rejection of H0 for 50 MC runs in period [300,350] (excluding warm-up) except for WW test (P-value presented), at 5% significance. ADF/PP H0: non-stationary—KPSS) H0: stationary—BDS H0: i.i.d., KS/WW H0: ergodic. Table 7. Comparison between policy regimes: statistical tests for detecting hysteresis Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Note: Frequencies of rejection of H0 for 50 MC runs in period [300,350] (excluding warm-up) except for WW test (P-value presented), at 5% significance. ADF/PP H0: non-stationary—KPSS) H0: stationary—BDS H0: i.i.d., KS/WW H0: ergodic. The results suggest that GDP, productivity, and wage growth rates more frequently exhibit stationary (no unit-roots) behaviors in both regimes. More borderline, the unemployment rate time series seems to be more commonly stationary among simulation runs in the Fordist regime, while more likely non-stationary in the Competitive case. The nonlinearity test indicates a quite nuanced situation: the unemployment series is the one more frequently displaying nonlinear structure, particularly in the Competitive regime, while the wage growth rates series seem more likely to be i.i.d. processes. Finally, the less powerful KS test cannot reject ergodicity for the majority of run pairs tested, while WW indicates non-ergodicity of all series. There are a few take-home messages from the tests. First, that mixed results, e.g., on ergodicity and stationarity, militate as such in favor of path dependency. In fact, they show the different statistical properties of alternative sample-paths: only an outright non-rejection of the null hypothesis could be claimed in support of the the lack of hysteresis. Second, but related, the tests aimed at the detection of some underlying, emergent, nonlinear structure are quite encouraging despite the limited length of the sample paths.25 Finally, we perform a global sensitivity analysis (SA) to explore the effects of alternative model parametrizations and to gain further insights on the robustness of our exercises on institutional shocks.26 Of the 57 parameters and initial conditions in the K + S model, we reduce the relevant parametric dimensionality to 29, by means of an Elementary Effect screening procedure which allows to discard from the analysis the parameters which do not significantly affect the selected model outputs.27 All the parameters tested in the SA, their “calibration” values, as well the tests statistics, are detailed in Table B1 (Appendix B). To understand the effect of each of the 29 parameters over the selected metrics, we perform a Sobol decomposition.28 Because of the relatively high computational costs to produce the decomposition using the original model, a simplified version of it—a meta-model—was built using the Kriging method and employed for the Sobol SA.29 The meta-model is estimated by numerical maximum likelihood using a set of observations (from the original model) sampled using a high-efficiency, nearly orthogonal Latin hypercube design of experiments (Cioppa and Lucas, 2007). The main indicator used for the SA is the accumulated GDP losses during the crises’ recovery periods (the crisis “cost”), as defined above. It seems a sensible choice, as it conveys information about both the duration and the intensity of the crises, as such among the key properties of hysteresis. Interestingly, this indicator is significantly influenced only by a limited set of parameters, and by no initial condition, including the learning rate parameter (τ), the retirement age (Tr), the replicator equation parameter (χ), the maximum technical advantage of the capital-good entrants (x5), the minimum capital ratio ( Φ1) and the expected capacity utilization (u) of the consumption-good entrants.30 The two parameters associated with the skills accumulation process, learning rate (τ) and retirement age (Tr), are jointly responsible for almost 80% of the variance of the losses indicator over the entire parametric space in both policy regimes.31 Figure 7A and B presents an exploration of the model response surface, using the Kriging meta-model, for the two critical skill-related parameters. The rugged surfaces, in particular in the Competitive regime, clearly indicate the nonlinear nature of the system, in tune with the hysteretic properties of the model. The SA seems to suggest that the prominent parameters influencing the level of hysteresis observed in the losses indicator are those directly connected with the workers skills accumulation process (τ and Tr), the firm entry mechanism ( Φ1, u and x5), and the market competitiveness (χ). Directly or in interaction among them, these five parameters account for 95% of the variation of the GDP crises losses in the model for the two scenarios. As can be seen in Figure 7, the Competitive regime tends to produce significant higher GDP crises losses irrespective of the model setup (notice that the peak losses in Plot (a), are at a lower z-axis level than the deepest valley in Plot (b)). Finally, the response surfaces in both regimes show that in general the higher the learning rate (τ), the higher is the accumulated GDP losses during the crises’ recovery periods. The latter positive marginal effect hints at the fact that the higher the firm-specific capabilities, the more difficult is to rebuild the workers skills destroyed by a crisis, turning back to the pre-crisis level. Figure 7. View largeDownload slide Global SA: response surfaces. Surfaces modelled using the Kriging meta-model. z-axis: recovery losses (rec_loss). Dot: calibration settings. Markers: maximum and minimum predicted crises losses. Figure 7. View largeDownload slide Global SA: response surfaces. Surfaces modelled using the Kriging meta-model. z-axis: recovery losses (rec_loss). Dot: calibration settings. Markers: maximum and minimum predicted crises losses. All in all, the statistical tests results indicate that model has a rather frequent tendency to show the properties usually associated with hysteresis in its main variables, in particular the unemployment rate, whenever hit by an endogenously produced crisis. Recoveries can take quite long times, and the losses experienced by the economy, both in terms of the GDP and the social cost of unemployment, are severe. It is also significant that such losses seem to increase after the introduction of structural reforms of the type discussed above. 6. Conclusions The Great Recession has forced a revival of the notion of hysteresis, as such a short-hand for the possibility of multiple equilibria/paths either in some transient periods or even in the longer term. The evidence has been overwhelming: not only the level trends of GDP and unemployment but even the growth rates in many countries are still persistently below the pre-2008 figures. Older candidates for the interpretation of such a behavior are unit-root processes in unemployment—as originally suggested by Blanchard and Summers (1986). But such interpretations are rather fragile in that they postulate the source of hysteresis in some deviations of reality from the standard frictionless model, e.g. the insider–outsider labor market rigidities. Here we have analyzed an opposite perspective. In tune with an expanding tradition of scholars, we have discussed the notions of hysteresis and path dependence, identifying in coordination failures and persistent effects of aggregate demand upon productivity the main sources of long-term jumps across multiple growth trajectories. In doing that, we have presented an ABM which intertwines a Schumpeterian engine of growth and a Keynesian generation of demand, declined under two institutional labor market variants, labeled as Fordist and Competitive regimes. The transition from the Fordist to the Competitive regime captured “structural reforms” aimed at increasing labor market flexibility. Does the latter reduce hysteresis? Not at all. The model is able to generically exhibit path dependence, nonlinearity, and non-ergodicity in its main macroeconomic variables, presenting both inter-regime and intra-regime hysteresis as a bottom-up emergent property. Moreover, the model suggests that both numerical and wage flexibility are quite prone to increase the hysteretic properties of the macroeconomic system. The K + S model leaves scope for many potential avenues for further research, addressing the links between the functioning of the capital, consumer, and labor markets. In particular, a straightforward extension of the current article would be the study of the effects of active labor market policies, declined under alternative training programs and hiring/firing schemes. Yet, another venue of research concerns the effect of hysteresis upon labor force participation. Acknowledgements The authors thank the Guest Editors of the Special Section; two anonymous referees, Engelbert Stockhammer and Federico Tamagni; and the participants to the International Conference Economics, Economic Policy and Sustainable Growth after the Crisis (Ancona, 2016), the 9th Workshop on Complex Evolving System Approach in Economics (Nice, 2017), the Workshop on Economic Growth, Macroeconomic Dynamics, and Agents Heterogeneity (Saint-Petersburg, 2017), the 10th EMAEE Conference (Strasbourg, 2017), the 22nd WEHIA (Milan, 2017), the 23rd CEF International Conference (New York, 2017), the 29th SASE Annual Meeting (Lyon, 2017), and the 29th Annual EAEPE Conference (Budapest, 2017). G.D., A.R., and M.E.V. gratefully acknowledge the support by the European Union’s Horizon 2020 research and innovation program under grant agreement number 649186 - ISIGrowth. Footnotes 1 See the presentations in Tesfatsion and Judd (2006); LeBaron and Tesfatsion (2008), and Fagiolo and Roventini (2012, 2017) for critical surveys on macro agent-based models (ABMs). For related ABMs which consider a decentralized labor market, see Dawid et al. (2014), Russo et al. (2016), Caiani et al. (2016a), and Caiani et al. (2016b), among the others. See also Bassi and Lang (2016) for an agent-based model with investment hysteresis. 2 For a more detailed discussion on the relationship between hysteresis and path dependency, see Castaldi and Dosi (2006); Setterfield (2009). 3 The hysteretic properties of economic systems are also studied in the Post Keynesian literature: see Skott (2005); Stockhammer (2011) which identify wage norms and fairness as a potential mechanism to produce an endogenous non-accelerating inflation rate of unemployment (NAIRU). 4 See Dosi and Virgillito (2018) for a further discussion. 5 For an enticing reconstruction of the discussion between Frisch and Schumpeter on the pendulum metaphor, see Louca (2001). 6 The order the channels are presented is not relevant in terms of the impact produced by each one upon hysteresis. 7 The subscript t indicates time dependence. From now on, agent-specific variables are denoted by a subscript i, in case of capital-good firms, j, for consumption-good firms, or ℓ, for workers. 8 The latter is a usual simplifying assumption which avoids the introduction of a multilevel capital-goods sector and keeps the innovation process more transparent (see, on a three-sector economy, Seppecher et al., 2017). 9 In the consumption-good market the replicator captures the law of motion which regulates the pace of adjustments in market shares among the ever-changing heterogeneous competitors. On the robustness of the empirical properties generated by the replicator dynamics, see Dosi et al. (2017e). 10 In the start of the simulation, initial workers ages are randomly drawn in the integer range [1,Tr] and all start from the same skills level. 11 The two regimes roughly capture two alternative wage-labor nexus in the language of the Regulation Theory (see, within a vast literature, Boyer and Saillard, 2005 and Amable, 2003). 12 For a detailed discussion upon the configurations and the parameter settings producing the above mentioned stylized facts, we refer to Dosi et al. (2010, 2017c). In the following we focus on the innovation, entry, and skills processes, and on the related variables and parameters. 13 Indeed, the change of the political structure and of the balance of power between capitalists and workers and the related results of a class struggle are phenomena which, while with profound economic roots, did occur at the sociopolitical level: the Thatcher–Reagan regime change has been an exogenous political transformation. Modeling the triggering mechanisms leading to the end of the welfare system is thus well beyond the scope of this article. 14 The presented series are the averages of 50 Monte Carlo simulation runs, over 500 periods. The initial 100 “warm-up” periods are not presented. 15 Note that the use of the two-sided Hodrick-Prescott (HP) filter artificially produces the diverging patterns of the two curves before time 100. 16 As discussed in Section 3.2, entry decision in the model is also driven by the average financial conditions of the firms in each sector. 17 In accordance with the behavioral rules set in the model (cf. Appendix A), the dynamics of innovation, of imitation, of new machines introduction, and, consequently, of the firms productivity growth is directly affected by the overall macroeconomic conditions, including those directly impacted by the reforms. This creates a (potentially hysteretic) reinforcing feedback process between the macro and the technological domains, which in part explains the observed results. 18 The real wage growth rates are 1.47% and 1.35% per period, respectively. 19 See Appendix A for details on the innovation process. 20 The diverging trend before time t = 100 is due to the two-sided HP filter we employ to detrend the series. 21 This is a necessary consequence of the firms unilaterally decided and homogeneously applied wage adjustments, so skills are the only heterogeneous metric among workers in a Fordist firm. 22 Note the log scale in the y-axis. 23 The maximum notional unemployment time is 120 periods, equivalent to the working life in the model (parameter Tr). 24 A crisis is defined by a 3% drop of the GDP in a single period which is not recovered in the next three periods. The pre-crisis level is calculated as the average GDP for the four periods before the crisis and the trend, and as the output of an HP filter at the period just before the crisis. The crisis is considered recovered when the GDP reaches back the pre-crisis trend level. 25 The choice of the adequate time window length is quite relevant when analyzing hysteresis, as detailed in Section 2.1, and it is not driven by the availability of simulated data. For this reason we split the analysis in inter-regime hysteresis, where the patterns are of long-term type, and the intra-regime hysteresis. The analysis used in the literature to detect hysteresis is always performed for relatively short time spans (usually under 20 years). For comparability with empirical data, to check for the intra-regime hysteresis we restricted the time span to 50 runs, which is closer to the empirical time horizons. Note that taking longer time spans would simply “dilute” some hysteretic properties of the series, like non-ergodicity or non-stationarity. 26 For technical details on the global SA methodology, see Dosi et al. (2017d). 27 Briefly, the Elementary Effects technique proposes both a specific design of experiments, to efficiently sample the parameter space under a one-factor-at-a-time, and some linear regression statistics, to evaluate direct and indirect (nonlinear/non-additive) effects of parameters on the model results (Morris, 1991, Saltelli et al., 2008). 28 The Sobol decomposition is a variance-based, global SA method consisting in the decomposition of the variance of the chosen model output into fractions according to the variances of the parameters selected for analysis, better dealing with nonlinearities and non-additive interactions than traditional local SA methods. It allows to disentangle both direct and interaction quantitative effects of the parameters on the chosen metrics (Sobol, 1993, Saltelli et al., 2008). 29 Summarizing, the Kriging meta-model “mimics” our original model by a simpler, mathematically tractable approximation. Kriging is an interpolation method that under fairly general assumptions provides the best linear unbiased predictors for the response of complex, nonlinear computer simulation models (Rasmussen and Williams, 2006, Salle and Yıldızoğlu, 2014). 30 All the equations and parameters are described in Sections 3.1 and 3.3 and in the Appendices. 31 The parameters calibration values, valid ranges, and the Sobol decomposition results are presented in Table B1 in Appendix B. References Abraham K. G. , Haltiwanger J. C. , Sandusky K. , Spletzer J. ( 2016 ), ‘The consequences of long term unemployment: Evidence from matched employer-employee data,’ NBER Working Paper 22665. National Bureau of Economic Research: Cambridge (MA), US . Amable B. ( 2003 ), The Diversity of Modern Capitalism . Oxford University Press : Oxford . Google Scholar CrossRef Search ADS Amable B. , Henry J. , Lordon F. , Topol R. ( 2004 ), ‘Complex remanence vs. simple persistence: are hysteresis and unit-root processes observationally equivalent?,’ in Barnett W. , Deissenberger C. , Feichtinger G. , (eds), Economic Complexity . Emerald Group : Bingley, UK , pp. 67 – 89 . Google Scholar CrossRef Search ADS Anderson P. , Arrow K. , Pines D. ( 1988 ), The Economy as a Complex Evolving System . Addison-Wesley : Boston, US . Arrow K. ( 1996 ), ‘ Technical information and industrial structure ,’ Industrial and Corporate Change , 5 ( 2 ), 645 – 652 . Google Scholar CrossRef Search ADS Arrow K. J. ( 1962 ), ‘ The economic implications of learning by doing ,’ The Review of Economic Studies , 29 ( 3 ), 155 – 173 . Google Scholar CrossRef Search ADS Arthur W. B. ( 1989 ), ‘ Competing technologies, increasing returns, and lock-in by historical events ,’ The Economic Journal , 99 ( 394 ), 116 – 131 . Google Scholar CrossRef Search ADS Ball L. , DeLong B. , Summers L. ( 2014 ), Fiscal policy and full employment, https://www.cbpp.org/research/full-employment/fiscal-policy-and-full-employment. Ball L. M. ( 2014 ), ‘Long-term damage from the great recession in OECD countries,’ NBER Working Paper 20185. National Bureau of Economic Research: Cambridge (MA), US. Bassi F. , Lang D. ( 2016 ), ‘ Investment hysteresis and potential output: a post-Keynesian–Kaleckian agent-based approach ,’ Economic Modelling , 52 , 35 – 49 . Google Scholar CrossRef Search ADS Beaudry P. , Galizia D. , Portier F. ( 2016 ), ‘Is the macroeconomy locally unstable and why should we care?,’ NBER Working Paper 22275. National Bureau of Economic Research: Cambridge (MA), US . Bertrand M. , Schoar A. , Thesmar D. ( 2007 ), ‘ Banking deregulation and industry structure: evidence from the French banking reforms of 1985 ,’ The Journal of Finance , 62 ( 2 ), 597 – 628 . ISSN 1540-6261. Google Scholar CrossRef Search ADS Blanchard O. J. , Leigh D. . ( 2013 ), ‘ Growth forecast errors and fiscal multipliers ,’ American Economic Review , 103 , 117 – 120 . Blanchard O. , Cerutti E. , Summers L. ( 2015 ), ‘Inflation and activity–two explorations and their monetary policy implications,’ NBER Working Paper 21726. National Bureau of Economic Research: Cambridge (MA), US . Blanchard O. J. , Summers L. H. ( 1986 ), ‘ Hysteresis and the European unemployment problem ,’ NBER Macroeconomics Annual , 1 , 15 – 78 . Google Scholar CrossRef Search ADS Blanchard O. J. , Summers L. H. ( 1987 ), ‘ Hysteresis in unemployment ,’ European Economic Review , 31 ( 1–2 ), 288 – 295 . Google Scholar CrossRef Search ADS Boyer R. , Saillard Y. ( 2005 ), Régulation Theory: The State of the Art . Routledge : London , Broock W. , Scheinkman J. A. , Dechert W. D. , LeBaron B. ( 1996 ), ‘ A test for independence based on the correlation dimension ,’ Econometric Reviews , 15 ( 3 ), 197 – 235 . Google Scholar CrossRef Search ADS Caiani A. , Godin A. , Caverzasi E. , Gallegati M. , Kinsella S. , Stiglitz J. E. ( 2016a ), ‘ Agent based-stock flow consistent macroeconomics: towards a benchmark model ,’ Journal of Economic Dynamics and Control , 69 , 375 – 408 . Google Scholar CrossRef Search ADS Caiani A. , Russo A. , Gallegati M. ( 2016b ), Does Inequality Hamper Innovation and Growth? Available at SSRN 2790911. Castaldi C. , Dosi G. ( 2006 ), ‘The grip of history and the scope for novelty: some results and open questions on path dependence in economic processes,’ in Wimmer A. , Kössler R. , editors, Understanding Change . Springer : NY, US . pp. 99 – 128 . Google Scholar CrossRef Search ADS Cioppa T. M. , Lucas T. W. ( 2007 ), ‘ Efficient nearly orthogonal and space-filling latin hypercubes ,’ Technometrics , 49 ( 1 ), 45 – 55 . Google Scholar CrossRef Search ADS David P. A. ( 1985 ), ‘ Clio and the economics of qwerty ,’ The American Economic Review , 75 ( 2 ), 332 – 337 . David P. A. ( 1994 ), ‘ Why are institutions the ‘carriers of history’?: Path dependence and the evolution of conventions, organizations and institutions ,’ Structural Change and Economic Dynamics , 5 ( 2 ), 205 – 220 . Google Scholar CrossRef Search ADS Dawid H. , Harting P. , Neugart M. ( 2014 ), ‘ Economic convergence: policy implications from a heterogeneous agent model ,’ Journal of Economic Dynamics and Control , 44 , 54 – 80 . Google Scholar CrossRef Search ADS Day R. H. , Chen P. ( 1993 ), Nonlinear Dynamics and Evolutionary Economics . Oxford University Press : Oxford , Decker R. A. , Haltiwanger J. , Jarmin R. S. , Miranda J. ( 2016 ), ‘Declining business dynamism: Implications for productivity?,’ Hutchins Center Working Paper. Brookings Institution: Cambridge (MA), US . Dosi G. , Kaniovski Y. ( 1994 ), ‘ On “badly behaved” dynamics ,’ Journal of Evolutionary Economics , 4 ( 2 ), 93 – 123 . Google Scholar CrossRef Search ADS Dosi G. , Virgillito M. E. (2018), ‘The “Schumpeterian” and the “Keynesian” Stiglitz: learning, coordination hurdles and growth trajectories,’ in Guzman M. (ed.), Toward a Just Society. Joseph Stiglitz and Twenty-First Century Economics . Columbia University Press , NY, US. Dosi G. , Fagiolo G. , Roventini A. ( 2010 ), ‘ Schumpeter meeting Keynes: a policy-friendly model of endogenous growth and business cycles ,’ Journal of Economic Dynamics and Control , 34 ( 9 ), 1748 – 1767 . Google Scholar CrossRef Search ADS Dosi G. , Fagiolo G. , Napoletano M. , Roventini A. ( 2013 ), ‘ Income distribution, credit and fiscal policies in an agent-based Keynesian model ,’ Journal of Economic Dynamics and Control , 37 ( 8 ), 1598 – 1625 . Google Scholar CrossRef Search ADS Dosi G. , Fagiolo G. , Napoletano M. , Roventini A. , Treibich T. ( 2015 ), ‘ Fiscal and monetary policies in complex evolving economies ,’ Journal of Economic Dynamics and Control , 52 , 166 – 189 . Google Scholar CrossRef Search ADS Dosi G. , Napoletano M. , Roventini A. , Treibich T. ( 2016a ), ‘The short-and long-run damages of fiscal austerity: keynes beyond Schumpeter,’ in Stiglitz J. , Guzman M. (eds), Contemporary Issues in Macroeconomics . Palgrave Macmillan , pp. 79 – 100 . Dosi G. , Pereira M. C. , Roventini A. , Virgillito M. E. ( 2016b ), ‘The effects of labour market reforms upon unemployment and income inequalities: an agent based model,’ LEM Papers Series 2016/27. Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced Studies: Pisa, Italy. Dosi G. , Napoletano M. , Roventini A. , Stiglitz J. , Treibich T. ( 2017a ), Rational Heuristics? Expectation and Behavior in Evolving Economies with Heterogeneous Interacting Agents. Lem Paper Series Forthcoming, Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced Studies. Dosi G. , Napoletano M. , Roventini A. , Treibich T. ( 2017b ), ‘ Micro and macro policies in the Keynes + Schumpeter evolutionary models ,’ Journal of Evolutionary Economics , 27 ( 1 ), 63 – 90 . Google Scholar CrossRef Search ADS Dosi G. , Pereira M. C. , Roventini A. , Virgillito M. E. ( 2017c ), ‘ When more flexibility yields more fragility: the microfoundations of Keynesian aggregate unemployment ,’ Journal of Economic Dynamics and Control , 81 , 162 – 186 . Google Scholar CrossRef Search ADS Dosi G. , Pereira M. C. , Virgillito M. E. ( 2017d ), ‘ On the robustness of the fat-tailed distribution of firm growth rates: a global sensitivity analysis ,’ Journal of Economic Interaction and Coordination . https://doi.org/10.1007/s11403-017-0193-4. Dosi G. , Pereira M. C. , Virgillito M. E. ( 2017e ), ‘ The footprint of evolutionary processes of learning and selection upon the statistical properties of industrial dynamics ,’ Industrial and Corporate Change , 26 ( 2 ), 187 – 210 . Fagiolo G. , Roventini A. ( 2012 ), ‘ Macroeconomic policy in DSGE and Agent-Based Models ,’ Revue de l’OFCE , 124 ( 5 ), 67 – 116 . Fagiolo G. , Roventini A. ( 2017 ), ‘ Macroeconomic policy in DSGE and Agent-Based Models redux: new developments and challenges ahead ,’ Journal of Artificial Societies and Social Simulation , 20 ( 1 ), 1 . Google Scholar CrossRef Search ADS Freeman C. , Clark J. , Soete L. ( 1982 ), Unemployment and Technical Innovation . Pinter Publisher : London and New York , Galí J. ( 2015 ), ‘Hysteresis and the European unemployment problem revisited,’ NBER Working Paper 21430. National Bureau of Economic Research: Cambridge (MA), US . Geroski P. ( 1991 ), Market Dynamic and Entry . Wiley-Blackwell , Geroski P. ( 1995 ), ‘ What do we know about entry?, ’ International Journal of Industrial Organization , 13 ( 4 ), 421 – 440 . The Post-Entry Performance of Firms. Google Scholar CrossRef Search ADS Ghayad R. ( 2013 ), ‘The jobless trap,’ Technical report. Northeastern University: Boston, US . Gomis R. M. , Khatiwada S. ( 2017 ), ‘Firm dynamics and business cycle: What doesn’t kill you makes you stronger?,’ Technical report. Economics Section, The Graduate Institute of International Studies: Geneva (CH). Gourio F. , Messer T. , Siemer M. ( 2016 ), ‘A missing generation of firms? Aggregate effects of the decline in new business formation.’ In 2016 Meeting Papers (No. 752). Society for Economic Dynamics , Toulouse (FR) . Gualdi S. , Tarzia M. , Zamponi F. , Bouchaud J.-P. ( 2015 ), ‘ Tipping points in macroeconomic agent-based models ,’ Journal of Economic Dynamics and Control , 50 , 29 – 61 . Google Scholar CrossRef Search ADS Hallett A. H. , Piscitelli L. ( 2002 ), ‘ Testing for hysteresis against nonlinear alternatives ,’ Journal of Economic Dynamics and Control , 27 ( 2 ), 303 – 327 . Google Scholar CrossRef Search ADS Jaimovich N. , Siu H. E. ( 2012 ), ‘The trend is the cycle: Job polarization and jobless recoveries,’ NBER Working Paper 18334. National Bureau of Economic Research: Cambridge (MA), US . Kerr W. R. , Nanda R. ( 2009 ), ‘ Democratizing entry: banking deregulations, financing constraints, and entrepreneurship ,’ Journal of Financial Economics , 94 ( 1 ), 124 – 149 . Google Scholar CrossRef Search ADS Kothari S. P. , Lewellen J. , Warner J. ( 2014 ), ‘The behavior of aggregate corporate investment,’ Technical Report 5112-14. MIT Sloan Research Paper. Lamberson P. , Page S. ( 2012 ), ‘ Tipping points ,’ Quarterly Journal of Political Science , 7 ( 2 ), 175 – 208 . Google Scholar CrossRef Search ADS LeBaron B. , Tesfatsion L. ( 2008 ), ‘ Modeling macroeconomies as open-ended dynamic systems of interacting agents ,’ American Economic Review , 98 ( 2 ), 246 – 250 . Google Scholar CrossRef Search ADS Lee Y. , Mukoyama T. ( 2015 ), ‘ Entry and exit of manufacturing plants over the business cycle ,’ European Economic Review , 77 , 20 – 27 . Google Scholar CrossRef Search ADS Lorenz H.-W. ( 1993 ), Nonlinear Dynamical Economics and Chaotic Motion , vol. 334 . Springer : NY, US . Google Scholar CrossRef Search ADS Louca F. ( 2001 ), ‘ Intriguing pendula: founding metaphors in the analysis of economic fluctuations ,’ Cambridge Journal of Economics , 25 , 25 – 55 . Google Scholar CrossRef Search ADS Lucidi F. , Kleinknecht A. ( 2009 ), ‘ Little innovation, many jobs: an econometric analysis of the Italian labour productivity crisis ,’ Cambridge Journal of Economics , 34 ( 3 ), 525 – 546 . Google Scholar CrossRef Search ADS Morris M. D. ( 1991 ), ‘ Factorial sampling plans for preliminary computational experiments ,’ Technometrics , 33 ( 2 ), 161 – 174 . Google Scholar CrossRef Search ADS Napoletano M. , Dosi G. , Fagiolo G. , Roventini A. ( 2012 ), ‘ Wage formation, investment behavior and growth regimes: an agent-based analysis ,’ Revue de l’OFCE , 5 ( 124 ), 235 – 261 . Ng S. , Wright J. H. ( 2013 ), ‘ Facts and challenges from the recession for forecasting and macroeconomic modeling ,’ Journal of Economic Literature , 51 ( 4 ), 1120 – 1154 . Google Scholar CrossRef Search ADS Piscitelli L. , Cross R. , Grinfeld M. , Lamba H. ( 2000 ), ‘ A test for strong hysteresis ,’ Computational Economics , 15 ( 1/2 ), 59 – 78 . Google Scholar CrossRef Search ADS Rasmussen C. , Williams C. ( 2006 ), Gaussian Processes for Machine Learning. MIT Press : Cambridge, MA . Reifschneider D. , Wascher W. , Wilcox D. ( 2015 ), ‘ Aggregate supply in the United States: recent developments and implications for the conduct of monetary policy ,’ IMF Economic Review , 63 ( 1 ), 71 – 109 . Google Scholar CrossRef Search ADS Rosser J. B. ( 2013 ), From Catastrophe to Chaos: A General Theory of Economic Discontinuities . Springer Science & Business Media : NY, US . Russo A. , Riccetti L. , Gallegati M. ( 2016 ), ‘ Increasing inequality, consumer credit and financial fragility in an agent based macroeconomic model ,’ Journal of Evolutionary Economics , 26 ( 1 ), 25 – 47 . Google Scholar CrossRef Search ADS Salle I. , Yıldızoğlu M. ( 2014 ), ‘ Efficient sampling and meta-modeling for computational economic models ,’ Computational Economics , 44 ( 4 ), 507 – 536 . Google Scholar CrossRef Search ADS Saltelli A. , Ratto M. , Andres T. , Campolongo F. , Cariboni J. , Gatelli D. , Saisana M. , Tarantola S. ( 2008 ), Global Sensitivity Analysis: The Primer . John Wiley & Sons : New York . Seppecher P. , Salle I. , Lavoie M. ( 2017 ), ‘What drives markups? Evolutionary pricing in an agent-based stock-flow consistent macroeconomic model,’ CEPN Working Papers 2017-03. Centre d’Economie de l’Université de Paris Nord: Paris (FR). Setterfield M. ( 2009 ), Path Dependency, Hysteresis and Macrodynamics . Springer , pp. 37 – 79 . Google Scholar CrossRef Search ADS Siemer M. ( 2014 ), ‘Firm entry and employment dynamics in the Great Recession,’ Technical report. Skott P. ( 2005 ), ‘ Fairness as a source of hysteresis in employment and relative wages ,’ Journal of Economic Behavior & Organization , 57 ( 3 ), 305 – 331 . ISSN 0167-2681. doi: https://doi.org/10.1016/ j.jebo.2003.11.011. URL http://www.sciencedirect.com/science/article/pii/S016726810400143X. Google Scholar CrossRef Search ADS Sobol I. M. ( 1993 ), ‘ Sensitivity analysis for non-linear mathematical models ,’ Mathematical Modelling and Computational Experiments , 1 ( 4 ), 407 – 414 . Stiglitz J. ( 1994 ), ‘Endogenous growth and cycles,’ in Shionoya Y. , Perlman M. (eds), Innovation in Technology, Industry and Institutions . The University of Michigan Press , pp. 121 – 156 . Stockhammer E. ( 2011 ), ‘ Wage norms, capital accumulation, and unemployment: a post-keynesian view ,’ Oxford Review of Economic Policy , 27 ( 2 ), 295 – 311 . Google Scholar CrossRef Search ADS Tesfatsion L. , Judd K. (eds) ( 2006 ), Handbook of Computational Economics II: Agent-Based Computational Economics . Elsevier : North Holland, Amsterdam . Wald A. , Wolfowitz J. ( 1940 ), ‘ On a test whether two samples are from the same population ,’ The Annals of Mathematical Statistics , 11 ( 2 ), 147 – 162 . Google Scholar CrossRef Search ADS Zhou H. , Dekker R. , Kleinknecht A. ( 2011 ), ‘ Flexible labor and innovation performance: evidence from longitudinal firm-level data ,’ Industrial and Corporate Change , 20 ( 3 ), 941 . Google Scholar CrossRef Search ADS Appendices Appendix A Capital- and consumption-good sectors and technical change The technology of capital-good firms is (Aiτ,Biτ). Aiτ is the labor productivity of the machine-tool manufactured by i for the consumption-good sector, while Biτ is the labor productivity to produce the machine. Superscript τ denotes the technology vintage being produced/used. Given the monetary average wage wi,t paid by firm i, the unit cost of production is: ci,t=wi,tBiτ. (6) With a fixed markup μ1∈ℝ+ pricing rule, prices pi,t are defined as: pi,t=(1+μ1)ci,t. (7) Firms in the capital-good industry “adaptively” strive to increase their market shares and their profits trying to improve technology via innovation and imitation. Firms invest in R&D a fraction ν∈(0,1] of their past sales Si,t−1: RDi,t=νSi,t−1. (8) R&D activity is performed by workers exclusively devoted to this activity, whose demand is: Li,tR&D=RDi,twi,t. (9) Firms split their R&D workers Li,tR&D between innovation ( INi,t) and imitation ( IMi,t) activities according to the parameter ξ∈[0,1]: INi,t=ξLi,tR&D (10) IMi,t=(1−ξ)Li,tR&D. (11) Innovation is a two-step process. The first one determines whether a firm obtains or not access to an innovation—irrespectively of whether it is ultimately a success or a failure—through a draw from a Bernoulli distribution with parameter: θi,tin=1−e−ζ1INi,t, (12) and parameter ζ1∈(0,1]. If a firm innovates, it may draw a new machine-embodying technology (Ai,tin,Bi,tin) according to: Ai,tin=Ai,t(1+xi,tA), (13) Bi,tin=Bi,t(1+xi,tB), (14) where xi,tA and xi,tB are two independent draws from a Beta (α1,β1) distribution, (α1,β1)∈ℝ2+ over the fixed support [x¯1,x¯1]⊂ℝ. Imitation also follows a two-step procedure. The access to imitation comes from sampling a Bernoulli with parameter: θi,tim=1−e−ζ2IMi,t (15) and ζ2∈(0,1]. Firms accessing the second stage are able to copy the technology (Aiim,Biim) of one of the competitors. Finally, they select the machine to produce according to the rule: min [pi,th+bc(Ai),th], h=τ,in,im, (16) where b∈ℝ+ is a payback parameter. Firms in consumption-good sector do not conduct R&D; instead they access new technologies incorporating new machines to their existing capital stock Ξj,t−1. Firms invest according to expected demand Dj,te, computed by an adaptive rule: Dj,te=g(Dj,t−1,…,Dj,t−h), 0<h<t, (17) where Dj,t−h is the actual demand faced by firms at time t–h ( h∈ℕ* is a parameter and g:ℝh→ℝ+ is the expectation function, here an unweighted moving average over four periods). The corresponding desired level of production Qj,td, considering the actual inventories from previous period Nj,t−1, is: Qj,td=(1+ι)Dj,te−Nj,t−1, (18) being Nj,td=ιDj,te the desired inventories and ι∈ℝ+, a parameter. If the desired capital stock Kjd —computed as a linear function of the desired level of production Qj,td —is higher than the current one, firms invest EIj,td to expand their production capacity: EIj,td=Kj,td−Kj,t−1. (19) Firms also invest SIj,td to replace machines by more productive vintages according to a fixed payback period (b > 0) rule, substituting machines Aiτ∈Ξj,t according to its obsolescence as well as the price of new machines: RSj,t={Aiτ∈Ξj,t:pi,t*cj,tAiτ−cj,t*≤b}, (20) where pi,t*∈ℝ+ and cj,t*∈ℝ+ are the price and unit cost of production upon the new machines. Given the stock of machines Ξj,t, firms compute average productivity πj,t and average unit cost of production cj,t, based on the average unit labor cost of production wj,t associated with each machine of vintage τ in its capital stock: cj,tAiτ=wj,tAiτ. (21) Consumption-good prices are set applying a markup μj,t on average unit costs: pj,t=(1+μj,t)cj,t. (22) Markup changes are regulated by the evolution of firm market shares (fj,t): μj,t=μj,t−1(1+υfj,t−1−fj,t−2fj,t−2), (23) with υ∈(0,1). Firm market shares evolve according to a replicator dynamics: fj,t=fj,t−1(1+χEj,t−E¯tE¯t), (24) where the firms competitivity Ej,t is defined based on the individual normalized prices p′j,t and unfilled demands l′j,t: Ej,t=−ω1p′j,t−1−ω1l′j,t−1, E¯t=1Ft2∑jEj,tfj,t−1, (25) being (ω1,ω1)∈ℝ2 parameters. Labor market and search and match process Labor demand in the consumption-good sector Lj,td is determined by desired production Qj,td and the average productivity of current capital stock Aj,t: Lj,td=Qj,tdAj,t. (26) In the capital-good sector, instead, Li,td considers orders Qi,t and labor productivity Bi,t. In what follows, only the behavior of the consumption-good sector (subscript j) is shown, as the capital-good firms operate under the same rules in the labor market, except they follow the wage offers from top-paying firms in the former sector. Firms decide whether to hire (or fire) workers according to the expected production Qj,td. If it is increasing, ΔLj,td new workers are (tentatively) hired in addition to the existing number Lj,t−1. Each firm (expectedly) gets a fraction of the number of applicant workers La,t in its candidates queue {ℓj,ts}, proportional to firm market share fj,t−1. In terms of statistical expectations: E(Lj,ts)=ωLa,tfj,t−1, (27) where ω∈ℝ+ is a parameter defining the number of job queues each seeker joins, in average. Considering the set of workers in {ℓj,ts}, each firm selects the subset of desired workers {ℓj,td} to make a job (wage) offer: {ℓj,td}={ℓj,t∈{ℓj,ts}:wℓ,tr<wj,to}, {ℓj,td}⊆{ℓj,ts}. (28) Firms target workers that would accept the wage offer wj,to, considering the wage wℓ,tr requested by workers, if any. Each firm hires workers up to its demand ΔLj,td, or to all workers in its queue, and the number of effectively hired workers (the set {ℓj,th}) is: #{ℓj,th}=ΔLj,t≤ΔLj,td≤Lj,ts=#{ℓj,ts}, ΔLj,t=Lj,t−Lj,t−1. (29) The search, wage determination, and firing processes differ according to the policy regime. In the Fordist regime, workers never quit jobs and firms fire employees only under losses ( Πj,t−1<0) and shrinking desired production ( Qj,td<Qj,t−1), except if exiting the market. Only unemployed workers search for jobs. Additionally, lowest skilled workers are fired first, while higher skilled workers are preferred when hiring, as in this regime wages are not bargained. Firms offer a wage: wj,to=wj,t−1o(1+WPj,t) bounded to wj,tmax=pj,t−1Aj,t−1, (30) that is accepted by the worker if she has no better offer. The positive wage premium is is defined as: WPj,t=ψ2ΔAtAt−1+ψ4ΔAj,tAj,t−1, ψ2+ψ4≤1, (31) being At the aggregate labor productivity and (ψ2,ψ4)∈ℝ2+ parameters. So, wages are linked to firm-specific performance and also to the aggregate productivity dynamics. wj,to is simultaneously applied to all firm’s workers. wj,to is bounded to a maximum break-even wage wj,tmax (the 0 unit profits myopic expectation). In the Competitive setting, firms freely fire workers and employees actively search for better jobs while employed, quitting when there is a better offer. When hiring or firing, firms give precedence to workers with a higher skills-to-wage ratio ( stℓ/wtℓ), contracting them first and dismissing last. The matching is done by a one-round bargaining process. Workers have a reservation wage equal to the unemployment benefit wtu they receive from the Government when unemployed, if any, and request a wage wℓ,tr during the job application: wℓ,tr={wℓ,t−1(1+ϵ) if employed in t-1 wℓ,ts if unemployed in t-1. (32) wℓ,t−1 is the current wage for the employed workers, and ϵ∈ℝ+ is a parameter. Unemployed workers have a gradually shrinking satisfying wage wℓ,ts, accounting for their recent wage history: wℓ,ts=max (wtu,1Ts∑h=1Tswℓ,t−h), (33) being Ts∈ℕ*, the time span parameter of the moving-average of the past income. An employed worker accepts the best offer wj,to she receives if it is higher than her current wage wℓ,t. An unemployed worker accepts the best offer she gets, if any, as all offers are at least equal to the unemployment benefit wtu. In all cases, Government establishes an institutional minimum wage wtmin, as the lower bound to the firm wage setting behavior: wtmin=wt−1min(1+ψ2ΔAtAt−1). (34) Model closure Government taxes firms profits at a fixed rate tr∈ℝ+ and provides a benefit wtu to unemployed workers which is a fraction of the current average wage: wtu=ψ1Lt−1D∑ℓ=1Lt−1Dwℓ,t−1, (35) where ψ∈[0,1] is a parameter, and LtD is the total labor demand. Therefore, the Government expenses are: Gt=wtu(LS−LtD). (36) Workers fully consume their income (if possible) and do not get credit. Accordingly, desired aggregate consumption Ctd depends on the income of both employed and unemployed workers plus the desired unsatisfied consumption from previous periods ( Ct−1d−Ct−1): Ctd=∑ℓwℓ,t+Gt+(Ct−1d−Ct−1). (37) The model applies the standard national account identities by the simple aggregation of agents’ stocks and flows. The aggregate value added by capital- and consumption-good firms Yt equals their aggregate production Qt1 and Qt2, respectively (there are no intermediate goods). In turn, it is equal to the sum of the effective consumption Ct, the total investment It, and the change in firm’s inventories ΔNt: Qt1+Qt2=Yt=Ct+It+ΔNt. (38) For further details, see Dosi et al. (2010) and Dosi et al. (2017c). Appendix B Table B1. Model parameters and initial conditions, calibration values, minimum–maximum range for sensitivity analisys, Elementary Effects μ* statistic, and Sobol decomposition direct and interaction effects indexes Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Note: Baseline policy-specific values (Fordist regime). SA statistics relative to GDP crises recovery losses indicator averages for both regimes. Appendix B Table B1. Model parameters and initial conditions, calibration values, minimum–maximum range for sensitivity analisys, Elementary Effects μ* statistic, and Sobol decomposition direct and interaction effects indexes Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Note: Baseline policy-specific values (Fordist regime). SA statistics relative to GDP crises recovery losses indicator averages for both regimes. Table B2. Stock-and-flow consistency: transaction flow matrix Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Note: (*) Government deficit/superavit is close to 0 in the long run. Table B2. Stock-and-flow consistency: transaction flow matrix Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Note: (*) Government deficit/superavit is close to 0 in the long run. Table B3. Regime-specific parameter values Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 Table B3. Regime-specific parameter values Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 © The Author(s) 2018. Published by Oxford University Press on behalf of Associazione ICC. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Industrial and Corporate Change Oxford University Press

Causes and consequences of hysteresis: aggregate demand, productivity, and employment

Loading next page...
 
/lp/ou_press/causes-and-consequences-of-hysteresis-aggregate-demand-productivity-QbMjqPmoOv
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of Associazione ICC.
ISSN
0960-6491
eISSN
1464-3650
D.O.I.
10.1093/icc/dty010
Publisher site
See Article on Publisher Site

Abstract

Abstract In this work we develop an agent-based model where hysteresis in major macroeconomic variables (e.g., gross domestic product, productivity, unemployment) emerges out of the decentralized interactions of heterogeneous firms and workers. Building upon the “Schumpeter meeting Keynes” family of models (cf. in particular Dosi et al. (2016b, 2017c)), we specify an endogenous process of accumulation of workers’ skills and a state-dependent process of firms entry. Indeed, hysteresis is ubiquitous. However, this is not due to market imperfections, but rather to the very functioning of decentralized economies characterized by coordination externalities and dynamic increasing returns. So, contrary to the insider–outsider hypothesis (Blanchard and Summers, 1986), the model does not support the findings that rigid industrial relations may foster hysteretic behavior in aggregate unemployment. On the contrary, this contribution provides evidence that during severe downturns, and thus declining aggregate demand, phenomena like decreasing investment and innovation rates, skills deterioration, and declining entry dynamics are better candidates to explain long-run unemployment spells and reduced output growth. In that, more rigid labor markets may well dampen hysteretic dynamics by sustaining aggregate demand, thus making the economy more resilient. 1. Introduction In this work, we explore the extent to which the labor-augmented “Schumpeter meeting Keynes” (K + S) model is able to display the endogenous emergence of hysteresis out of the interaction of heterogeneous firms and workers. The article focuses on both the causes and the consequences of the hysteretical properties of macroeconomic time series, including gross domestic product (GDP), productivity, and unemployment. Further, refining upon Dosi et al. (2016b, 2017c), we introduce an endogenous process of accumulation of workers’ skills and a state-dependent process of firms entry, studying their hysteretic effects. As we shall briefly discuss below, there are different notions of hysteresis. Basically, they boil down to three interpretations of the phenomenon (more in Piscitelli et al., 2000, Hallett and Piscitelli, 2002, Amable et al., 2004). The first is formulated in terms of the persistence in the deviations from some equilibrium path; the second is defined as a random-walk dynamics in equilibrium itself; the third, we believe a more genuine one, is in terms of the heterogeneous and nonlinear responses of a system characterized by multiple equilibria or path-dependent trajectories. Even if Piscitelli et al. (2000: 59–60) define the former two as bastard usages of the notion of hysteresis, they have been so far the most common ones in economics. In this work we shall adopt the third notion which encompasses the phenomena of remanence, super-hysteresis, persistence, nonlinearity, and path dependency. Nonetheless, in an archetypical example, Blanchard and Summers (1986) used the second of the foregoing interpretations in an attempt to explain the structural unemployment in the late 1980s in many European countries, at around 10% and quite far from the predicted 2–3% equilibrium level: The recent European experience has led to the development of alternative theories of unemployment embodying the idea that the equilibrium unemployment rate depends on the history of the actual unemployment rate. Such theories may be labelled hysteresis theories after the term in the physical sciences referring to situations where equilibrium is path-dependent (Blanchard and Summers, 1987: 1). Two alternative hypotheses were proposed by these authors to explain the emergence of hysteresis, a first one resting on the membership channel according to which only insider workers are able to exert pressure in the wage setting process, and a second one based on the duration channel because the long-term unemployed workers are less relevant in the wage determination process. In the latter case, unemployment duration can (i) induce a process of worker skills deterioration, implying that the long-term unemployed worker experiences a fall in its productivity; and, (ii) trigger search discouragement in unemployed people, less re-employable, and so less prone to search in the labor market. Together with the supply-side channels emphasized since the 80s, after the current crisis, some acknowledgements have gone to aggregate demand shocks conceived as potential sources of hysteresis. Therefore, the notion of hysteresis has been extended from unemployment to permanent output loss. Blanchard et al. (2015) revisit hysteresis as the permanent effect exerted by crises on the the levels of output relative to the pre-crises ones. That work provides evidence of a persistent output gap in 69% of the cases, among 22 countries in the period 1960–2010, where in 47% of them the recession was followed by an increasing output gap, meaning that recessionary periods affected not only the levels but also the subsequent growth rates, an effect named by Ball (2014) as super-hysteresis. In fact, Ball (2014) reports that over 23 countries in the period 2007–2014, most of them have been hit by severe recessions, and some of them, like Greece, faced up to 30% losses in potential output. […] in most countries the loss of potential output is almost as large as the shortfall of actual output from its pre-crisis trend. This finding implies that hysteresis effects have been very strong during the Great Recession. Second, in the countries hit hardest by the recession, the growth rate of potential output is significantly lower today than it was before 2008. This growth slowdown means that the level of potential output is likely to fall even farther below its pre-crisis trend in the years to come (Ball, 2014: 2). The empirical detection of hysteresis, of course, goes together with the analysis of its determinants. Agent-based models are particularly suitable to the task as one knows by construction the micro data-generating process and thus can explore the possible hysteretic features of aggregate variables as emergent properties of the evolutionary dynamics.1 The model, built upon the “Schumpeter meeting Keynes” family of models (Dosi et al., 2010, Napoletano et al., 2012, Dosi et al., 2013, 2015, 2017a and Dosi et al., 2017c), as we shall see, is able to generically yield hysteresis in the macro variables under scrutiny both inter-regimes and intra-regimes, i.e., across institutional setups governing the labor markets. Indeed, hysteresis is ubiquitous. According to our analysis, hysteresis is not due to market imperfections but rather to the very functioning of decentralized economies characterized by coordination externalities and dynamic increasing returns. Contrary to what suggested by, e.g., Blanchard and Summers (1986), our model does not support the hypothesis that rigid industrial relations, via the insider–outsider channel, are the driving source of hysteresis in aggregate unemployment. On the contrary, more in line with Ball et al. (2014), our results indicate that during severe downturns and thus declining aggregate demand, phenomena like decreasing investment and innovation rates, skills deterioration, and declining entry dynamics are better candidates to explain long-term unemployment spells and reduced output growth. In such a framework, more rigid labor markets, by supporting aggregate demand, do not foster hysteresis but rather may well dampen it, thus making the economy more resilient. The article is organized as follows. Section 2 discusses the nature and the sources of hysteresis. In Section 3, we present the model structure. The empirical regularities matched by the K + S model are discussed in Section 4. In Section 5, we study the emergence and the causes of hysteresis. Finally, Section 6 concludes. 2. The nature and determinants of hysteresis In this section, we provide a brief exploration on the sources and potential channels which might induce hysteretic behaviors in the macroeconomic variables. 2.1 The nature of hysteresis Hysteresis, a concept adopted from the natural sciences but with similar instances in economics, is a nonlinear mechanism, often implying multiple (alternative) time trajectories and equilibria. In a very broad perspective, a dynamical system can be considered hysteretical when the time trajectories of some or all of its variables do exhibit path dependency, in turn also implying non-ergodicity. The very notion of multiple paths for the development of both socioeconomic and natural complex systems ultimately rests on the idea that history is an essential part of the interpretation of many dynamic phenomena. The property that history matters is also intimately related to that of time irreversibility, that is, a situation where it is not possible, even theoretically, to “reverse the arrow of time” and still expects to recover invariant properties of the system under investigation. Reviewing the literature on complex systems is beyond the scope of this article. Suffice to recall the distinction between nonlinear deterministic systems and stochastic ones, both however displaying forms of path dependency. Concerning the former, instantiations are bifurcation, chaotic, and catastrophe dynamics (see Lorenz, 1993). With respect to the latter, Generalized Polya Urns are a well-known example. Both families of processes are often characterized by the presence of tipping points whereby a tip is a threshold point (variable or parameter) which, when reached, might induce irreversible changes on the evolution of the state-space (see Lamberson and Page, 2012 for a detailed discussion in social sciences). On empirical grounds, in tackling path-dependent phenomena in the social sciences (but also in, e.g., biology), an intrinsic difficulty rests in the fact that frequently only one of the many possible realizations of the system, dependent on its initial state, is empirically observed. In that, how much is history-dependence shaped by initial conditions or conversely how does it relate to irreversible effects of some particular unfolding events (e.g., crises or regime changes)? Related, how do the set of all possible evolutionary paths are shaped and constrained by the structure inherited from the past?2 In economics—at least in the dominant theory as distinct from e.g. economic history—the very notion of hysteresis has only been acknowledged with some skepticism and often in the most restrictive interpretations. In the 1980s and 1990s, a stream of literature has faced head-on the challenge of nonlinearity of growth processes and thus the multiplicity of alternative paths and the related hysteretic properties (insightful examples are the contributions in Anderson et al., 1988, Day and Chen, 1993 and Rosser, 2013).3 However, such a stream of investigation was progressively marginalized, possibly due to its “revolutionary” theoretical implications, particularly in terms of equilibria existence, selection, and the associated welfare theorems. Fundamentally, any form of innovation/knowledge accumulation/learning is associated with dynamic increasing returns and thus nonlinearities (Arrow (1996) witnesses from the General Equilibrium side). Illustrative applications of path-dependent stochastic systems to technology diffusion are in David (1985), Arthur (1989), and Dosi and Kaniovski (1994). Finally, an analysis of tipping points in hybrid agent-based models (ABMs) have been performed in Gualdi et al. (2015). A usual “safer” path has been that of formalizing the phenomenon based on linear stochastic models with close-to-unit-root auto-regressive processes. In this perspective, Blanchard and Summers (1986) identify hysteresis in the unemployment series whenever the coefficient of persistence ρ in the equation Ut=ρUt−1+αt+ϵt+θϵt−1 was estimated to be greater or equal to 1. Whether a (close to) unit-root process is an adequate sign of hysteresis has been strongly debated. In general, this modeling approach is based on a somewhat naive epistemology—like “Which processes should present unit-roots? The natural rate of unemployment, the inflation target, or the wage setting curve?”—, but without jeopardizing the underlying unique equilibrium assumption. So, for example, Galí (2015) explores, without conclusive results, three alternative sources to a unit-root process of the European unemployment rate, testing whether it lies (i) in the natural rate of unemployment (Utn=Ut−1n+ϵt), (ii) in the central bank inflation target (πt*=πt−1*+ϵt*), or (iii) in the insider–outsider hypothesis (à la Blanchard–Summers) via alternative specifications for the New Keynesian Wage Phillips Curve. The obvious dissatisfaction with the unit-root process approach is currently yielding a revival of the detection of nonlinearities in empirical macroeconomics. For example, Beaudry et al. (2016), while examining empirical time series like unemployment and working hours, do find evidence of recurrent cyclical patterns, not detectable when estimating auto-regressive linear stochastic models. However, the critique to the unit-root process approach is deeper and concerns its very underlying theory: as suggested by Piscitelli et al. (2000), Hallett and Piscitelli (2002), Amable et al. (2004), and Bassi and Lang (2016), genuine models of hysteresis should embed a nonlinear structure—or at least do not discard nonlinearity in advance. According to Piscitelli et al. (2000), three features characterize hysteretic processes, namely, nonlinearity, selectivity, and remanence. Being this memory process nonlinear, reversing a shock may not drive the system to recover its starting point. Moreover, selectivity means that not all shocks affect the system in the same way in different circumstances. Finally, remanence entails that temporary or non-recurrent shocks may lead to permanent new system states. Widespread origins of hysteresis in the socioeconomic domain are, first, feedback mechanisms related to coordination externalities, and, second, amplification processes stemming from some form of increasing returns.4 In particular, it is frequently associated with (i) positive feedbacks between levels of aggregate activities and innovative search, and (ii) powerful interactions between aggregate demand and diffusion of innovations. Whenever one abandons the unfortunate idea that the macroeconomic system is held up to some mysteriously stable and unique equilibrium path, it could well be, for example, that negative demand shocks exert persistent effects because less aggregate demand entails less innovative search, which in turn yields less innovation stemming from technological shocks: [During recessionary phases], typically firms also reduce their expenditures in R&D and productivity-enhancing expenditures. The reduction in output reduces opportunities to “learn by doing.” Thus, the attempt to pare all unnecessary expenditures may have a concomitant effect on long-run productivity growth. In this view, the loss from a recession may be more than just the large, but temporary, costs of idle and wasted resources: the growth path of the economy may be permanently lowered (Stiglitz, 1994: 122). Despite the 2008 crisis, many economists continue to believe in some version of the model underlying the Example A in Figure 1: the economy is bound to “spring back,” with no permanent loss to its long-run equilibrium rate of growth. The econometric side of this view is the Frisch-like idea of the economy as a “pendulum,” responding to exogenous shocks.5 In this perspective, it seems almost a “miracle” that in the empirical literature one recently finds impulse response functions with multipliers significantly greater than 1. This, we suggest, is a witness of the depth of the current crisis (see Blanchard and Leigh, 2013). Figure 1. View largeDownload slide Effects of recessions: (A) short-run (no hysteresis), (B) long-run (hysteresis), (C) permanent/divergent (super-hysteresis). Source: (A and B): Stiglitz, 1994, p. 123. Figure 1. View largeDownload slide Effects of recessions: (A) short-run (no hysteresis), (B) long-run (hysteresis), (C) permanent/divergent (super-hysteresis). Source: (A and B): Stiglitz, 1994, p. 123. However, a small but significant minority of the profession has been forced by the evidence to accept Case B in Figure 1: recession-induced output losses are permanent because even if the system goes back to the pre-crisis rate of growth, then that is associated with an absolute level gap growing exponentially over time. Moreover, as discussed in Stiglitz (1994), imperfect capital markets and credit rationing may well exacerbate the effects of recessions, hampering the recovery of the growth rate even further. Beyond that, recurrent negative demand shocks, such as those deriving from austerity or labor market flexibilization policies, might yield reduced long-term rates of growth: this is what is shown in Dosi et al. (2016a) and Dosi et al. (2017c). In the latter scenario, as in the Example C in Figure 1, the pre- and post-crisis growth trajectories diverge, implying a reduced long-run rate of the output growth. 2.2 Innovation, diffusion, and investment At the empirical level, a first microeconomic channel6 which might induce hysteresis is the lower innovation rate associated with a reduction in the aggregate demand, which turns out in a decline in the productivity growth. Indeed, R&D expenditures are pro-cyclical. Moreover, the diffusion of new technologies and the adoption of capital-embodied, best-practice techniques slow down during crises. Reifschneider et al. (2015) document a drop in the yearly rate of growth of R&D expenditure in the United States from 3.6% during the pre-crisis period (1990–2007), on average, to 1.6% after 2007. Not only the propensity to innovate but also the process of adoption and diffusion of innovation is slowed down by the contraction of aggregate demand. Both phenomena have been emphasized long ago by Freeman et al. (1982) in their search for the patterns and determinants of long-term fluctuations in growth and employment, and, more recently, theoretically investigated in Dosi et al. (2016a, 2017b). Together with the slower rates of innovation, a process of destruction of the installed productive capacity, due to the lack of sales prospects, seems markedly happening in the post-2008. Indeed, even non-Keynesian commentators have identified the current economic crisis as one stemming from the lack of aggregate demand. As the interest rate reached its zero lower bound without fostering any surge in the investment rate, only accelerator-type investment processes seem able to explain the deteriorating dynamics of the productive capacity. Consistently with the accelerator hypothesis, Kothari et al. (2014) report fresh evidence that investments are ultimately affected by the dynamics of sales, rather than by the interest rate. Overall, lower innovation, diffusion and investment rates seem very plausible candidates to explain the current slowdown in productivity. In turn, the fundamental point is that such changes may well bear a long-term impact, that is, hysteretic effects, on the future dynamics of productivity, GDP, and employment. 2.3 Entry dynamics The second microeconomic channel is the declining entry rate of firms in the market, which has been recently investigated especially in the United States (see Gourio et al., 2016). Entry rates have declined since 2006 by about 27%, a widespread phenomenon across all sectors of the economy. This has been accompanied by steady exit rates and, consequently, also shrinking net entry rates. One direct effect of less entry is the reduced creation of new job opportunities. Decker et al. (2016) document a long-term pattern in the declining business dynamism which the authors attribute, mainly, to the contracting share of young firms. In a similar vein, Siemer (2014) introduces the hypothesis of a missing generation of entrants after the 2008 crisis as a result of the tightening financial constraints, primarily affecting young firms. According to his estimates, the more finance-dependent entrant firms reduced their rates of job creation between 4.8 and 10.5 percentage points relative to the less finance-constrained incumbents. In fact, constrained access to credit may represent an important barrier to entry, together with the usual setup costs, particularly during crises and the associated tight finance availability. Conversely, periods of easy access to debt may induce a higher entry rate (see among others Kerr and Nanda, 2009 and Bertrand et al., 2007). All in all, both in bad and good times, the entry dynamics seems to be a potentially relevant source of hysteresis. 2.4 Skills deterioration A third microeconomic channel which might trigger hysteresis is the workers’ skills deterioration process. Once the economy enters a long recessionary phase, firms tend to fire workers. During severe recessions, like the 2008 crisis, unemployment, which under milder downturns could be in principle temporary and cyclical, turns out to be persistent, implying that many workers experience long unemployment spells. Unemployed workers, of course, stop learning-by-doing, lose contact with the new practices and techniques introduced by firms and gradually deteriorate their existing skills. As the economy recovers and the unemployed are finally hired, their productivity is lower than incumbent workers, reducing the overall productivity. Looking at the recent figures, Reifschneider et al. (2015) document that the share of workers who have been unemployed for more than 26 weeks peaked at 45% in 2011, and it was still about 30% in 2013. On a similar vein, Jaimovich and Siu (2012) analyze the speed of economic recovery during different economic recessions (1970, 1975, 1982, 1991, 2001, 2009) in the United States. Their findings suggest that while in the first three recessions aggregate employment begun to expand within 6 months of the trough of the downturn, during the last three crises employment continued to contract for about 20 months before turning around. Therefore, at the end of 2013 employment had not returned to the pre-crisis level. In turn, Abraham et al. (2016) studying the effect of long-term unemployment on employment probability, and earnings find evidence that long unemployment duration is negatively associated with both job-finding rates and earning opportunities, while Ghayad (2013), based on a résumé review study, reports that employers have a strong rejection for long-term unemployed applicants, even in case of equivalent or superior résumé qualification. Hence, the effects of long unemployment episodes upon skills and job-finding probabilities are yet another important candidate to be a source of macroeconomic hysteresis. 3. The model We build a general disequilibrium, stock-and-flow consistent (see Table B2 in Appendix B), agent-based model, populated by heterogeneous firms and workers who behave according to bounded-rational rules. More specifically, we extend the K + S model (Dosi et al., 2010) with explicitly decentralized interactions among firms and workers in the labor market (Dosi et al., 2016b, 2017c), further adding an endogenous process of workers’ skills accumulation and variable number of firms in each market. The two-sector economy in the model is composed of three populations of heterogeneous agents, Ft1 capital-good firms, Ft2 consumption-good firms, LS consumers/workers, plus a bank and the Government.7 The basic structure of the model is depicted in Figure 2. Capital-good firms invest in R&D and produce heterogeneous machine-tools whose productivity stochastically evolves over time. Consumption-good firms combine machines bought from capital-good firms and labor to produce a homogeneous product for consumers. There is a minimal financial system represented by a single bank that provides credit to firms to finance production and investment plans. Credit is allocated to each firm according to their own demand, which is constrained by their past performance, according to a loan-to-sales cap rule applied by the bank. Conversely, credit supply is completely elastic, adapting to the approved credit demand. Workers submit job applications to a small random subset of firms. Firms hire according to their individual adaptive demand expectations. The government levies taxes on firms’ profits, pays unemployment benefits, and sets minimum wages, according to the policy setting, absorbing excess profits and losses from the bank and keeping a relatively balanced budget in the long run. Figure 2. View largeDownload slide The model structure. Boxes in bold style represent heterogeneous agents’ populations. Figure 2. View largeDownload slide The model structure. Boxes in bold style represent heterogeneous agents’ populations. In the following, we first summarize the functioning of the capital- and the consumption-good sectors of our economy, with a focus on the entry process, and then present the labor market dynamics, detailing the skills accumulation and deterioration mechanisms. Finally, we describe the two alternative policy regime settings under which the model has been explored. In Appendix A, we briefly present the firms’, the workers’, and the Government behavioral rules (for details, see also Dosi et al., 2010 and Dosi et al., 2017c). The model main variables, its configuration, and the parameter setup are presented in Appendix B. 3.1 The capital- and consumption-good sectors The capital-good industry is the locus where innovation is endogenously generated in the model. Capital-good firms develop new machine-embodied techniques or imitate the ones of their competitors to produce and sell more productive and cheaper machinery. On demand, they supply machine-tools to consumption-good firms, producing with labor as the only input.8 The capital-good market is characterized by imperfect information and Schumpeterian competition driven by technological innovation. Machine-tool firms signal the price and productivity of their machines to the current customers as well to a subset of potential new ones, and invest a fraction of past revenues in R&D aimed at searching for new machines or copy existing ones. Prices are set using a fixed markup over (labor) costs of production. Consumption-good firms produce a homogeneous good employing capital (composed by different “vintages” of machines) and labor under constant returns to scale. Desired production is determined according to adaptive (myopic) demand expectations. Given the actual inventories, if the current capital stock is not sufficient to produce the desired output, firms order new machines to expand their installed capacity, paying in advance—drawing on their retained past profits or, up to some limit, on bank credit. Moreover, they replace old machines according to a payback period rule. As new machines embed state-of-the-art technologies, the labor productivity of consumption-good firms increases over time according to the mix of vintages of machines in their capital stocks. Consumption-good firms choose in every period their capital-good supplier comparing the price and the productivity of the machines they are aware of. Firms then fix their prices applying a variable markup rule on their production costs, trying to balance profit margins and market shares. More specifically, firms increase their markup and price whenever their market share is expanding and vice versa. Imperfect information is also the normal state of the consumption-good market, so consumers do not instantaneously switch to the most competitive (cheaper) producer. Market shares evolve according to a (quasi) replicator dynamics: more competitive firms expand, while firms with relatively lower competitiveness levels shrink, or exit the market.9 3.2 The entry and exit processes We expanded the earlier K + S model to account for a variable number of firms in both the consumption- and the capital-good sectors ( Ft1, Ft2). In this new version, entry and exit are now independent processes. As before, firms leave the market whenever their market shares get close to 0 or their net assets turn negative (bankruptcy). However, we now define the number of entrants by means of the random variables bt1 and bt2: btz=Ft−1z[(1−o)MAtz+oπtz]  (lower bounded to 0), (1) where z∈{1,2} denotes the sector (capital- or consumption-good, respectively), Ft−1z is the existing number of incumbent firms, MAtz the “financial attractiveness” of the industry, 1≤o≤1 is a mix balance parameter, and πtz is a random draw from a uniform distribution on the fixed support [x¯2,x¯2]. The number of entrants stochastically depends on the number of incumbents (recalling a spin-off process of the former from the latter) with the financial conditions influencing the decision of would be entrants. The industry-specific “financial attractiveness” MAtz in period t is defined as: MAtz=MCtz−MCt−1z  (bounded to [x¯2,x¯2]). (2) MCtz is calculated on firms’ balance sheets as the (log) ratio between the aggregate stocks of liquid assets NWy,t (bank deposits) and bank debt Deby,t: MCtz= log  (∑yNWy,t−1)− log  (∑yDeby,t−1), (3) in each sector, y∈{i,j}, accordingly. So, MCtz measures the sectoral liquidity-to-debt ratio and thus the tightness of the credit market, and MAtz is a proxy to its dynamics. Correspondingly, negative (positive) values of MAtz represent leveraged (deleveraged) markets, meaning that debt is growing faster (slower) than the accumulation of cash equivalents. This means that whenever the overall liquidity-to-debt ratio is shrinking would-be firms are more inclined to enter, and vice versa. The adopted formulation for the entry process tries to model some well-known facts in the industrial dynamics and business cycle literature: (i) the number of entrants is roughly proportional to the number of incumbent firms (Geroski, 1991, 1995), (ii) entry is affected by the easiness of access to credit (Kerr and Nanda, 2009; Bertrand et al., 2007), and (iii) the process is pro-cyclical (Gomis and Khatiwada, 2017; Lee and Mukoyama, 2015). 3.3 The labor market and skills dynamics The labor market in the model implements a fully decentralized search and hiring process between workers and firms (more on that in Dosi et al., 2016b, 2017c). The aggregate supply of labor LS is fixed and all workers are available to be hired in any period. Moreover, also the labor market is characterized by imperfect information. When unemployed, workers submit a certain number of job applications to firms. Employed workers may apply or not for better positions, according to the institutional setup (see Section 3.5 below). Larger firms, in terms of market share, have a proportionally higher probability of receiving job applications, which are organized in separated, firm-specific queues. Firms decide about their individual labor demand based on the received orders (capital-good sector), the expected demand (consumption-good sector), and the expected labor productivity levels. Considering the number and the productivity of the already employed workers, firms decide to (i) hire new workers, (ii) fire part of the existing ones, or (iii) keep the existing labor force. Each hiring firm defines a unique wage offer for the applicant workers, based on its internal conditions and the received applications. Workers select the best offer they get from the firms to which they submitted applications, if any. If already employed (depending on the institutional regime), they quit the current job if a better wage offer is received. There is no second round of bargaining between workers and firms in the same period, and, so, firms have no guarantee of fulfilling all the open positions (no market clearing). Moreover, there is no firing or hiring transaction costs. We extended the K + S model to account for the process of workers’ skills accumulation and deterioration. Such a process is driven by the worker-specific job tenures, assuming learning-by-doing mechanism when employed and a gradual deterioration of skills while unemployed, and assuming firms keep introducing new techniques all the time, deprecating the skills of unemployed workers. The skill level sℓ,t>0 of each worker ℓ evolves over time as a multiplicative process: sℓ,t={(1+τ)sℓ,t−1 if employed in t−111+τsℓ,t−1 if unemployed in t−1, (4) with the learning rate τ≥0 a parameter. As a consequence, when worker ℓ is employed her skills improve over time, as she becomes more experienced in her task. Conversely, unemployed workers lose skills. In particular, when a worker is hired, she may immediately acquire the minimum level of skills already present in the firm (the existing worker with the lowest skills), if above her present level. Also, workers have a fixed working life. After a fixed number of periods Tr∈ℕ* in the labor market, workers retire and are replaced by younger ones,10 whose skills are equivalent to the current minimum level in the incumbent firms. Workers’ skills define their individual (potential) productivity Aℓ,t: Aℓ,t=sℓ,ts¯tAiτ,  s¯t=1LS∑ℓsℓ,t, (5) where s¯t is the average worker skills level, and Aiτ is the warranted productivity of the machinery vintage the worker operates. The ratio sℓ,t/s¯t, or the worker normalized productivity, represents her ability to produce more (if sℓ,t>s¯t) or less (otherwise) when using a certain machine technology, in relation to the warranted vintage productivity. Note that the sectoral aggregation over the firm-level effective productivities Aj,t is a truly emergent properties of the model, resulting, simultaneously, from the technical innovation dynamics (mainly, the introduction of new vintages Aiτ), the worker skills accumulation/deterioration process, and the effective demand, which guides firms when deciding Qj,td, the capital stock dynamics, and the employed machine mix (see Appendix A for more details). The influence of the workers’ skills upon production reflects a learning by tenure/doing mechanism well established in the literature at least since the seminal contribution of Arrow (1962). On the empirical side, for the links between job tenure, capability accumulation, and firm productivity, see Zhou et al. (2011) and Lucidi and Kleinknecht (2009), among others. 3.4 Time line of events In each simulation time step, which can be taken to roughly represent a quarter, firms and workers behavioral rules are applied according to the following time line: machines ordered in the previous period (if any) are delivered; capital-good firms perform R&D and signal their machines to consumption-good firms; consumption-good firms decide on how much to produce, invest, and hire/fire; to fulfill production and investment plans, firms allocate cash flows, and (if needed) borrow from bank; firms send/receive machine-tool orders for the next period (if applicable); firms open job queues and job seekers send applications to them (“queue”); wages are set (indexation or bargaining) and job vacancies are partly or totally filled; workers (employed and unemployed) update their skills; government collects taxes and pays unemployment subsidies; consumption-good market opens and the market shares evolve according to competitiveness; firms in both sectors compute their profits, pay wages, and repay debt; exit takes place, firms with near-zero market share, or negative net assets are eschewed from the market; prospective entrants decide to enter according to the markets conditions; and aggregate variables are computed and the cycle restarts. 3.5 Alternative labor market policy regimes We employ the model described above to study two alternative policy regimes, which we call Fordist (our baseline) and Competitive.11 The policy regimes are telegraphically sketched in Table 1 (see Table B3 in Appendix B for the regime-specific parameter values). Table 1. Main characteristics of tested policy regimes Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Table 1. Main characteristics of tested policy regimes Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Fordist (baseline) Competitive Wage sensitivity to unemployment Low (rigid) High (flexible) Workers search activity Unemployed only Unemployed and employed Labor firing restrictions Under losses only None Workers hiring priority Higher skills Lower payback Workers firing priority Lower skills Higher payback Unemployment benefits Yes Yes (reduced) Minimum wage productivity indexation Full Partial Under the Fordist regime, wages are insensitive to the labor market conditions and indexed on a convex combination between economy-wide and firm-level productivity growth. There is a sort of covenant between firms and workers concerning “long term” employment: firms fire only when their profits become negative, while workers are loyal to employers and do not seek for alternative jobs. When hiring/firing, firms aim to keep the more skilled worker. Labor market institutions contemplate a minimum wage fully indexed to the aggregate economy productivity and unemployment benefits financed by taxes on profits. Conversely, in the Competitive regime, flexible wages respond to unemployment in a decentralized labor market dynamics, and are set by means of an asymmetric bargaining process where firms have the last say. Employed workers search for better paid jobs with some positive probability and firms freely adjust (fire) their excess workforce according to their planned production. Hiring/firing workers by firms are based on a trade-off between skills and wages, using a simple payback comparison rule. The Competitive regime is also characterized by different labor institutions: minimum wage is only partially indexed to productivity, and unemployment benefits—and the associated taxes on profits—are relatively lower. The simulation exercises in Section 5 are built so that there is a regime transition at a certain time step, capturing a set of labor market “structural reforms.” This institutional shock is aimed at spurring flexibility on the relations among agents in the labor market and implies that the social compromise embodied in the Fordist regime is replaced by the Competitive one. 4. Empirical validation The K + S model is able to generate endogenous growth and business cycles, emergent crises, and to reproduce a rich set of macro (e.g., relative volatility, co-movements, etc.) and micro (firm size distributions, firm productivity dynamics, etc.) stylized facts (see Dosi et al., 2010, 2013, 2015, 2017b). The detailed list of empirical regularities matched by the model is reported in Table 2. In addition, the labor-enhanced version of the model (Dosi et al., 2016b, 2017c), which explicitly accounts for microeconomic firm–worker interactions, has already proved to be able to robustly reproduce most of the labor market macro empirical regularities (cf. the bottom part of Table 2).12 Table 2. Stylized facts matched by the K + S model at different aggregation levels Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function Table 2. Stylized facts matched by the K + S model at different aggregation levels Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function Microeconomic stylized facts Aggregate-level stylized facts Skewed firm size distributions Endogenous self-sustained growth with persistent fluctuations Fat-tailed firm growth rates distributions Fat-tailed GDP growth rate distribution Heterogeneous productivity across firms Endogenous volatility of GDP, Consumption, and investment Persistent productivity differentials Cross-correlation of macro variables Lumpy investment rates of firms Pro-cyclical aggregate R&D investment Heterogeneous skills distribution Persistent and counter-cyclical unemployment Fat-tailed unemployment time distribution Endogenous volatility of productivity, unemployment, vacancy, separation, and hiring rates Unemployment and inequality correlation Pro-cyclical workers skills accumulation Beveridge curve Okun curve Wage curve Matching function The extensions to the K + S model proposed here add some new empirical regularities matched by the model. First, the new labor force learning dynamics produces fat-tailed worker-level skill distributions and firm-level productivity ones, consistent with the empirical evidence suggesting the presence of both firm- and worker-specific ample heterogeneity. Second, the more realistic entry dynamics increases the number of variables that match the cross-correlation/lag structures among aggregate macro indicators. Table 3 shows the time series correlation structure of firms net entry (entries minus exits) in the market, total bank debt, and average liquidity-to-sales ratios with respect to the GDP. These and the other macro variables correlation structures (not presented) seem reasonably in line with empirical evidence, as detailed in Dosi et al. (2010). Table 3. Correlation structure with respect to GDP on selected variables Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Note: All results significant at 5% level. MC standard errors in parentheses. Non-rate series are Baxter–King bandpass-filtered (6, 32, 12). Table 3. Correlation structure with respect to GDP on selected variables Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Fordist t−4 t−3 t−2 t−1 0 t + 1 t + 2 t + 3 t + 4 Net entry 0.09 0.13 0.14 0.07 −0.05 −0.18 −0.25 −0.25 −0.17 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.21 0.29 0.34 0.35 0.30 0.21 0.11 0.02 −0.03 (0.02) (0.03) (0.03) (0.04) (0.04) (0.04) (0.04) (0.03) (0.02) Liquidity-to-sales −0.12 −0.31 −0.52 −0.65 −0.66 −0.51 −0.26 −0.00 0.19 (0.03) (0.03) (0.02) (0.03) (0.03) (0.03) (0.02) (0.02) (0.02) Competitive t−4 t−3 t−2 t−1 0 t+1 t+2 t+3 t+4 Net entry 0.07 0.12 0.15 0.15 0.11 0.03 −0.07 −0.16 −0.21 (0.02) (0.02 (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) Total firm debt 0.11 0.11 0.08 0.03 −0.03 −0.08 −0.09 −0.07 −0.03 (0.03) (0.04 (0.04) (0.03) (0.03) (0.03) (0.02) (0.03) (0.03) Liquidity-to-sales −0.24 −0.50 −0.72 −0.85 −0.83 −0.64 −0.35 −0.02 0.25 (0.02) (0.01 (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) Note: All results significant at 5% level. MC standard errors in parentheses. Non-rate series are Baxter–King bandpass-filtered (6, 32, 12). 5. At the roots of hysteresis Let us study the emergence of hysteresis in our model, addressing its possible causes and discussing its consequences for the economic dynamics. We will first study inter-regime long-run hysteresis (cf. Figure 1). We will then analyze the emergence of intra-regime (transient) hysteresis (Section 5.2). 5.1 Regime change: super-hysteresis We begin with the long-run dynamics of the model, when affected by an institutional shock, namely, the introduction of “structural reforms” aimed at increasing the flexibility of the labor market, leaving however untouched the technological fundamentals. In our policy typology, the reforms are supposed to move the labor market regime from a Fordist to a Competitve setup (see Section 3.5 above). In that, we are also implicitly testing the insider–outsider hypothesis of hysteresis proposed by Blanchard and Summers (1987). In our model, the transition from a Fordist toward a Competitive type of labor relations captures the structural reforms, aimed at achieving both numerical (easier firing) and wage flexibility (wages more respondent to unemployment), as illustrated in Table 1.13 The normative implication of such hypothesis is the advocacy of a more flexible labor market, where unions have lower bargaining power in the wage formation process, with the aim of making wages more respondent to unemployment conditions. In Figure 3, we report the time series of the main macroeconomic variables in the two regimes.14 The institutional shock occurs at time t = 100 (the vertical dotted line). The widening GDP gap between the two regimes, as presented in Figure 3A, shows how the structural reforms determine super-hysteresis (i.e., a permanently lower growth rate of the GDP), whereby the effects propagate in the very long run (see also Dosi et al., 2017c, 2016b). The actual level of the long-run capacity utilization increases from the 85% to 90% after the introduction of the Competitive regime (cf. Figure 3B), hinting at a process of underinvestment due to the steeper fluctuations in investments opportunities for firms. In the Competitive regime, as a result of the more flexible wage dynamics, increased GDP volatility, and their effect on the aggregate demand, firms reduce their average expansionary investments, which depend on the difference between (demand-led) desired and installed production capacity (see Equation (19) in Appendix A), pushing down the number of machines ordered from the capital-good sector. Therefore, firms decrease the gap between the effective production and the potential capacity, leading to a cyclical surge in the capacity utilization which tends to yield self-rationing. Capital accumulation is slower when structural reforms are in place: the long-run growth rate falls from 1.55% to 1.44% per period. Figure 3C shows the dynamics of unemployment and vacancy rates, which are negatively correlated, consistent with a Beveridge Curve, while unemployment is significantly higher in the Competitive regime. The negative effects of structural reforms spill over the long-run: the number of successful innovations in the capital-good sector takes a lower trajectory (Figure 3D), and the average level of workers skills is significantly reduced (Figure 3E). Finally, the trend of the net entry15 of firms in the market is more turbulent after the reforms, also as a consequence of a higher level of volatility in credit conditions (Figure 3F).16 Figure 3. View largeDownload slide Macroeconomic dynamics in alternate policy regimes. Lines represent 50 MC runs averages (Fordist: black; Competitive: blue/gray). Figure 3. View largeDownload slide Macroeconomic dynamics in alternate policy regimes. Lines represent 50 MC runs averages (Fordist: black; Competitive: blue/gray). The different performance of the two regimes is quantitatively summarized in Table 4, which presents the averages, the ratios between selected variables of the two setups, and the P-values for a t test comparing the averages. The results confirm, at a 5% significance level, that after the introduction of structural reforms the short- and long-run performance of the economy significantly worsens. Note that as the technological configuration of the model is invariant between the two regime specifications, the differences in terms of productivity, innovation, and imitation rates are entirely caused by the institutional shock.17 Table 4. Comparison between policy regimes Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Note: Averages for 50 MC runs in period [200,400] (excluding warm-up). P-value for a two-means t-test, H0: no difference between regimes. Table 4. Comparison between policy regimes Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Time series Fordist (1) Competitive (2) Ratio (2)/(1) P-value GDP growth rate 0.0148 0.0135 0.9118 0.044 Capacity utilization 0.8712 0.9038 1.0374 0.000 Productivity growth rate 0.0147 0.0134 0.9084 0.034 Innovation rate 0.0937 0.0719 0.7677 0.001 Imitation rate 0.0253 0.0189 0.7476 0.004 Unemployment rate 0.0152 0.2640 17.400 0.000 Vacancy rate 0.0976 0.1439 1.4749 0.000 Worker tenure 27.861 4.9561 0.1779 0.000 Worker skills 1.7288 1.3418 0.7762 0.000 Wages standard deviation 0.0618 0.1710 2.7672 0.000 Note: Averages for 50 MC runs in period [200,400] (excluding warm-up). P-value for a two-means t-test, H0: no difference between regimes. What are the drivers of the soaring super-hysteresis in the model? The huge surge in unemployment reflects the widening gap between the long-run dynamics of real wages in the two regimes,18 which, in turn, leads to the emergence of Keynesian unemployment due to the contraction of aggregate demand, the slowdown in skills accumulation and actual productivity growth. Figure 4 shows the box-plot comparison between the Monte Carlo simulation runs for the two regimes, for the long-term consequences in terms of the innovation and imitation rates, productivity growth, job tenure, workers’ skills, and net entry of firms (see Section 2). The results in the first row of plots (Figure 4A–C) indicate a reduction in the innovation and imitation rates in the majority of the simulation runs—the latter variables are calculated as the rate of successful innovators and imitators in the capital-good sector—and, as a consequence, in the productivity growth rate. This is an indirect outcome of the fall in the aggregate demand, which yields lower R&D expenditure by firms.19 In the same direction, the results in the second row of Figure 4 show the quite significant fall on the average tenure period (Plot d) and the ensuing slower pace of the workers skills accumulation (Plot e), which, in turn, also has a direct and negative effect on the growth of productivity. Finally, the dynamics of net entry (number of entrants minus the exiting firms) is presented in Plot f.20 In the Competitive regime the financial cycle is amplified due to the increased volatility, exacerbating the entry dynamics: in good times there are more entrants in Competitive than in the Fordist regime, which exhibits a stabler financial cycle, while the opposite occurs in bad times. Both emergent phenomena, i.e. the more pronounced leverage cycle and the tighter availability of credit, have been empirically documented by Ng and Wright (2013) from the last three recessions (1990, 2001, and 2007). Figure 4. View largeDownload slide Performance comparison between policy regimes (Fordist: left; Competitive: right). Summary statistics for 50 MC runs in period [200,400] (excluding warm-up). Bar: median; box: second to third quartile; whiskers: maximum–minimum; dots: outliers. Figure 4. View largeDownload slide Performance comparison between policy regimes (Fordist: left; Competitive: right). Summary statistics for 50 MC runs in period [200,400] (excluding warm-up). Bar: median; box: second to third quartile; whiskers: maximum–minimum; dots: outliers. The transmission channels in the model operate through both numerical and wage flexibility. First, higher numerical flexibility, where workers are more freely fired, determines a sharp drop in workers job tenure and, indirectly, has a negative effect on skills accumulation and, consequently, on productivity. Not only the firing rule but also the firing order criteria affect the dynamics of productivity growth. In the Fordist regime, firms first hire (fire) workers with higher (lower) skills.21 Conversely, in the Competitive case, firms use the skills-to-wage “payback” ratio as a decision guide to preferentially hire (fire) workers with superior (inferior) short-term “returns.” Such a behavior has a negative impact on the aggregate skill level of the incumbent workers over time. On the other hand, higher wage flexibility, by limiting the wage indexation upon the productivity gains, causes a straightforward drop in the aggregate demand via the reduced consumption of workers. In turn, the shrinking sales opportunities drive a fall in investment and labor demand, which induces more unemployment, characterizing a typical Keynesian feedback-amplified downturn. Moreover, the slower economy also impacts upon the entry/exit and the innovation/imitation rates, via the overall cut in total R&D expenditure and the higher volatility in the number of operating firms. In fact, Table 3 shows the significant level of correlation between the business cycle and the net entry of firms in the market. The severe effects of super-hysteresis are particularly well illustrated by the probability distributions for the time unemployed workers need to find a new job, presented in Figure 5.22 As shown by the huge increase in the distribution support, long-term unemployment is by far higher in the Competitive case.23 Figure 5. View largeDownload slide Actual probability distribution vs normal fit of worker unemployment time. Each t corresponds to a quarter. Figure 5. View largeDownload slide Actual probability distribution vs normal fit of worker unemployment time. Each t corresponds to a quarter. To sum up, our experiments generically yield super-hysteresis stemming from an institutional shock. Indeed, institutions are a “carrier of history” (David, 1994) also here. However, contrary to the insider–outsider hypothesis (Blanchard and Summers, 1987), “pro-market” institutions bear a negative hysteretic effect. The model suggests that structural reforms aimed at increasing the flexibility in the labor market may well spur even more hysteresis instead of reducing it. Granted that, in the next section, we focus on intra-regime hysteresis phenomena. 5.2 Detecting intra-regime hysteresis Assessing the emergence of intra-regime hysteresis is not a trivial task, as there is no conclusive test or even widely accepted criteria for this. However, there are several properties and techniques which do help uncover particular aspects of hysteresis. In the following, we present a set of analytical methods, summarized in Table 5, which provide evidence of the presence of hysteretical properties in the K + S model. In line with the literature, we study whether the time series generated by the model present evidence of (i) remanence, (ii) persistency, (iii) nonlinearity, (iv) path dependency, and (v) super-hysteresis. Needless to say, these properties are to some degree overlapping. Table 5. Selected tests to evaluate hysteretic properties in times series Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Table 5. Selected tests to evaluate hysteretic properties in times series Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Property Test Reference Remanence Duration of recovery of employment and GDP after crises Jaimovich and Siu, 2012 Persistency Unit-root tests for stationarity Blanchard and Summers, 1986 Nonlinearity BDS test Broock et al., 1996 Path dependence Ergodicity tests Wald and Wolfowitz, 1940 Super-hysteresis Different GDP growth trend (slope) after crises Blanchard et al., 2015 Figure 6 illustrates the number of periods (gray area) necessary to put the economy back to the pre-crisis growth trend (dashed line) in typical simulation runs.24 The analysis is inspired by previous work (Blanchard et al., 2015) and simply performs an extrapolation of the long-run GDP trend to detect the recovery from crises under the possible presence of hysteresis. The results show the coexistence of shorter business cycle downturns with longer, hysteretical crises, requiring significant more times for the economy to recover. Note also the presence of super-hysteresis, revealed by the different slopes of the peak-to-peak GDP trends (dashed lines). Figure 6. View largeDownload slide GDP recovery after crises. Typical simulation runs. Dashed line: pre-crisis trends; gray boxes: recovery periods. Figure 6. View largeDownload slide GDP recovery after crises. Typical simulation runs. Dashed line: pre-crisis trends; gray boxes: recovery periods. Table 6 reports the average recovery duration for both the GDP and the mean unemployment time (the average period a worker takes to find a new job). While the duration of the GDP trend recovery is similar among regimes (around 16 quarters), the unemployment time takes almost five times more to return to its pre-crisis level in the Competitive case. To better assess the severity of the crises, we also track the peak GDP trend deviation during the recovery period (the farther the GDP gets from the pre-crisis trend) and the accumulated GDP losses in comparison to the trend (the crisis “cost”). The model robustly shows how Competitive regime crises are about twice deeper than in the Fordist scenario. The accumulated GDP loss comparison leads to a similar conclusion. Table 6. Comparison between policy regimes: crises, GDP and unemployment time recovery Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Note: Averages for 50 MC runs in period [200,400] (excluding warm-up), MC standard errors in parentheses. Table 6. Comparison between policy regimes: crises, GDP and unemployment time recovery Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Fordist Competitive Number of crises 6.15 5.77 (0.44) (0.28) Crises peak 0.23 0.51 (0.01) (0.02) Crises losses 2.38 4.18 (0.33) (0.42) Recovery duration  -GDP 15.64 16.97 (1.43) (1.04)  -Unemployment time 6.83 31.22 (0.55) (9.04) Note: Averages for 50 MC runs in period [200,400] (excluding warm-up), MC standard errors in parentheses. In Table 7, we report a set of statistical tests to detect unit-roots vs. stationarity (Augmented Dickey–Fuller/ADF, Phillips–Perron/PP, and Kwiatkowski–Phillips–Schmidt–Shin/KPSS tests), i.i.d. vs. nonlinear processes (Brock–Dechert–Scheinkman/BDS test), and ergodicity (Kolmogorov–Smirnov/KS and Wald–Wolfowitz/WW tests). Except for the WW case, the tests are applied for individual Monte Carlo simulation runs (or multiple run-pair combinations, in the case of KS) and, so, the results present the frequency of the rejection of the null hypothesis for the set of 50 runs at the usual 5% significance level (see Table 7 for the definition of H0 in each case). Table 7. Comparison between policy regimes: statistical tests for detecting hysteresis Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Note: Frequencies of rejection of H0 for 50 MC runs in period [300,350] (excluding warm-up) except for WW test (P-value presented), at 5% significance. ADF/PP H0: non-stationary—KPSS) H0: stationary—BDS H0: i.i.d., KS/WW H0: ergodic. Table 7. Comparison between policy regimes: statistical tests for detecting hysteresis Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Fordist ADF PP KPSS BDS KS WW GDP growth rate 0.80 1.00 0.00 0.30 0.23 0.00 Productivity growth rate 0.76 1.00 0.02 0.44 0.12 0.00 Wage growth rate 0.60 1.00 0.12 0.16 0.40 0.00 Unemployment rate 0.40 0.60 0.16 0.50 0.33 0.01 Competitive ADF PP KPSS BDS KS WW GDP growth rate 0.54 0.98 0.00 0.42 0.11 0.00 Productivity growth rate 0.64 1.00 0.02 0.62 0.19 0.00 Wage growth rate 0.42 1.00 0.14 0.30 0.38 0.02 Unemployment rate 0.24 0.00 0.26 1.00 0.49 0.00 Note: Frequencies of rejection of H0 for 50 MC runs in period [300,350] (excluding warm-up) except for WW test (P-value presented), at 5% significance. ADF/PP H0: non-stationary—KPSS) H0: stationary—BDS H0: i.i.d., KS/WW H0: ergodic. The results suggest that GDP, productivity, and wage growth rates more frequently exhibit stationary (no unit-roots) behaviors in both regimes. More borderline, the unemployment rate time series seems to be more commonly stationary among simulation runs in the Fordist regime, while more likely non-stationary in the Competitive case. The nonlinearity test indicates a quite nuanced situation: the unemployment series is the one more frequently displaying nonlinear structure, particularly in the Competitive regime, while the wage growth rates series seem more likely to be i.i.d. processes. Finally, the less powerful KS test cannot reject ergodicity for the majority of run pairs tested, while WW indicates non-ergodicity of all series. There are a few take-home messages from the tests. First, that mixed results, e.g., on ergodicity and stationarity, militate as such in favor of path dependency. In fact, they show the different statistical properties of alternative sample-paths: only an outright non-rejection of the null hypothesis could be claimed in support of the the lack of hysteresis. Second, but related, the tests aimed at the detection of some underlying, emergent, nonlinear structure are quite encouraging despite the limited length of the sample paths.25 Finally, we perform a global sensitivity analysis (SA) to explore the effects of alternative model parametrizations and to gain further insights on the robustness of our exercises on institutional shocks.26 Of the 57 parameters and initial conditions in the K + S model, we reduce the relevant parametric dimensionality to 29, by means of an Elementary Effect screening procedure which allows to discard from the analysis the parameters which do not significantly affect the selected model outputs.27 All the parameters tested in the SA, their “calibration” values, as well the tests statistics, are detailed in Table B1 (Appendix B). To understand the effect of each of the 29 parameters over the selected metrics, we perform a Sobol decomposition.28 Because of the relatively high computational costs to produce the decomposition using the original model, a simplified version of it—a meta-model—was built using the Kriging method and employed for the Sobol SA.29 The meta-model is estimated by numerical maximum likelihood using a set of observations (from the original model) sampled using a high-efficiency, nearly orthogonal Latin hypercube design of experiments (Cioppa and Lucas, 2007). The main indicator used for the SA is the accumulated GDP losses during the crises’ recovery periods (the crisis “cost”), as defined above. It seems a sensible choice, as it conveys information about both the duration and the intensity of the crises, as such among the key properties of hysteresis. Interestingly, this indicator is significantly influenced only by a limited set of parameters, and by no initial condition, including the learning rate parameter (τ), the retirement age (Tr), the replicator equation parameter (χ), the maximum technical advantage of the capital-good entrants (x5), the minimum capital ratio ( Φ1) and the expected capacity utilization (u) of the consumption-good entrants.30 The two parameters associated with the skills accumulation process, learning rate (τ) and retirement age (Tr), are jointly responsible for almost 80% of the variance of the losses indicator over the entire parametric space in both policy regimes.31 Figure 7A and B presents an exploration of the model response surface, using the Kriging meta-model, for the two critical skill-related parameters. The rugged surfaces, in particular in the Competitive regime, clearly indicate the nonlinear nature of the system, in tune with the hysteretic properties of the model. The SA seems to suggest that the prominent parameters influencing the level of hysteresis observed in the losses indicator are those directly connected with the workers skills accumulation process (τ and Tr), the firm entry mechanism ( Φ1, u and x5), and the market competitiveness (χ). Directly or in interaction among them, these five parameters account for 95% of the variation of the GDP crises losses in the model for the two scenarios. As can be seen in Figure 7, the Competitive regime tends to produce significant higher GDP crises losses irrespective of the model setup (notice that the peak losses in Plot (a), are at a lower z-axis level than the deepest valley in Plot (b)). Finally, the response surfaces in both regimes show that in general the higher the learning rate (τ), the higher is the accumulated GDP losses during the crises’ recovery periods. The latter positive marginal effect hints at the fact that the higher the firm-specific capabilities, the more difficult is to rebuild the workers skills destroyed by a crisis, turning back to the pre-crisis level. Figure 7. View largeDownload slide Global SA: response surfaces. Surfaces modelled using the Kriging meta-model. z-axis: recovery losses (rec_loss). Dot: calibration settings. Markers: maximum and minimum predicted crises losses. Figure 7. View largeDownload slide Global SA: response surfaces. Surfaces modelled using the Kriging meta-model. z-axis: recovery losses (rec_loss). Dot: calibration settings. Markers: maximum and minimum predicted crises losses. All in all, the statistical tests results indicate that model has a rather frequent tendency to show the properties usually associated with hysteresis in its main variables, in particular the unemployment rate, whenever hit by an endogenously produced crisis. Recoveries can take quite long times, and the losses experienced by the economy, both in terms of the GDP and the social cost of unemployment, are severe. It is also significant that such losses seem to increase after the introduction of structural reforms of the type discussed above. 6. Conclusions The Great Recession has forced a revival of the notion of hysteresis, as such a short-hand for the possibility of multiple equilibria/paths either in some transient periods or even in the longer term. The evidence has been overwhelming: not only the level trends of GDP and unemployment but even the growth rates in many countries are still persistently below the pre-2008 figures. Older candidates for the interpretation of such a behavior are unit-root processes in unemployment—as originally suggested by Blanchard and Summers (1986). But such interpretations are rather fragile in that they postulate the source of hysteresis in some deviations of reality from the standard frictionless model, e.g. the insider–outsider labor market rigidities. Here we have analyzed an opposite perspective. In tune with an expanding tradition of scholars, we have discussed the notions of hysteresis and path dependence, identifying in coordination failures and persistent effects of aggregate demand upon productivity the main sources of long-term jumps across multiple growth trajectories. In doing that, we have presented an ABM which intertwines a Schumpeterian engine of growth and a Keynesian generation of demand, declined under two institutional labor market variants, labeled as Fordist and Competitive regimes. The transition from the Fordist to the Competitive regime captured “structural reforms” aimed at increasing labor market flexibility. Does the latter reduce hysteresis? Not at all. The model is able to generically exhibit path dependence, nonlinearity, and non-ergodicity in its main macroeconomic variables, presenting both inter-regime and intra-regime hysteresis as a bottom-up emergent property. Moreover, the model suggests that both numerical and wage flexibility are quite prone to increase the hysteretic properties of the macroeconomic system. The K + S model leaves scope for many potential avenues for further research, addressing the links between the functioning of the capital, consumer, and labor markets. In particular, a straightforward extension of the current article would be the study of the effects of active labor market policies, declined under alternative training programs and hiring/firing schemes. Yet, another venue of research concerns the effect of hysteresis upon labor force participation. Acknowledgements The authors thank the Guest Editors of the Special Section; two anonymous referees, Engelbert Stockhammer and Federico Tamagni; and the participants to the International Conference Economics, Economic Policy and Sustainable Growth after the Crisis (Ancona, 2016), the 9th Workshop on Complex Evolving System Approach in Economics (Nice, 2017), the Workshop on Economic Growth, Macroeconomic Dynamics, and Agents Heterogeneity (Saint-Petersburg, 2017), the 10th EMAEE Conference (Strasbourg, 2017), the 22nd WEHIA (Milan, 2017), the 23rd CEF International Conference (New York, 2017), the 29th SASE Annual Meeting (Lyon, 2017), and the 29th Annual EAEPE Conference (Budapest, 2017). G.D., A.R., and M.E.V. gratefully acknowledge the support by the European Union’s Horizon 2020 research and innovation program under grant agreement number 649186 - ISIGrowth. Footnotes 1 See the presentations in Tesfatsion and Judd (2006); LeBaron and Tesfatsion (2008), and Fagiolo and Roventini (2012, 2017) for critical surveys on macro agent-based models (ABMs). For related ABMs which consider a decentralized labor market, see Dawid et al. (2014), Russo et al. (2016), Caiani et al. (2016a), and Caiani et al. (2016b), among the others. See also Bassi and Lang (2016) for an agent-based model with investment hysteresis. 2 For a more detailed discussion on the relationship between hysteresis and path dependency, see Castaldi and Dosi (2006); Setterfield (2009). 3 The hysteretic properties of economic systems are also studied in the Post Keynesian literature: see Skott (2005); Stockhammer (2011) which identify wage norms and fairness as a potential mechanism to produce an endogenous non-accelerating inflation rate of unemployment (NAIRU). 4 See Dosi and Virgillito (2018) for a further discussion. 5 For an enticing reconstruction of the discussion between Frisch and Schumpeter on the pendulum metaphor, see Louca (2001). 6 The order the channels are presented is not relevant in terms of the impact produced by each one upon hysteresis. 7 The subscript t indicates time dependence. From now on, agent-specific variables are denoted by a subscript i, in case of capital-good firms, j, for consumption-good firms, or ℓ, for workers. 8 The latter is a usual simplifying assumption which avoids the introduction of a multilevel capital-goods sector and keeps the innovation process more transparent (see, on a three-sector economy, Seppecher et al., 2017). 9 In the consumption-good market the replicator captures the law of motion which regulates the pace of adjustments in market shares among the ever-changing heterogeneous competitors. On the robustness of the empirical properties generated by the replicator dynamics, see Dosi et al. (2017e). 10 In the start of the simulation, initial workers ages are randomly drawn in the integer range [1,Tr] and all start from the same skills level. 11 The two regimes roughly capture two alternative wage-labor nexus in the language of the Regulation Theory (see, within a vast literature, Boyer and Saillard, 2005 and Amable, 2003). 12 For a detailed discussion upon the configurations and the parameter settings producing the above mentioned stylized facts, we refer to Dosi et al. (2010, 2017c). In the following we focus on the innovation, entry, and skills processes, and on the related variables and parameters. 13 Indeed, the change of the political structure and of the balance of power between capitalists and workers and the related results of a class struggle are phenomena which, while with profound economic roots, did occur at the sociopolitical level: the Thatcher–Reagan regime change has been an exogenous political transformation. Modeling the triggering mechanisms leading to the end of the welfare system is thus well beyond the scope of this article. 14 The presented series are the averages of 50 Monte Carlo simulation runs, over 500 periods. The initial 100 “warm-up” periods are not presented. 15 Note that the use of the two-sided Hodrick-Prescott (HP) filter artificially produces the diverging patterns of the two curves before time 100. 16 As discussed in Section 3.2, entry decision in the model is also driven by the average financial conditions of the firms in each sector. 17 In accordance with the behavioral rules set in the model (cf. Appendix A), the dynamics of innovation, of imitation, of new machines introduction, and, consequently, of the firms productivity growth is directly affected by the overall macroeconomic conditions, including those directly impacted by the reforms. This creates a (potentially hysteretic) reinforcing feedback process between the macro and the technological domains, which in part explains the observed results. 18 The real wage growth rates are 1.47% and 1.35% per period, respectively. 19 See Appendix A for details on the innovation process. 20 The diverging trend before time t = 100 is due to the two-sided HP filter we employ to detrend the series. 21 This is a necessary consequence of the firms unilaterally decided and homogeneously applied wage adjustments, so skills are the only heterogeneous metric among workers in a Fordist firm. 22 Note the log scale in the y-axis. 23 The maximum notional unemployment time is 120 periods, equivalent to the working life in the model (parameter Tr). 24 A crisis is defined by a 3% drop of the GDP in a single period which is not recovered in the next three periods. The pre-crisis level is calculated as the average GDP for the four periods before the crisis and the trend, and as the output of an HP filter at the period just before the crisis. The crisis is considered recovered when the GDP reaches back the pre-crisis trend level. 25 The choice of the adequate time window length is quite relevant when analyzing hysteresis, as detailed in Section 2.1, and it is not driven by the availability of simulated data. For this reason we split the analysis in inter-regime hysteresis, where the patterns are of long-term type, and the intra-regime hysteresis. The analysis used in the literature to detect hysteresis is always performed for relatively short time spans (usually under 20 years). For comparability with empirical data, to check for the intra-regime hysteresis we restricted the time span to 50 runs, which is closer to the empirical time horizons. Note that taking longer time spans would simply “dilute” some hysteretic properties of the series, like non-ergodicity or non-stationarity. 26 For technical details on the global SA methodology, see Dosi et al. (2017d). 27 Briefly, the Elementary Effects technique proposes both a specific design of experiments, to efficiently sample the parameter space under a one-factor-at-a-time, and some linear regression statistics, to evaluate direct and indirect (nonlinear/non-additive) effects of parameters on the model results (Morris, 1991, Saltelli et al., 2008). 28 The Sobol decomposition is a variance-based, global SA method consisting in the decomposition of the variance of the chosen model output into fractions according to the variances of the parameters selected for analysis, better dealing with nonlinearities and non-additive interactions than traditional local SA methods. It allows to disentangle both direct and interaction quantitative effects of the parameters on the chosen metrics (Sobol, 1993, Saltelli et al., 2008). 29 Summarizing, the Kriging meta-model “mimics” our original model by a simpler, mathematically tractable approximation. Kriging is an interpolation method that under fairly general assumptions provides the best linear unbiased predictors for the response of complex, nonlinear computer simulation models (Rasmussen and Williams, 2006, Salle and Yıldızoğlu, 2014). 30 All the equations and parameters are described in Sections 3.1 and 3.3 and in the Appendices. 31 The parameters calibration values, valid ranges, and the Sobol decomposition results are presented in Table B1 in Appendix B. References Abraham K. G. , Haltiwanger J. C. , Sandusky K. , Spletzer J. ( 2016 ), ‘The consequences of long term unemployment: Evidence from matched employer-employee data,’ NBER Working Paper 22665. National Bureau of Economic Research: Cambridge (MA), US . Amable B. ( 2003 ), The Diversity of Modern Capitalism . Oxford University Press : Oxford . Google Scholar CrossRef Search ADS Amable B. , Henry J. , Lordon F. , Topol R. ( 2004 ), ‘Complex remanence vs. simple persistence: are hysteresis and unit-root processes observationally equivalent?,’ in Barnett W. , Deissenberger C. , Feichtinger G. , (eds), Economic Complexity . Emerald Group : Bingley, UK , pp. 67 – 89 . Google Scholar CrossRef Search ADS Anderson P. , Arrow K. , Pines D. ( 1988 ), The Economy as a Complex Evolving System . Addison-Wesley : Boston, US . Arrow K. ( 1996 ), ‘ Technical information and industrial structure ,’ Industrial and Corporate Change , 5 ( 2 ), 645 – 652 . Google Scholar CrossRef Search ADS Arrow K. J. ( 1962 ), ‘ The economic implications of learning by doing ,’ The Review of Economic Studies , 29 ( 3 ), 155 – 173 . Google Scholar CrossRef Search ADS Arthur W. B. ( 1989 ), ‘ Competing technologies, increasing returns, and lock-in by historical events ,’ The Economic Journal , 99 ( 394 ), 116 – 131 . Google Scholar CrossRef Search ADS Ball L. , DeLong B. , Summers L. ( 2014 ), Fiscal policy and full employment, https://www.cbpp.org/research/full-employment/fiscal-policy-and-full-employment. Ball L. M. ( 2014 ), ‘Long-term damage from the great recession in OECD countries,’ NBER Working Paper 20185. National Bureau of Economic Research: Cambridge (MA), US. Bassi F. , Lang D. ( 2016 ), ‘ Investment hysteresis and potential output: a post-Keynesian–Kaleckian agent-based approach ,’ Economic Modelling , 52 , 35 – 49 . Google Scholar CrossRef Search ADS Beaudry P. , Galizia D. , Portier F. ( 2016 ), ‘Is the macroeconomy locally unstable and why should we care?,’ NBER Working Paper 22275. National Bureau of Economic Research: Cambridge (MA), US . Bertrand M. , Schoar A. , Thesmar D. ( 2007 ), ‘ Banking deregulation and industry structure: evidence from the French banking reforms of 1985 ,’ The Journal of Finance , 62 ( 2 ), 597 – 628 . ISSN 1540-6261. Google Scholar CrossRef Search ADS Blanchard O. J. , Leigh D. . ( 2013 ), ‘ Growth forecast errors and fiscal multipliers ,’ American Economic Review , 103 , 117 – 120 . Blanchard O. , Cerutti E. , Summers L. ( 2015 ), ‘Inflation and activity–two explorations and their monetary policy implications,’ NBER Working Paper 21726. National Bureau of Economic Research: Cambridge (MA), US . Blanchard O. J. , Summers L. H. ( 1986 ), ‘ Hysteresis and the European unemployment problem ,’ NBER Macroeconomics Annual , 1 , 15 – 78 . Google Scholar CrossRef Search ADS Blanchard O. J. , Summers L. H. ( 1987 ), ‘ Hysteresis in unemployment ,’ European Economic Review , 31 ( 1–2 ), 288 – 295 . Google Scholar CrossRef Search ADS Boyer R. , Saillard Y. ( 2005 ), Régulation Theory: The State of the Art . Routledge : London , Broock W. , Scheinkman J. A. , Dechert W. D. , LeBaron B. ( 1996 ), ‘ A test for independence based on the correlation dimension ,’ Econometric Reviews , 15 ( 3 ), 197 – 235 . Google Scholar CrossRef Search ADS Caiani A. , Godin A. , Caverzasi E. , Gallegati M. , Kinsella S. , Stiglitz J. E. ( 2016a ), ‘ Agent based-stock flow consistent macroeconomics: towards a benchmark model ,’ Journal of Economic Dynamics and Control , 69 , 375 – 408 . Google Scholar CrossRef Search ADS Caiani A. , Russo A. , Gallegati M. ( 2016b ), Does Inequality Hamper Innovation and Growth? Available at SSRN 2790911. Castaldi C. , Dosi G. ( 2006 ), ‘The grip of history and the scope for novelty: some results and open questions on path dependence in economic processes,’ in Wimmer A. , Kössler R. , editors, Understanding Change . Springer : NY, US . pp. 99 – 128 . Google Scholar CrossRef Search ADS Cioppa T. M. , Lucas T. W. ( 2007 ), ‘ Efficient nearly orthogonal and space-filling latin hypercubes ,’ Technometrics , 49 ( 1 ), 45 – 55 . Google Scholar CrossRef Search ADS David P. A. ( 1985 ), ‘ Clio and the economics of qwerty ,’ The American Economic Review , 75 ( 2 ), 332 – 337 . David P. A. ( 1994 ), ‘ Why are institutions the ‘carriers of history’?: Path dependence and the evolution of conventions, organizations and institutions ,’ Structural Change and Economic Dynamics , 5 ( 2 ), 205 – 220 . Google Scholar CrossRef Search ADS Dawid H. , Harting P. , Neugart M. ( 2014 ), ‘ Economic convergence: policy implications from a heterogeneous agent model ,’ Journal of Economic Dynamics and Control , 44 , 54 – 80 . Google Scholar CrossRef Search ADS Day R. H. , Chen P. ( 1993 ), Nonlinear Dynamics and Evolutionary Economics . Oxford University Press : Oxford , Decker R. A. , Haltiwanger J. , Jarmin R. S. , Miranda J. ( 2016 ), ‘Declining business dynamism: Implications for productivity?,’ Hutchins Center Working Paper. Brookings Institution: Cambridge (MA), US . Dosi G. , Kaniovski Y. ( 1994 ), ‘ On “badly behaved” dynamics ,’ Journal of Evolutionary Economics , 4 ( 2 ), 93 – 123 . Google Scholar CrossRef Search ADS Dosi G. , Virgillito M. E. (2018), ‘The “Schumpeterian” and the “Keynesian” Stiglitz: learning, coordination hurdles and growth trajectories,’ in Guzman M. (ed.), Toward a Just Society. Joseph Stiglitz and Twenty-First Century Economics . Columbia University Press , NY, US. Dosi G. , Fagiolo G. , Roventini A. ( 2010 ), ‘ Schumpeter meeting Keynes: a policy-friendly model of endogenous growth and business cycles ,’ Journal of Economic Dynamics and Control , 34 ( 9 ), 1748 – 1767 . Google Scholar CrossRef Search ADS Dosi G. , Fagiolo G. , Napoletano M. , Roventini A. ( 2013 ), ‘ Income distribution, credit and fiscal policies in an agent-based Keynesian model ,’ Journal of Economic Dynamics and Control , 37 ( 8 ), 1598 – 1625 . Google Scholar CrossRef Search ADS Dosi G. , Fagiolo G. , Napoletano M. , Roventini A. , Treibich T. ( 2015 ), ‘ Fiscal and monetary policies in complex evolving economies ,’ Journal of Economic Dynamics and Control , 52 , 166 – 189 . Google Scholar CrossRef Search ADS Dosi G. , Napoletano M. , Roventini A. , Treibich T. ( 2016a ), ‘The short-and long-run damages of fiscal austerity: keynes beyond Schumpeter,’ in Stiglitz J. , Guzman M. (eds), Contemporary Issues in Macroeconomics . Palgrave Macmillan , pp. 79 – 100 . Dosi G. , Pereira M. C. , Roventini A. , Virgillito M. E. ( 2016b ), ‘The effects of labour market reforms upon unemployment and income inequalities: an agent based model,’ LEM Papers Series 2016/27. Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced Studies: Pisa, Italy. Dosi G. , Napoletano M. , Roventini A. , Stiglitz J. , Treibich T. ( 2017a ), Rational Heuristics? Expectation and Behavior in Evolving Economies with Heterogeneous Interacting Agents. Lem Paper Series Forthcoming, Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced Studies. Dosi G. , Napoletano M. , Roventini A. , Treibich T. ( 2017b ), ‘ Micro and macro policies in the Keynes + Schumpeter evolutionary models ,’ Journal of Evolutionary Economics , 27 ( 1 ), 63 – 90 . Google Scholar CrossRef Search ADS Dosi G. , Pereira M. C. , Roventini A. , Virgillito M. E. ( 2017c ), ‘ When more flexibility yields more fragility: the microfoundations of Keynesian aggregate unemployment ,’ Journal of Economic Dynamics and Control , 81 , 162 – 186 . Google Scholar CrossRef Search ADS Dosi G. , Pereira M. C. , Virgillito M. E. ( 2017d ), ‘ On the robustness of the fat-tailed distribution of firm growth rates: a global sensitivity analysis ,’ Journal of Economic Interaction and Coordination . https://doi.org/10.1007/s11403-017-0193-4. Dosi G. , Pereira M. C. , Virgillito M. E. ( 2017e ), ‘ The footprint of evolutionary processes of learning and selection upon the statistical properties of industrial dynamics ,’ Industrial and Corporate Change , 26 ( 2 ), 187 – 210 . Fagiolo G. , Roventini A. ( 2012 ), ‘ Macroeconomic policy in DSGE and Agent-Based Models ,’ Revue de l’OFCE , 124 ( 5 ), 67 – 116 . Fagiolo G. , Roventini A. ( 2017 ), ‘ Macroeconomic policy in DSGE and Agent-Based Models redux: new developments and challenges ahead ,’ Journal of Artificial Societies and Social Simulation , 20 ( 1 ), 1 . Google Scholar CrossRef Search ADS Freeman C. , Clark J. , Soete L. ( 1982 ), Unemployment and Technical Innovation . Pinter Publisher : London and New York , Galí J. ( 2015 ), ‘Hysteresis and the European unemployment problem revisited,’ NBER Working Paper 21430. National Bureau of Economic Research: Cambridge (MA), US . Geroski P. ( 1991 ), Market Dynamic and Entry . Wiley-Blackwell , Geroski P. ( 1995 ), ‘ What do we know about entry?, ’ International Journal of Industrial Organization , 13 ( 4 ), 421 – 440 . The Post-Entry Performance of Firms. Google Scholar CrossRef Search ADS Ghayad R. ( 2013 ), ‘The jobless trap,’ Technical report. Northeastern University: Boston, US . Gomis R. M. , Khatiwada S. ( 2017 ), ‘Firm dynamics and business cycle: What doesn’t kill you makes you stronger?,’ Technical report. Economics Section, The Graduate Institute of International Studies: Geneva (CH). Gourio F. , Messer T. , Siemer M. ( 2016 ), ‘A missing generation of firms? Aggregate effects of the decline in new business formation.’ In 2016 Meeting Papers (No. 752). Society for Economic Dynamics , Toulouse (FR) . Gualdi S. , Tarzia M. , Zamponi F. , Bouchaud J.-P. ( 2015 ), ‘ Tipping points in macroeconomic agent-based models ,’ Journal of Economic Dynamics and Control , 50 , 29 – 61 . Google Scholar CrossRef Search ADS Hallett A. H. , Piscitelli L. ( 2002 ), ‘ Testing for hysteresis against nonlinear alternatives ,’ Journal of Economic Dynamics and Control , 27 ( 2 ), 303 – 327 . Google Scholar CrossRef Search ADS Jaimovich N. , Siu H. E. ( 2012 ), ‘The trend is the cycle: Job polarization and jobless recoveries,’ NBER Working Paper 18334. National Bureau of Economic Research: Cambridge (MA), US . Kerr W. R. , Nanda R. ( 2009 ), ‘ Democratizing entry: banking deregulations, financing constraints, and entrepreneurship ,’ Journal of Financial Economics , 94 ( 1 ), 124 – 149 . Google Scholar CrossRef Search ADS Kothari S. P. , Lewellen J. , Warner J. ( 2014 ), ‘The behavior of aggregate corporate investment,’ Technical Report 5112-14. MIT Sloan Research Paper. Lamberson P. , Page S. ( 2012 ), ‘ Tipping points ,’ Quarterly Journal of Political Science , 7 ( 2 ), 175 – 208 . Google Scholar CrossRef Search ADS LeBaron B. , Tesfatsion L. ( 2008 ), ‘ Modeling macroeconomies as open-ended dynamic systems of interacting agents ,’ American Economic Review , 98 ( 2 ), 246 – 250 . Google Scholar CrossRef Search ADS Lee Y. , Mukoyama T. ( 2015 ), ‘ Entry and exit of manufacturing plants over the business cycle ,’ European Economic Review , 77 , 20 – 27 . Google Scholar CrossRef Search ADS Lorenz H.-W. ( 1993 ), Nonlinear Dynamical Economics and Chaotic Motion , vol. 334 . Springer : NY, US . Google Scholar CrossRef Search ADS Louca F. ( 2001 ), ‘ Intriguing pendula: founding metaphors in the analysis of economic fluctuations ,’ Cambridge Journal of Economics , 25 , 25 – 55 . Google Scholar CrossRef Search ADS Lucidi F. , Kleinknecht A. ( 2009 ), ‘ Little innovation, many jobs: an econometric analysis of the Italian labour productivity crisis ,’ Cambridge Journal of Economics , 34 ( 3 ), 525 – 546 . Google Scholar CrossRef Search ADS Morris M. D. ( 1991 ), ‘ Factorial sampling plans for preliminary computational experiments ,’ Technometrics , 33 ( 2 ), 161 – 174 . Google Scholar CrossRef Search ADS Napoletano M. , Dosi G. , Fagiolo G. , Roventini A. ( 2012 ), ‘ Wage formation, investment behavior and growth regimes: an agent-based analysis ,’ Revue de l’OFCE , 5 ( 124 ), 235 – 261 . Ng S. , Wright J. H. ( 2013 ), ‘ Facts and challenges from the recession for forecasting and macroeconomic modeling ,’ Journal of Economic Literature , 51 ( 4 ), 1120 – 1154 . Google Scholar CrossRef Search ADS Piscitelli L. , Cross R. , Grinfeld M. , Lamba H. ( 2000 ), ‘ A test for strong hysteresis ,’ Computational Economics , 15 ( 1/2 ), 59 – 78 . Google Scholar CrossRef Search ADS Rasmussen C. , Williams C. ( 2006 ), Gaussian Processes for Machine Learning. MIT Press : Cambridge, MA . Reifschneider D. , Wascher W. , Wilcox D. ( 2015 ), ‘ Aggregate supply in the United States: recent developments and implications for the conduct of monetary policy ,’ IMF Economic Review , 63 ( 1 ), 71 – 109 . Google Scholar CrossRef Search ADS Rosser J. B. ( 2013 ), From Catastrophe to Chaos: A General Theory of Economic Discontinuities . Springer Science & Business Media : NY, US . Russo A. , Riccetti L. , Gallegati M. ( 2016 ), ‘ Increasing inequality, consumer credit and financial fragility in an agent based macroeconomic model ,’ Journal of Evolutionary Economics , 26 ( 1 ), 25 – 47 . Google Scholar CrossRef Search ADS Salle I. , Yıldızoğlu M. ( 2014 ), ‘ Efficient sampling and meta-modeling for computational economic models ,’ Computational Economics , 44 ( 4 ), 507 – 536 . Google Scholar CrossRef Search ADS Saltelli A. , Ratto M. , Andres T. , Campolongo F. , Cariboni J. , Gatelli D. , Saisana M. , Tarantola S. ( 2008 ), Global Sensitivity Analysis: The Primer . John Wiley & Sons : New York . Seppecher P. , Salle I. , Lavoie M. ( 2017 ), ‘What drives markups? Evolutionary pricing in an agent-based stock-flow consistent macroeconomic model,’ CEPN Working Papers 2017-03. Centre d’Economie de l’Université de Paris Nord: Paris (FR). Setterfield M. ( 2009 ), Path Dependency, Hysteresis and Macrodynamics . Springer , pp. 37 – 79 . Google Scholar CrossRef Search ADS Siemer M. ( 2014 ), ‘Firm entry and employment dynamics in the Great Recession,’ Technical report. Skott P. ( 2005 ), ‘ Fairness as a source of hysteresis in employment and relative wages ,’ Journal of Economic Behavior & Organization , 57 ( 3 ), 305 – 331 . ISSN 0167-2681. doi: https://doi.org/10.1016/ j.jebo.2003.11.011. URL http://www.sciencedirect.com/science/article/pii/S016726810400143X. Google Scholar CrossRef Search ADS Sobol I. M. ( 1993 ), ‘ Sensitivity analysis for non-linear mathematical models ,’ Mathematical Modelling and Computational Experiments , 1 ( 4 ), 407 – 414 . Stiglitz J. ( 1994 ), ‘Endogenous growth and cycles,’ in Shionoya Y. , Perlman M. (eds), Innovation in Technology, Industry and Institutions . The University of Michigan Press , pp. 121 – 156 . Stockhammer E. ( 2011 ), ‘ Wage norms, capital accumulation, and unemployment: a post-keynesian view ,’ Oxford Review of Economic Policy , 27 ( 2 ), 295 – 311 . Google Scholar CrossRef Search ADS Tesfatsion L. , Judd K. (eds) ( 2006 ), Handbook of Computational Economics II: Agent-Based Computational Economics . Elsevier : North Holland, Amsterdam . Wald A. , Wolfowitz J. ( 1940 ), ‘ On a test whether two samples are from the same population ,’ The Annals of Mathematical Statistics , 11 ( 2 ), 147 – 162 . Google Scholar CrossRef Search ADS Zhou H. , Dekker R. , Kleinknecht A. ( 2011 ), ‘ Flexible labor and innovation performance: evidence from longitudinal firm-level data ,’ Industrial and Corporate Change , 20 ( 3 ), 941 . Google Scholar CrossRef Search ADS Appendices Appendix A Capital- and consumption-good sectors and technical change The technology of capital-good firms is (Aiτ,Biτ). Aiτ is the labor productivity of the machine-tool manufactured by i for the consumption-good sector, while Biτ is the labor productivity to produce the machine. Superscript τ denotes the technology vintage being produced/used. Given the monetary average wage wi,t paid by firm i, the unit cost of production is: ci,t=wi,tBiτ. (6) With a fixed markup μ1∈ℝ+ pricing rule, prices pi,t are defined as: pi,t=(1+μ1)ci,t. (7) Firms in the capital-good industry “adaptively” strive to increase their market shares and their profits trying to improve technology via innovation and imitation. Firms invest in R&D a fraction ν∈(0,1] of their past sales Si,t−1: RDi,t=νSi,t−1. (8) R&D activity is performed by workers exclusively devoted to this activity, whose demand is: Li,tR&D=RDi,twi,t. (9) Firms split their R&D workers Li,tR&D between innovation ( INi,t) and imitation ( IMi,t) activities according to the parameter ξ∈[0,1]: INi,t=ξLi,tR&D (10) IMi,t=(1−ξ)Li,tR&D. (11) Innovation is a two-step process. The first one determines whether a firm obtains or not access to an innovation—irrespectively of whether it is ultimately a success or a failure—through a draw from a Bernoulli distribution with parameter: θi,tin=1−e−ζ1INi,t, (12) and parameter ζ1∈(0,1]. If a firm innovates, it may draw a new machine-embodying technology (Ai,tin,Bi,tin) according to: Ai,tin=Ai,t(1+xi,tA), (13) Bi,tin=Bi,t(1+xi,tB), (14) where xi,tA and xi,tB are two independent draws from a Beta (α1,β1) distribution, (α1,β1)∈ℝ2+ over the fixed support [x¯1,x¯1]⊂ℝ. Imitation also follows a two-step procedure. The access to imitation comes from sampling a Bernoulli with parameter: θi,tim=1−e−ζ2IMi,t (15) and ζ2∈(0,1]. Firms accessing the second stage are able to copy the technology (Aiim,Biim) of one of the competitors. Finally, they select the machine to produce according to the rule: min [pi,th+bc(Ai),th], h=τ,in,im, (16) where b∈ℝ+ is a payback parameter. Firms in consumption-good sector do not conduct R&D; instead they access new technologies incorporating new machines to their existing capital stock Ξj,t−1. Firms invest according to expected demand Dj,te, computed by an adaptive rule: Dj,te=g(Dj,t−1,…,Dj,t−h), 0<h<t, (17) where Dj,t−h is the actual demand faced by firms at time t–h ( h∈ℕ* is a parameter and g:ℝh→ℝ+ is the expectation function, here an unweighted moving average over four periods). The corresponding desired level of production Qj,td, considering the actual inventories from previous period Nj,t−1, is: Qj,td=(1+ι)Dj,te−Nj,t−1, (18) being Nj,td=ιDj,te the desired inventories and ι∈ℝ+, a parameter. If the desired capital stock Kjd —computed as a linear function of the desired level of production Qj,td —is higher than the current one, firms invest EIj,td to expand their production capacity: EIj,td=Kj,td−Kj,t−1. (19) Firms also invest SIj,td to replace machines by more productive vintages according to a fixed payback period (b > 0) rule, substituting machines Aiτ∈Ξj,t according to its obsolescence as well as the price of new machines: RSj,t={Aiτ∈Ξj,t:pi,t*cj,tAiτ−cj,t*≤b}, (20) where pi,t*∈ℝ+ and cj,t*∈ℝ+ are the price and unit cost of production upon the new machines. Given the stock of machines Ξj,t, firms compute average productivity πj,t and average unit cost of production cj,t, based on the average unit labor cost of production wj,t associated with each machine of vintage τ in its capital stock: cj,tAiτ=wj,tAiτ. (21) Consumption-good prices are set applying a markup μj,t on average unit costs: pj,t=(1+μj,t)cj,t. (22) Markup changes are regulated by the evolution of firm market shares (fj,t): μj,t=μj,t−1(1+υfj,t−1−fj,t−2fj,t−2), (23) with υ∈(0,1). Firm market shares evolve according to a replicator dynamics: fj,t=fj,t−1(1+χEj,t−E¯tE¯t), (24) where the firms competitivity Ej,t is defined based on the individual normalized prices p′j,t and unfilled demands l′j,t: Ej,t=−ω1p′j,t−1−ω1l′j,t−1, E¯t=1Ft2∑jEj,tfj,t−1, (25) being (ω1,ω1)∈ℝ2 parameters. Labor market and search and match process Labor demand in the consumption-good sector Lj,td is determined by desired production Qj,td and the average productivity of current capital stock Aj,t: Lj,td=Qj,tdAj,t. (26) In the capital-good sector, instead, Li,td considers orders Qi,t and labor productivity Bi,t. In what follows, only the behavior of the consumption-good sector (subscript j) is shown, as the capital-good firms operate under the same rules in the labor market, except they follow the wage offers from top-paying firms in the former sector. Firms decide whether to hire (or fire) workers according to the expected production Qj,td. If it is increasing, ΔLj,td new workers are (tentatively) hired in addition to the existing number Lj,t−1. Each firm (expectedly) gets a fraction of the number of applicant workers La,t in its candidates queue {ℓj,ts}, proportional to firm market share fj,t−1. In terms of statistical expectations: E(Lj,ts)=ωLa,tfj,t−1, (27) where ω∈ℝ+ is a parameter defining the number of job queues each seeker joins, in average. Considering the set of workers in {ℓj,ts}, each firm selects the subset of desired workers {ℓj,td} to make a job (wage) offer: {ℓj,td}={ℓj,t∈{ℓj,ts}:wℓ,tr<wj,to}, {ℓj,td}⊆{ℓj,ts}. (28) Firms target workers that would accept the wage offer wj,to, considering the wage wℓ,tr requested by workers, if any. Each firm hires workers up to its demand ΔLj,td, or to all workers in its queue, and the number of effectively hired workers (the set {ℓj,th}) is: #{ℓj,th}=ΔLj,t≤ΔLj,td≤Lj,ts=#{ℓj,ts}, ΔLj,t=Lj,t−Lj,t−1. (29) The search, wage determination, and firing processes differ according to the policy regime. In the Fordist regime, workers never quit jobs and firms fire employees only under losses ( Πj,t−1<0) and shrinking desired production ( Qj,td<Qj,t−1), except if exiting the market. Only unemployed workers search for jobs. Additionally, lowest skilled workers are fired first, while higher skilled workers are preferred when hiring, as in this regime wages are not bargained. Firms offer a wage: wj,to=wj,t−1o(1+WPj,t) bounded to wj,tmax=pj,t−1Aj,t−1, (30) that is accepted by the worker if she has no better offer. The positive wage premium is is defined as: WPj,t=ψ2ΔAtAt−1+ψ4ΔAj,tAj,t−1, ψ2+ψ4≤1, (31) being At the aggregate labor productivity and (ψ2,ψ4)∈ℝ2+ parameters. So, wages are linked to firm-specific performance and also to the aggregate productivity dynamics. wj,to is simultaneously applied to all firm’s workers. wj,to is bounded to a maximum break-even wage wj,tmax (the 0 unit profits myopic expectation). In the Competitive setting, firms freely fire workers and employees actively search for better jobs while employed, quitting when there is a better offer. When hiring or firing, firms give precedence to workers with a higher skills-to-wage ratio ( stℓ/wtℓ), contracting them first and dismissing last. The matching is done by a one-round bargaining process. Workers have a reservation wage equal to the unemployment benefit wtu they receive from the Government when unemployed, if any, and request a wage wℓ,tr during the job application: wℓ,tr={wℓ,t−1(1+ϵ) if employed in t-1 wℓ,ts if unemployed in t-1. (32) wℓ,t−1 is the current wage for the employed workers, and ϵ∈ℝ+ is a parameter. Unemployed workers have a gradually shrinking satisfying wage wℓ,ts, accounting for their recent wage history: wℓ,ts=max (wtu,1Ts∑h=1Tswℓ,t−h), (33) being Ts∈ℕ*, the time span parameter of the moving-average of the past income. An employed worker accepts the best offer wj,to she receives if it is higher than her current wage wℓ,t. An unemployed worker accepts the best offer she gets, if any, as all offers are at least equal to the unemployment benefit wtu. In all cases, Government establishes an institutional minimum wage wtmin, as the lower bound to the firm wage setting behavior: wtmin=wt−1min(1+ψ2ΔAtAt−1). (34) Model closure Government taxes firms profits at a fixed rate tr∈ℝ+ and provides a benefit wtu to unemployed workers which is a fraction of the current average wage: wtu=ψ1Lt−1D∑ℓ=1Lt−1Dwℓ,t−1, (35) where ψ∈[0,1] is a parameter, and LtD is the total labor demand. Therefore, the Government expenses are: Gt=wtu(LS−LtD). (36) Workers fully consume their income (if possible) and do not get credit. Accordingly, desired aggregate consumption Ctd depends on the income of both employed and unemployed workers plus the desired unsatisfied consumption from previous periods ( Ct−1d−Ct−1): Ctd=∑ℓwℓ,t+Gt+(Ct−1d−Ct−1). (37) The model applies the standard national account identities by the simple aggregation of agents’ stocks and flows. The aggregate value added by capital- and consumption-good firms Yt equals their aggregate production Qt1 and Qt2, respectively (there are no intermediate goods). In turn, it is equal to the sum of the effective consumption Ct, the total investment It, and the change in firm’s inventories ΔNt: Qt1+Qt2=Yt=Ct+It+ΔNt. (38) For further details, see Dosi et al. (2010) and Dosi et al. (2017c). Appendix B Table B1. Model parameters and initial conditions, calibration values, minimum–maximum range for sensitivity analisys, Elementary Effects μ* statistic, and Sobol decomposition direct and interaction effects indexes Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Note: Baseline policy-specific values (Fordist regime). SA statistics relative to GDP crises recovery losses indicator averages for both regimes. Appendix B Table B1. Model parameters and initial conditions, calibration values, minimum–maximum range for sensitivity analisys, Elementary Effects μ* statistic, and Sobol decomposition direct and interaction effects indexes Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Symbol Description Value Minimum Maximum μ* Direct Interaction Policy ϕ Unemployment subsidy rate on average wage 0.40 0.00 1.00 4.82 – – r Interest rate 0.01 0.00 0.10 8.27 0.006 0.001 tr Tax rate 0.10 0.00 0.30 4.24 0.001 0.001 (Λ,Λmin) Prudential limit on debt (sales multiple/fixed floor) (2, 20,000) (1, 0) (4, 100000) (6.89,2.07) (−,−) (−,−) Labor market ϵ Minimum desired wage increase rate 0.020 0.005 0.200 6.33 0.000 0.000 τ Skills accumulation rate 0.010 0.001 0.100 11.0 0.714 0.030 Tr Number of periods before retirement (work life) 120 60 240 3.96 0.032 0.012 Ts Number of wage memory periods 0 1 8 0.66 – – (ω,ωun) Number of firms to send applications (employed/unemployed) (0, 5) (1, 1) (20, 20) (2.87,8.92) (−,0.002) (−,0.001) (ψ2,ψ4) Aggregate/firm-level productivity pass-trough (0.50,0.50) (0.95,0.00) (1.05,1.00) (11.1,5.38) (−,−) (−,−) Technology η Maximum machine tools useful life 20 10 40 10.9 0.000 0.002 ν R&D investment propensity over sales 0.04 0.01 0.20 2.58 – – ξ Share of R&D expenditure in imitation 0.50 0.20 0.80 9.78 – – b Payback period for machine replacement 3 1 10 7.72 0.007 0.001 dimmach Machine-tool unit production capacity 40 10 100 7.88 0.014 0.002 (α1,β1) Beta distribution parameters (innovation process) (3, 3) (1, 1) (5, 5) (8.96,5.21) – – (α2,β2) Beta distribution parameters (entrant productivity) (2, 4) (1, 1) (5, 5) (5.89,10.3) (−,0.000) (−,0.001) (ζ1,ζ2) Search capabilities for innovation/imitation (0.30,0.30) (0.10,0.10) (0.60,0.60) (6.91,4.91) (−,−) (−,−) [x¯1,x¯1] Beta distribution support (innovation process) [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (4.16,4.74) (−,0.012) (−,0.001) Industrial dynamics γ Share of new customers for capital-good firm 0.50 0.20 0.80 8.45 – – ι Desired inventories share 0.10 0.00 0.30 5.98 0.000 0.001 μ1 Markup in capital-good sector 0.05 0.01 0.20 7.76 0.000 0.001 o Weight of market conditions for entry decision 0.50 0.00 1.00 3.80 – – χ Replicator dynamics coefficient (competition intensity) 1.0 0.2 5.0 9.13 0.056 0.001 υ Markup adjustment coefficient 0.04 0.01 0.10 5.05 – – u Planned utilization by consumption-good entrant 0.75 0.50 1.00 5.35 0.034 0.001 x5 Maximum technical advantage of capital-good entrant 0.30 0.00 1.00 8.97 0.030 0.001 exit1 Minimum orders to stay in capital-good sector 1 1 5 3.90 – – exit2 Minimum share to stay in consumption-good sector 10−5 10−6 10−3 3.38 – – [Φ1,Φ2] Minimum/maximum capital ratio for consumption-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (5.43,11.6) (0.031,−) (0.001,−) [Φ3,Φ4] Minimum/maximum net wealth ratio for capital-good entrant [0.10,0.90] [0.00,0.50] [0.50,1.00] (8.68,5.00) (0.001,0.003) (0.001,0.001) (ω1,ω2) Competitiveness weight for price/unfilled demand (1.0,1.0) (0.2,0.2) (5.0,5.0) (7.97,12.5) (−,0.004) (−,0.000) [x¯2,x¯2] Entry randomness distribution support and limit [−0.15,0.15] [−0.3,0.1] [−0.1,0.3] (8.91,10.9) (0.002,0.001) (0.001,0.002) [Fmin1,Fmax1] Minimum/maximum number of capital-good firms [10,100] [10,20] [20,100] (15.3,19.9) (−,0.001) (−,0.003) [Fmin2,Fmax2] Minimum/maximum number of consumption-good firms [50,500] [50,200] [200,500] (5.90,6.59) (−,0.014) (−,0.012) Initial conditions μ02 Initial markup in consumption-good sector 0.20 0.10 0.50 10.54 0.003 0.001 K0 Initial capital stock in consumption-good sector 800 200 3000 3.72 – – L0S Number of workers 250,000 50000 100,0000 8.17 0.012 0.001 (F01,F02) Initial number of capital/consumption-good firms (20, 200) (10, 50) (100, 500) (6.39,7.49) (−,−) (−,−) (NW01,NW02) Initial net wealth in capital/consumption-good sector (10,000, 5000) (2000, 2000) (50,000, 50,000) (5.62,5.59) (0.005,0.001) ( 0.008,0.005) Note: Baseline policy-specific values (Fordist regime). SA statistics relative to GDP crises recovery losses indicator averages for both regimes. Table B2. Stock-and-flow consistency: transaction flow matrix Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Note: (*) Government deficit/superavit is close to 0 in the long run. Table B2. Stock-and-flow consistency: transaction flow matrix Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Workers Capital-good firms Consumption-good firms Bank Government ∑ current capital current capital current capital Consumption –C +C 0 Investment +I – I 0 Government expenditures +G –G 0 Wages +W −W1 −W2 0 Profits, firms −Π1 +Π1 −Π2 +Π2 0 Profits, bank −Πb +Πb 0 Debt interests −rDebt−11 −rDebt−12 +rDebt−1 0 Deposits interests +rNWt−11 +rNWt−12 −rNWt−1 0 Taxes −Tax1 −Tax2 +Tax 0 Change in debt +ΔDeb1 +ΔDeb2 −ΔDeb 0 Change in deposits −ΔNW1 −ΔNW2 +ΔNW 0 ∑ 0 0 0 0 0 0 0 0* 0* Note: (*) Government deficit/superavit is close to 0 in the long run. Table B3. Regime-specific parameter values Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 Table B3. Regime-specific parameter values Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 Parameter Description Fordist Competitive ω Number of firms to send applications 0 5 ϕ Unemployment subsidy rate on average wage 0.4 0.2 Ts Number of wage memory periods 0 4 r Interest rate 0.010 0.005 tr Tax rate 0.015 0.010 © The Author(s) 2018. Published by Oxford University Press on behalf of Associazione ICC. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com

Journal

Industrial and Corporate ChangeOxford University Press

Published: Apr 7, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off