Financialization, Technological Change, and Trade Union Decline

Financialization, Technological Change, and Trade Union Decline Abstract Recent research finds that financialization and technological change have had a variety of negative effects on labor, including reducing low-skill workers’ wages and increasing income inequality. In this article, I examine the effect on trade unions of one type of financialization, equity market development and one type of technological change, routine-biased technological change. I argue that we should conceptualize trade union strength in two dimensions: (a) the strength of their institutional structures, such as the degree of wage bargaining coordination and the degree to which firms can deviate from collective agreements; (b) the strength of their membership. Using data for 21 OECD countries from 1970 to 2010, I find a negative effect of equity market development on unions’ institutional structures, but not on union membership. Contrarily, I find that routine-biased technological change has a negative effect on union density, but an inconsistent relationship with the strength of unions’ institutional structures. 1. Introduction Labor markets in advanced democracies have undergone dramatic changes in recent decades. There has been a shift from industrial to service employment, the Internet has revolutionized service delivery, and economic inequality has increased. One of the major contributions to the latter phenomenon has been the decline of trade unions. While trade unions remain quite strong in several advanced democracies, they have been weakened in most countries in several ways, with decreasing coverage of collective bargaining agreements, reduction in the scope of collective agreements and declining membership (Baccaro and Howell, 2011). Work on trade union decline has devoted great attention to how structural economic changes, such as technological change, offshoring and global capital flows have adversely affected trade unions. One of these structural changes, which has been the focus of much recent work in economic sociology, is ‘financialization’, an increased role for financial actors in firms’ management and an increased importance of revenue from financial activities to firms’ bottom lines (van der Zwan, 2014). While there is a substantial literature on how financialization adversely impacts a variety of labor outcomes, including raising the probability of firm downsizing (Jung, 2015) and increasing income inequality (Lin and Tomaskovic-Devey, 2013), there has been little work on how financialization affects unions. Another important structural change in Western democracies has been declining employment in ‘routine task’ jobs: jobs in which employees performed conceptually simple, repetitive tasks. As computing power has increased, these jobs, which can be programed as algorithms, can be performed by machines (Autor et al., 2003). Recent work in labor economics has found that this ‘routine-biased’ technological change has largely affected jobs that were in the middle of the wage distribution, such as assembly line and clerical work (Spitz-Oener 2006). As a result of this, there has been labor market ‘polarization,’ an increase in employment in low- and high-skill jobs, across advanced democracies (Goos et al., 2014). In this article, I examine how financialization and routine-biased technological change have affected various measures of trade union strength. While there are several different definitions of financialization in the economic sociology literature, I focus on the role of equities markets. I argue that stock market development, in particular, should have a negative effect on trade union strength because equity investors gain more from reduced labor costs than holders of other financial instruments, such as bonds. Investors will press managers to seek greater flexibility in their employment practices, including in collective agreements. Because much management compensation comes in the form of stock options and managers fear firm takeovers if their stock price remains low, they have personal incentives to pursue such policies. I also argue that routine task employment will be positively associated with trade union strength, i.e. when there is high employment in routine task jobs, trade unions will be stronger. Before increases in computing power, industrial production and many types of clerical work were very labor intensive, requiring large numbers of similarly skilled individuals. Because these individuals had similar interests and were concentrated within workplaces, they had a great deal of power to demand strong unions. When routine task employment declines, this source of collective power is lost. New jobs are less likely to be unionized as they are typically in smaller firms, which do not have a history or culture of unionization. In contrast to much work on union strength, which focuses on union density or wage bargaining coordination separately, I focus on multiple measures: union density, wage bargaining coordination, wage bargaining centralization and the presence of ‘opening clauses’ in collective agreements, which allow firms to opt out of collective agreement provisions when facing economic hardship. I conceptualize these variables into two categories: institutional structure variables and membership. Recent work on institutional change argues that there is typically wide latitude for flexibility within institutions and that there may be institutional change without change in the formal structure of an institution (Mahoney and Thelen, 2010). I argue that wage bargaining coordination, centralization and opening clauses capture different aspects of institutional structure while union density captures membership. Furthermore, stock market development and routine-biased technological change may differentially affect these variables. In order to assess these arguments empirically, I analyze data for 21 OECD countries from 1970 to 2010. Using error correction models (ECM), which include both changes and levels of the independent variables, I find that routine task employment and stock market development have differential effects on the institutional structure and membership variables. I find that changes in stock market development adversely affect the institutional structure variables but have little consistent effect on union density. I find that union density is more likely to grow when routine task employment is growing and that the rate of growth is highest at higher levels of routine task employment. I find, however, that the level of routine task employment is actually a negative predictor of both wage bargaining coordination and centralization, suggesting that while routine task employment has a positive effect on union membership, institutional structures do not become stronger when routine task employment is high. I run two types of robustness checks to help rule out reverse causality: (a) models where I lag changes in the independent variables and (b) models where I regress stock market development and routine task employment on the various measures of union strength. I find that the original results are largely robust to the first type of robustness check. Regressions of stock market development and routine task employment on the four measures of union strength show no consistent relationships, suggesting that reverse causality is not driving the results. I proceed as follows: in Section 2, I present my conceptualization of the trade union variables. In Section 3, I review the literature on trade union decline. In Section 4, I both review the literature on how financialization and routine-biased technological change have affected labor and develop my arguments for how these should affect trade unions. This is followed in Section 5 by a discussion of data, methods and predictions for each of the main independent variables. In Section 6, I present the main results and the two robustness checks. Section 7 follows with a discussion of the results in broader context and concludes. 2. Conceptualizing union strength: institutional structures and union membership There have been a variety of explanations for both union density decline and decentralization of collective bargaining, but there has been little work which analyzes these two phenomena together. It is important to do so, however, because their timing differs and the factors which affect them may differ as well. Union density decline has been gradual and consistent in most countries over time, while usage of opening clauses increased starting in the 1990s and decentralization of wage bargaining occurred largely in the 1980s and early 1990s. Structural changes, such as technological change and offshoring, may affect union membership, but national institutions, such as wage bargaining coordination, may experience little change due to path dependence or institutional complementarities. In addition to differential timing in how they have changed, I argue that these union variables can be conceptualized into two categories: institutional structure, including wage bargaining coordination, centralization, and opening clauses, and union membership, including union density. While institutional structures can certainly help explain union membership (and vice versa), institutional structures can change without necessarily affecting membership. If, for example, a high degree of cross-sector coordination in wage bargaining creates conflict between different skill groups of workers, individual unions may be able to better grow their membership if cross-sector coordination is reduced. Job losses due to the decline of heavily unionized occupations may cause a decline in union membership without causing a decline in the institutional structure. At the same time, within levels of wage bargaining coordination and centralization, the institutional structure can be more or less rigid. This is in part a function of the presence of opening clauses, which can increase the scope of bargaining at the local level and allow employers to deviate from the conditions of centralized collective agreements when facing economic hardship. Recent work has shown that the broadest structures of collective bargaining, wage bargaining coordination and centralization, have not greatly changed in recent years (Du Caju et al., 2008; Thelen, 2014). Collective bargaining has occurred primarily at the industry level in most European countries since the 1990s. But despite stability in the broadest level of institutional structure, there has been a substantial degree of within-structure weakening. In the Nordic countries, the content of industry-level collective agreements has changed, with collective agreements primarily specifying minimum wages, rather than full wage scales as in the post-war decades (Ahlberg and Bruun, 2005). In Germany, industry-level collective agreements have increasingly included ‘opening clauses’, which allow participating employers to deviate from the provisions in the agreement under certain economic conditions. These opening clauses represent both a weakening at the central level, in that national unions and employers’ associations allow individual employers to deviate from them, and at the local level, in that their usage means that the agreement’s conditions do not always govern employment. Union density has declined in almost all advanced democracies. Although this does not necessarily prevent unions from setting high wages through collective agreements, it may make it more difficult for unions to rally enough people to effectively strike.1 While much work of historical institutionalism has focused on abrupt change at ‘critical junctures’, unsettled times at which social norms have become weakened and there is a possibility to redesign institutions (Capoccia and Kelemen, 2007), recent work has argued that institutional change comes largely through gradual but continual change. Mahoney and Thelen (2010) argue that rule interpretation and enforcement autonomy can fundamentally change the functioning of an institution without changing its structure. Regarding industrial relations institutions, Thelen (2014) argues that the gradual hollowing out of centralized collective agreements and their replacement by less redistributive local bargaining, a process which she calls ‘embedded flexibilization,’ has allowed unions in these countries to maintain a high degree of solidarity. Centralized collective agreements no longer strictly link wage increases in different occupations and regions to each other, which reduces the between-skill group conflict that existed in these countries in the 1980s. Yet they are still responsible for setting high minimum wages, which has allowed Scandinavian countries to largely avoid working poverty. So it is still important to study institutional structures, both because they continue to matter for important outcomes, and because it is an open question whether they are amenable to the same types of explanations as within-institution strength and union membership. 3. Previous explanations for trade union decline While existing scholarship tends to separately address institutional structure and within-institutional structure change in wage bargaining institutions separately, scholars have given similar types of explanations for both. Explanations for trade union decline can be grouped into four broad categories: (a) political (b) national institutions (c) globalization (d) deindustrialization. One of the foremost explanations for union decline generally has been that right-wing politicians have become more anti-union and have actively worked to weaken unions (Brady, 2007). The most famous examples are the USA, where Ronald Reagan fired striking air-traffic controllers in 1981, which began an anti-union turn in American politics and the UK, where Margaret Thatcher’s Conservative Party passed far-reaching union reforms in the 1980s, removing much of unions’ strike immunity and implementing more stringent conditions on union votes. Another type of explanation emphasizes the role of national institutions in strengthening or weakening collective bargaining. This approach is associated with Varieties of Capitalism, according to which employers assent to strong unions in coordinated market economies (CME), such as Germany and Sweden, because they produce products that require specific skills and more cooperative labor relations (Hall and Soskice, 2001). Other scholars have found that organizations like coordinated wage bargaining and works councils help strengthen union membership (Scruggs and Lange, 2002). Western (1997) found that union membership is higher in countries with a Ghent unemployment system, where union membership is required to participate in the unemployment insurance system. Ahlquist (2010) found that coordinated wage bargaining is more likely to occur in countries in which unions have established a centralized strike fund. Perhaps the foremost types of explanations have been based on various aspects of globalization, such as increased trade, capital mobility and immigration, and on deindustrialization. As countries reduce barriers to trade, manufacturers in highly developed countries can begin to take advantage of lower labor costs in developing countries. Offshoring and trade impact largely lower skill workers in manufacturing and industry, who were likely to be union members.2 Increased capital mobility allows firms to locate production in countries with lower labor costs and repatriate profits to their home countries. Along similar lines, scholars have found that increased foreign direct investment decreases unions’ wage premia and union density (Choi, 2001; Slaughter, 2007). Finally, Lee (2005) found that inward immigration reduced union density. Deindustrialization is another prominent recent explanation for trade union decline, as industrial jobs were among those most likely to be unionized (Hirsch, 2008). There is a simple explanation based on deindustrialization, in which unions decline due to attrition when factories are offshored or jobs replaced by new technology. But there is also a more sophisticated version, in which technological change differentially shapes the bargaining power of different skill groups of workers, who then have differing preferences over unionization. With such skill-biased technological change (SBTC), both employment and wages in high-skill occupations increase due to their complementarity with new technology. These decrease, however, in lower-skill occupations. Because of this increased differential in individual bargaining leverage, high-skill workers, who have more bargaining leverage, are less willing to join low-skill workers to support unions (Acemoglu et al., 2001). A similar explanation has been given for the decline of coordinated wage bargaining in Europe in the 1980s and 1990s. New technology gave rise to ‘diversified quality production’, which increased the global competitiveness of high-skill manufacturing workers (Iversen, 1996; Pontusson and Swenson, 1996). These workers demanded higher wage differentials and unions representing them withdrew from multi-sector wage bargaining institutions. 4. Financialization, technological change, and labor market outcomes 4.1 Finance and labor While there has been a substantial amount of work on globalization, deindustrialization and union decline, recent developments in economic sociology on financialization and in labor economics on routine-biased technological change suggest new mechanisms through which to advance the study of trade union decline. In foundational work on the relationship between finance and labor, Hall and Soskice (2001) argued that ‘patient capital’ provided by banks in long-term relationships with firms is a central institution of CMEs. It enabled long-term, stable relationships with unions, in which banks promised to finance skill investments and protect workers’ specific skills during market downturns in return for success in niche, specific skill-intensive markets. But Hall and Soskice also recognized that financial markets had become more global, that financial actors had become more heterogeneous, and that finance providers might become less willing to underwrite such relationships. Indeed Hardie et al. (2013) find that even large banks in CMEs have become less willing to provide patient capital because they are dependent on international markets for funding, which makes them less able to have such long-term commitments. Empirical work on finance and labor has found that liberalized finance results in lower shares of firm revenue going to labor over a variety of outcome variables. Bertrand et al. (2007) found that after reforms reducing government intervention into bank lending in France in the mid-1980s, average wage increases were substantially lower in more bank-dependent sectors. They also found that worse-performing firms were more likely to outsource. Scholars have found that adoption of ‘shareholder value’ practices increased dividend payouts to shareholders and decreased the share of revenue going to workers (Lazonick and O’Sullivan, 2000; Beyer and Hassel, 2002). Lin and Tomaskovic-Devey (2013) found that American firms’ increasing dependence on financial income in recent decades was associated with a lower labor share of income, increased top executive share of compensation, and increasing earnings dispersion among workers. Jung (2015) found that increased reliance on institutional investors was associated with a greater probability of workforce downsizing for a sample of American firms from 1981 to 2006. Similar to my own analysis, Black et al. (2007), using cross-sectional data from the 1990s, found that wage bargaining centralization is lower in countries with higher equity market development. Also closely related, Darcillon (2015) found that higher shares of employment and total value added in finance are associated with lower levels of a composite measure of workers’ bargaining power. In this article, I focus specifically on how equity market development should affect trade unions. While previous work has focused on broader conceptualizations of financialization, such as employment in finance and share of value added in financial activities (Darcillon, 2015), I focus on stock market development. This is both because equity investors, despite their heterogeneity, should have certain common objectives with respect to firms and because firm managers will be influenced by investor preferences and the price of shares, both as components of their compensation and as performance metrics.3 One shortcoming of previous work on financialization and labor outcomes is that different types of investors may have very different preferences regarding firm performance and that measures combining different types of financial investments may obscure more than they reveal. Equity investors, for example, have different incentives than bond investors. The former make money when share prices rise while the latter earn a set rate of return that can only be wiped out through bankruptcy or firm restructuring. Because of this, bondholders should not be as concerned with firms’ profitability and should be more patient investors than equity holders (Deeg and Hardie, 2016). How should we expect equities and equity prices to matter for management of the firm? First, it is important to distinguish between how they will matter for financial actors and how they will matter for management. Equities entitle holders to a share of the firms’ earnings and they make money on their shares when the share prices increases. Their primary motive should be to achieve higher share prices, which comes through increased profitability. A firm may increase its profitability through firm growth, but it may also achieve this by limiting itself to a set of core, profitable activities and by reducing the share of revenue going to other stakeholders. In recent decades, the focus has shifted from growing the firm to focusing on a narrower set of competencies and aiming for high profitability (Dallery, 2009). Dallery (2009) argues the preference for high profitability has resulted in lower investment and growth rates than would be preferred by both management and labor. Shareholder value, the idea that the primary purpose of corporations is to make money for their shareholders (owners), has become a powerful rallying cry for investors (Lazonick and O’Sullivan, 2000; van der Zwan, 2014). Two common demands have been for managers to increase the size of dividends paid on shares and to buy back the firm’s stock, which reduces the number of outstanding shares, thus raising share prices. Money spent on share buybacks could have otherwise been spent on capital upgrades or increased employment/wages (Gospel and Pendleton, 2003). Because institutions are under constant pressure to deliver returns, fund managers are under high pressure to deliver constant performance. They face a great deal of competition from other fund managers and evaluate their analysts’ performance frequently. There are also longstanding concerns that equities investors, while exhibiting heterogeneity, are often biased toward short-term performance, preferring measures that enhance stock performance in the short term at the expense of longer-term commitments (Jackson and Petraki, 2011). Dallery (2009) argues that investors’ preference for increasing profitability by lowering investment and firm growth will be especially strong if these investors have short-term time horizons. Amable et al. (2005) argue that time-horizons are important for behavior toward labor because when investors have long-time horizons, a cooperative relationship with labor will better maximize firm performance. Unless stock-based compensation requires an extended holding period, this can encourage management to take short-term measures, such as cutting labor costs or under-investment in capital to boost the stock price and their compensation. There are several ways in which these investor preferences can directly influence management. One, due to past efforts of equity investors, a substantial percentage of manager pay now comes from stock options, which incentivize managers to take actions that will ensure high stock prices (Bond et al., 2012). As mentioned above, there is a great deal of pressure from investors to buy back the firm’s shares in order to boost its stock price and to increase the dividends paid per share. Additionally, even though equities markets are secondary markets and do not directly fund the firm’s daily activities, lenders make decisions upon these prices and if the stock price underperforms, the firm may become a target for takeovers by activist investors (Höpner, 2005). These investors would likely replace existing management to realize latent profitability potential. From a behavioral finance perspective, managers derive a great deal of personal prestige from stock prices and may be willing to forego investment projects in order to make expected quarterly numbers (Graham et al., 2005). Even if they are not directly pressured by investors to do so, stock prices are commonly seen as important metrics of performance and managers will try to boost them to signal both to markets and the public that they are competently managing the firm (Aspara et al., 2014).4 How do these preferences matter specifically for trade unions? One way to increase profitability is to ensure a high degree of flexibility in labor contracts. Generally, investors will want greater management flexibility with respect to setting labor policy.5 They should be more opposed to institutions that impose rigidity on management’s policy setting. Examples of institutions which investors should oppose for this reason are centralized and coordinated wage bargaining, where decisions for all firms within an industry or region are made at a central level. This is problematic for the creation of shareholder value because these wages may be too high for certain firms, given their economic performance. With respect to unions, this means that the growth of the shareholder value paradigm should weaken institutions like centralized and coordinated wage bargaining, which will allow more localized union/management discretion and greater response to local conditions. It should also increase investors’ and managers’ interest in including opening clauses in collective agreements. These would allow managers greater autonomy to reduce wages and labor costs to limit losses during economic downturns. Investors’ preferences may not, however, equally affect all union variables equally. While a preference for firm-level policy autonomy suggests an aversion to centralized wage bargaining, the implications for union density, which measures union membership as a percentage of the workforce, are less clear. As long as union membership does not translate into greater policy rigidity, investors are less likely to be concerned with this. On the other hand, specific tactics that management may use to weaken unions and reduce the share of firm revenue going to labor, such as outsourcing non-core tasks to work agencies, offshoring aspects of production to lower labor cost countries, or relying more heavily on part-time and temporary workers, could either prevent workers from unionizing or weaken their incentives to do so.6 Then stock market development could have an indirect negative effect on union density. And generally if investors and management push for more decentralized wage setting and this in turn lead to lower wage increases for union members, they may be less likely to join unions and pay union dues because they do not see the costs of union membership as outweighing the benefits.7 4.2 Technological change and labor One of the predominant explanations for trade union decline has been deindustrialization, a decline in employment in industry and manufacturing (Lee, 2005; Hirsch, 2008). Economists have incorporated the heterogeneous effect of technological change across skill groups into models of workers’ preferences for trade union representation and trade union decline (Acemoglu et al., 2001; Dinlersoz and Greenwood, 2016). These theories have been based on SBTC, which holds that under recent technological change, there is a positive correlation between a worker’s skill level and their wages/employment opportunities (Goldin and Katz, 2008). In other words, the highest skill workers have the best employment opportunities and make the highest wages, whereas the lowest skilled workers have the poorest employment opportunities and have seen their wages either stagnate or decline. The models of SBTC and union decline posit that because technological change has increased the individual bargaining power of high-skill workers, they will be less likely to join lower-skill workers to support unions than they were in the past. Recent work in labor economics has, however, called the SBTC explanation for changes in occupational employment into question. Autor et al. (2003) argued that technological change, specifically improvements in computing power over the past few decades, has allowed for the replacement of routine task jobs.8 They found the greatest employment decline for US occupations from 1960 to 1998 in occupations that were routine task intensive. Further research has found decline in routine task employment to be prevalent across advanced democracies and that these jobs were most heavily concentrated in the middle of the wage distribution, leading to labor market ‘polarization’ (Spitz-Oener, 2006; Goos et al., 2014). How would we expect the decline of routine task employment and labor market polarization to affect trade unions? The occupations with the highest routine task content were those in industry/manufacturing and clerical work. These occupations were likely to be highly unionized because they required many workers in large workplaces performing repetitive tasks in order to produce finished goods or clerical tasks. Workers’ similar levels of skills and individual bargaining ability gave them a great deal of common interest in supporting unions and the collective power to do so. With the decline in routine task employment however, employment has sorted into high and low-skill occupations, increasing between-skill group heterogeneity in individual bargaining power. At the same time, the competitive pool for low-skill jobs has increased due to loss of employment in routine task occupations and these workers’ lower suitability for high-skill occupations (Jaimovich and Siu, 2014). This gives low-skilled workers less leverage to pressure employers to allow them to unionize. Therefore, we would expect routine task-biased technological change to cause union decline through three mechanisms: (a) that routine task occupations were among the most likely to be unionized and job loss in these occupations would cause a corresponding decline in union density; (b) that routine-biased technological change increases between-worker skill heterogeneity, which reduces workers demand to join together to support unions; (c) that formerly routine task workers can only substitute for workers in lower-skill jobs, which creates more competition for these jobs and gives employers greater leverage to resist workers’ attempts at unionization. In the previous section, I argued that we would expect investors to want an increase in firm-level flexibility over wage-setting policy and that higher stock market development should be associated with reductions in coordinated and centralized wage bargaining as well as increased usage of opening clauses, but not necessarily a decrease in union density. In contrast to stock market development, the clearest prediction for routine-biased technological change is for union density. Workers were particularly likely to be unionized in factories during the Fordist era because production required large numbers of similarly skilled workers, giving them a common interest in union representation. Research on plant-level unionization finds that newer and smaller workplaces are less likely to be unionized (Schnabel, 2013). If union membership is voluntary, workers in these workplaces would be less likely to want to become union members. There is no legacy of union relationships as in older firms and no norm among the workers of union membership at that firm. The relationship between routine task employment and unions’ institutional structures is less clear. As routine task employment declines, it may be more difficult to maintain coordinated wage bargaining because the workforce becomes more heterogeneous. But if unions and employers allow increased local and firm-level bargaining and greater possibility of deviation through opening clauses, formal coordination may continue to exist at the sector level. It might also be the case that through high minimum wages, high levels of wage bargaining coordination price out lower-skill jobs, which would mean that the occupational structure becomes less polarized when coordination and centralization are high. In any case, the prediction for routine-biased technological change is clearer for union membership than for unions’ institutional structures. 5. Data and methods In order to empirically test these theories, I have assembled a dataset for 21 OECD countries from 1970 to 2010. My main variable StMkt is an average of stock market capitalization/GDP and stock market value traded/GDP compiled largely from data for 1975–2004 from Claessens et al. (2006) complimented by data from the Financial Development and Structure Dataset (Beck et al., 2000). These two elements capture the amount and flow of equity investment, both of which should affect firms’ decision-making. I take this measure of stock market development as a proxy for the pressure that equity prices place on investors’ preferences and managements’ decision-making. When a country’s level of stock market development is higher, the stock market plays a greater role in the economy and the importance of equity investors’ preferences for delivering shareholder value will be higher, as will managements’ responsiveness to equity prices. Data on occupational employment come from LABORSTA, which contains ISCO one-digit occupational employment for OECD countries from 1970 to 2010.9 Codings of occupational routine task intensity come from Autor et al. (2003). I generate a variable RTI, a country-year measure of the routine task intensity of overall employment, by weighting employment in each occupational category as a percentage of total employment by its routine task intensity score from Autor et al. Data for union variables, including union density, coordination, centralization, and presence of opening clauses, works councils and a strike fund come from Visser (2013). Data on economic variables come from the Comparative Welfare States Dataset and various sources within (Brady et al., 2014). Data on migration come from OECD Stats Extracts and the 1977 and 1985 UN Demographic Yearbooks (United Nations, 1978, 1987). I linearly interpolated missing observations within an otherwise complete block of observations within country panels.10 Due to the presence of both autocorrelation and panel non-stationarity, I use ECM, as recommended by DeBoef and Keele (2008).11 The basic ECM takes the following form:   ΔYt= α0+α1*Yt−1+ β0*ΔXt+ β1*Xt−1+ εt where:   α1*= (α1– 1)β0*= β0β1*= β0+β112 I include four types of models for all dependent variables: models with random effects, with country fixed effects, and with country and either 5-year or year fixed effects, all accounting for AR(1) autocorrelation and with panel corrected standard errors (Beck and Katz, 1995).13 While I present models both without and with country fixed effects, I prefer the country fixed effects models because in addition to the possibility of unobserved, relatively fixed differences between countries, equities ownership differs in ways which may have implication for labor relations. In the Netherlands and Denmark, for example, unions play an important role in the governance of pension funds, which have large stakes in equities (Jackson and Thelen, 2015; McCarthy et al., 2016). As a result, we would expect stock market development to have a less adverse effect on labor relations in these countries. In absence of a measure of the degree of union involvement in pension funds, country fixed effects help address these differences between countries and give us the effect of within-country changes in stock market development on the various trade union outcomes. In addition to helping address the issue of panel non-stationarity, the ECM is a dynamic model which gives estimates for the effects of changes versus lagged levels in the independent variables (β0*ΔXt and β1*Xt−1, respectively) on changes in the dependent variable (ΔYt). This raises further interesting and substantively important questions about what we would expect the effect of changes versus levels of stock market development and routine task employment on the various dependent variables to be. We should expect to see a negative relationship between changes in stock market development and changes in the institutional structure variables because rigid institutional structures entail reduced flexibility in wages and working conditions. As stock market development increases, the presence of investors increases and the pressure of the demand to reduce labor costs increases. This should result in the decline of centralized wage bargaining institutions, which remove decision-making power over labor from the hands of management, and increased usage of opening clauses, which allow greater scope for local bargaining and flexibility in response to hardship. The expected relationship between the lagged level of stock market development Xt−1 and change in wage bargaining coordination and centralization is not, however, so clear. A negative relationship between high levels of stock market development and bargaining coordination would mean that at high levels of stock market development, there is greater decline in bargaining coordination than at low levels of stock market development. It is unclear that this would be the case. We would expect institutional structures to settle after undergoing change in response to initial changes in stock market development, which would mean that reduction in coordination would be achieved before stock markets reached high levels of development. There will also likely be a limit to how much collective bargaining institutional structures can change. We would not expect a country like Sweden, which had highly centralized bargaining, to converge to a US-style industrial relations system. Furthermore, investors may be willing to accept formally centralized wage setting if it moderates wage outcomes (Wallerstein, 1990). Because of this, we might expect to see a negative relationship between stock market development levels and change in institutional structures—in other words, that institutional structures decline when stock market development is still relatively low, but then stabilize. The relationship between stock market development and union membership should generally be negative. On one hand, if investors are primarily concerned with flexibility, attacks on union membership may not be necessary, as long as collective agreements allow flexibility. But we would expect investors to support many types of measures that would make union membership less attractive to workers. For example, we would expect investors to approve of firms outsourcing non-core jobs when it is possible to do so. If these outsourcing firms are not unionized, this would reduce union density. It would also put downward pressure on wages in unionized firms, as employers would be able to threaten unions with outsourcing. And in countries in which workers vote on union certification for their establishment, we would expect investors to support management efforts to fight this. As with unions’ institutional structures, however, it is not clear that union density decline should be greatest when stock market development is high. We might expect that union density declines are highest when stock market development begins to increase but levels are still relatively low. Unlike with institutional structures, however, the lower bound for union density is much lower and decline could continue at high levels of stock market development. Therefore, the predicted relationship between stock market levels and union density is unclear. How should levels versus changes in routine task employment affect unions’ institutional structures and membership? The effect on membership is relatively clear; as routine task employment was heavily concentrated in industry and workers in industry were heavily unionized, we should see a decline in union density with a decline in routine task employment. Union density decline may also occur through a second mechanism: when routine task employment declines, there is greater competition for lower-wage jobs in the service sector, which allows management to place greater pressure on unions not to unionize. Moreover, we should see variation in the degree of union density decline with levels of routine task employment. When routine task employment is high, we should expect union density to grow. But when it is low, union density should decline. As explained in the previous section, the relationship between routine task employment and unions’ institutional structures is somewhat ambiguous. Unlike union density, which can undergo continuous change, institutional structures like wage bargaining coordination are less likely to undergo gradual change and more likely to rapidly change at critical junctures. When routine task employment is high and union membership is strong, unions should preserve their institutional structures. They should also be able to retain this institutional strength for some time after routine task employment begins to decline. But eventually, unions will come under pressure to weaken collective agreements in order to preserve jobs. As a result, we should expect a decline in routine task employment to be associated with decline in wage bargaining coordination/centralization and with increased usage of opening clauses. It is somewhat less clear that there is a level of routine task employment at which would which expect coordination to decrease and opening clause usage to increase. Unions should be able to avoid institutional structure weakening for a while after routine task employment begins to decline, but will eventually face a trade-off between preserving centralized collective agreements without opening clauses and job security. Nevertheless, this should occur well before routine task employment has declined to its lowest levels and therefore we should expect a decline in coordination/centralization and an increase in opening clause usage at relatively high levels of routine task employment. Table 1 summarizes the predictions for the effect of increases in and levels of stock market development and routine task employment on changes in institutional structure (wage bargaining coordination/centralization), within-institutional-structure strength (opening clauses), and union membership (union density): Table 1. Predicted Effects of Increases and Levels of Stock Market Development and Routine Task Employment   Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable    Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable  Table 1. Predicted Effects of Increases and Levels of Stock Market Development and Routine Task Employment   Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable    Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable  6. Results and interpretation 6.1 Main results Due to the number of parameters in the models, I present simplified regression tables including only the main variables of interest in Tables 2–5. The dependent variables in these tables are union density, opening clauses, wage bargaining coordination and wage bargaining centralization, respectively. Full regression results with discussion are in the Supplementary Appendix. Columns 1 and 2 of each table are simple models, with the four main variables without and with country fixed effects. Columns 3–6 add controls, then two types of time fixed effects: for 5-year blocks in column 5 and for year in column 6. The 5-year fixed effects help address common period shocks while year fixed effects help address common year shocks. Table 2. Change in union density Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 2. Change in union density Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 3. Change in opening clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 3. Change in opening clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 4. Change in wage bargaining coordination Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 4. Change in wage bargaining coordination Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 5. Change in wage bargaining centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 5. Change in wage bargaining centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Contrary to the predictions, the stock market development variables have a fairly inconsistent relationship with union density. While the signs on both ΔStmkt and Stmktt−1 are mostly negative they are rarely significant. The results for both changes and levels of RTI are consistent with the predictions. They are positive in all of the regressions, meaning that a decline in routine task employment is associated with a decline in union density and that union density growth is higher at high levels of routine task employment. The coefficients are significant across all models for ΔRTI and are significant across all fixed effects models for RTIt−1. The dependent variable is measured as the change in the percentage of the workforce that is unionized and the independent variable is a standardized measure of the routine task intensity of employment, with mean of zero and SD ∼0.14.14 To give some sense of the substantive meaning, Model 4 would predict that for 1 SD increase in routine task employment, roughly the difference between the USA in 2007 and 1986, union density would be 5.92 percentage points higher. Table 3 presents the results for opening clauses. The opening clauses variable is coded 1–6, where 1 represents a complete lack of room for deviation from the conditions of sector-level collective agreements, while 5 indicates extensive presence and use of exceptions and 6 indicates lack of presence of sector-level collective agreements. Changes in stock market development are associated with increased usage of opening clauses across all models. Holding everything else constant, a 50 percentage point increase in stock market development, which is about the difference in Germany from 1980 to 2000, would be associated with a 0.1 unit increase (6-point scale) in opening clauses. This is a small effect, but it is important to remember that several countries, like Canada and the USA, have no variation in the dependent variable because they are always at the maximum value of the variable. The coefficients are significant in all models, except for the model with both country and year fixed effects. This is consistent with the argument that investors and managers under the influence of equity prices will want to increase flexibility in collective agreements. As with union density, there is no consistent relationship between Stmktt−1 and use of opening clauses. The signs on both RTI variables are almost always negative, consistent with the predictions. This suggests that as routine task employment decreases, employers are more able to use opening clauses, perhaps because workers are less able to prevent the internal weakening of institutional structures. The coefficients for ΔRTI are not, however, significant. Unlike routine task employment changes, it was more difficult to predict at which levels of routine task employment we would expect opening clause usage to increase. The negative coefficients suggest that opening clause usage is more likely to increase at lower levels of routine task employment, consistent with the idea that unions are able to avoid the trade-off between collective agreement strength and job protection until routine task employment has already substantially declined. Table 4 presents the results for wage bargaining coordination. The results for stock market development are consistent with those from the opening clauses regressions. The signs on ΔStmkt are all negative and significant while the signs on Stmktt−1 change and are almost entirely insignificant. This suggests that wage bargaining coordination declines quickly in response to increases in stock market development, but that it is not as sensitive to the level of stock market development. To give a sense of the magnitude, a change in German stock market development from the level in 1980 to the level in 2000 would be associated with about a 0.2 unit decrease (5-point scale) in wage bargaining coordination. ΔRTI has little consistent relationship with wage bargaining coordination and the sign on RTIt−1 depends on whether the regressions include fixed effects. Without fixed effects, RTIt−1 has a positive and significant relationship with wage bargaining coordination. But when we include fixed effects, this relationship is always negative. This is consistent with the idea that wage bargaining coordination will begin to decline when routine task employment is still relatively high. Note, however, the difference between these results and the opening clauses results; opening clauses were more likely to increase at lower levels of routine task employment. It is not surprising that there is a difference between the random effects and fixed effects results for RTI because we would expect that relatively constant country-level differences, such as culture, would affect the degree of change in institutional structures. In addition to reasons mentioned above, the negative correlation between RTI levels and change in wage bargaining coordination could be due to reverse causality. Under highly coordinated wage bargaining, it could be more difficult for employers to lay off workers. Oesch and Menes (2011) find that employment growth in low-wage sectors was higher in Spain and the UK, countries with lower wage floors, than in Germany and Switzerland. I further examine the possibility of reverse causality below in Section 6.2. As we can see in Table 5, the results for wage bargaining centralization are similar to those for wage bargaining coordination. Changes in stock market development again have a negative effect, while neither changes in RTI nor levels of stock market development show a consistent relationship. Again, RTIt−1 switches signs from positive in the random effects models to negative in the fixed effects models with full controls, although these coefficients are not significant as they were for wage bargaining coordination. 6.2 Robustness checks As with all observational data work not based on a strong natural experiment, it is not clear that we can reasonably interpret these estimates as causal. A causal interpretation would require at least two things: (a) that unobserved variables, such as countries’ legal institutions, do not explain stock market development or employment composition; (b) that there is not reverse causality, where wage bargaining institutions cause changes in stock market development or the composition of employment. That unobserved variables could explain stock market development or employment composition is a cause for concern. It could be, for example, that a country’s legal institutions or its culture help explain why some countries have higher stock market development or routine task employment than others. While not exhaustive, the inclusion of various types of fixed effects should help address the issue of selection effects as there are stark differences between countries in stock market development, which are likely based on unobserved country-level factors. Country fixed effects should, for example, help address concerns about labor and corporate governance legal institutions, which might explain stock market development, but are relatively constant within countries over time. As we saw in the preceding regressions, the inclusion of country fixed effects matters for the results. This is likely due in large part to the fact that countries have differences dating back to before the period of study. The 5-year and year fixed effects can help deal with common underlying factors that can explain stock market development and decline in routine task employment across these countries. Of these concerns, reverse causality is likely the largest problem for interpreting the coefficients. This is especially the case for the ΔXs, where simultaneity bias is a strong possibility. We would expect investors to shy away from countries with strong, and especially strong centralized union movements, as these reduce management autonomy in firm-level decision-making and would want to keep the ratio of firm revenues divided between workers and investors high. It could also be that decreases in wage bargaining coordination/centralization or introduction of opening clauses would cause stock markets to rise. Nevertheless, countries with very strong unions, such as Sweden, have had these institutions for the entire postwar period and still have many strong, internationally active firms. Investors may be willing to invest in these firms, despite union strength. As long as union actions are predictable, they can be factored in as one among many costs of doing business. Although there is no method to perfectly address this concern, one way which may help address it is to use a different lag structure in the ΔXs. In order to address this, I lagged the ΔXs by 1 year, resulting in the following model:   ΔYt= α0+α1*Yt−1+ β0*ΔXt−1+ β1*ΔXt−1+ εt If the contemporaneous correlations in the ΔXs are due to largely to reverse causality, we should find that the relationship which existed between ΔY and ΔX is either weaker or entirely disappears between ΔY and ΔXt−1. I present four regressions with full controls and the different types of fixed effects for each of the dependent variables in Tables 6,7. The results are similar to the original regressions, although there are a few exceptions. The coefficient for RTI in the union density regressions is now negative, which weakens the support for the argument although the coefficient on RTIt−1 remains positive and consistently significant. The relationship between ΔStmkt and opening clauses disappears, but the negative relationship between RTIt−1 and opening clauses remains. Generally, these results support those in the original regressions that decline in routine task employment reduces union density and that stock market development often has a negative effect, but that it is not highly robust. Table 6. Robustness Check: Change in Union Density, Opening Clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 6. Robustness Check: Change in Union Density, Opening Clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 7. Robustness Check: Change in Wage Bargaining Coordination, Centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 7. Robustness Check: Change in Wage Bargaining Coordination, Centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 8. Robustness Check: Stock Market Development, Routine Task Employment as the Dependent Variable Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 8. Robustness Check: Stock Market Development, Routine Task Employment as the Dependent Variable Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 The results for wage bargaining coordination and centralization in Table 7 are very similar to the original results. The coefficient on stock market development changes remains negative and significant across all specifications. In the previous regressions, RTIt−1 had a negative and significant relationship in many of the fixed effects models. This relationship disappears, but now ΔRTI has a consistently negative coefficient across models, significant in all but one of six models. The substantive conclusion is again fairly similar: the effect of stock market development on union institutions occurs is found in short-term changes, not in levels and the negative relationship is more robust for coordination/centralization than it is for either union density or opening clauses. Contrarily, higher routine task employment has a positive effect on union density and decreases usage of opening clauses, but it also appears to be consistent with lower levels of wage bargaining coordination and centralization. As an additional robustness check for reverse causality I run these same regressions, but with stock market development and routine task employment as dependent variables and the four union variables as predictors. In columns 1–4 and 6–9 of Table 8, I include the change and level parameters for each of the union variables separately without controls but with country fixed effects. Columns 5 and 10 add the control variables. If reverse causality were a concern and strong unions adversely affected stock market development, we would expect to see negative and significant coefficients on union density, bargaining coordination and bargaining centralization and a positive and significant coefficient on opening clauses. As we can see, the various union institutional variables all display inconsistent signs across the models and are seldom significant predictors of either stock market development or routine task employment. This can increase our confidence that the results are not due to reverse causality.15 7. Discussion and conclusion What conclusions can we draw from this analysis about the effect of stock market development and the decline of routine task employment on union strength? One conclusion is that these two variables affect unions in different ways. Perhaps the most surprising result is that decline in routine task employment does not have a negative effect on wage bargaining coordination and centralization. Many of the coefficients on RTI levels are negative and significant, showing that decline of wage bargaining coordination/centralization is stronger at low levels of routine task employment. One interesting possibility is that coordination and centralization are affected by between-firm differences within sectors. If many firms in a sector are similarly affected by technological change, there may not be any pressure to rethink multi-firm bargaining institutions. Stock market development, which has a consistent, negative relationship with wage bargaining coordination and centralization, may differentially affect firms within a sector, increasing differences between firms in whether management is willing to participate in such agreements and causing institutional decline. Future research should investigate the possibility that technological change, stock market development, or other conceptualizations of financialization have differential effects on within-institutional structure and institutional structure change because they have differential effects across firms within a sector. These findings also have important implications for debates on how different dimensions of unions’ strength mediate economic outcomes in Western democracies. While unions have been internally weakened, with both declining union density and collective agreements becoming both less comprehensive in their conditions and more flexible in their application, centralized institutional structures can still matter a great deal for substantive outcomes. While lower-skill workers in manufacturing and service sectors in Germany have borne the brunt of liberalization, higher-skill workers in the manufacturing sector continue to benefit from industry-level collective agreements and strong plant-level representation through works councils (Jackson and Thelen, 2015). In the Nordic countries, industry-level collective agreements usually set only minimum wages, but because these are present in all sectors and unions are still quite powerful, these countries have some of the highest minimum wages in the world, despite the fact that none of them has a statutory minimum wage (Meyer, 2016).16 For strong firms, these agreements are of minimal importance, as workers tend to make above the minimum rates. But in weaker firms, and especially for firms in sectors which are low-wage sectors in other countries (such as fast food or transportation), the sector-level collective agreements guarantee relatively high wages for all. As a result, there is almost no phenomenon of ‘the working poor’ in Sweden and other Scandinavian countries, unlike the USA or Germany (Gautie and Schmitt, 2010). One of the most notable trends across Western political economies has been the increase in economic inequality, which is broadly attributed to the decline of trade unions, among other factors (Scheve and Stasavage, 2009; Western and Rosenfeld, 2011). While recent work has attributed the growth in American inequality to financialization (Lin and Tomaskovic-Devey, 2013), the link between technological change and inequality has been more contested (Kristal and Cohen, 2017; cf. Autor et al., 2008). Regardless of how stock market development and technological change affect inequality directly, they may have an indirect effect on inequality through their effect on trade unions. My results suggest that while these affect unions in different ways, they both contribute to their decline. Future research should further investigate the mediating effect of trade unions on the relationship between financialization, technological change and inequality. This article is only one small contribution toward the broader project of explaining how finance and technological change matter for labor relations. One possibility, which I did not explore, is that the characteristics of the equities holders, such as whether they are pension funds, private equity firms or hedge funds may matter for labor relations. Pension funds tend to be large holders of equities and when unions have a strong say in how these funds invest, they may insist on worker-friendly employment practices. Contrarily, if private equity firms or hedge funds are large equities holders, we might expect an even more aggressive stance toward labor. Gospel et al. (2011) argue that private equity firms have mid-term time horizons and are most likely to take an activist role in corporate governance, while hedge funds are less likely to take an activist role, but have shorter time horizons and have as a primary goal pressuring managers to increase returns to shareholders. Nevertheless, in case studies of select firms in Spain, Germany and the UK, they find little evidence of a substantial change in labor relations after assumption of ownership by private equity funds or increased ownership by hedge funds. Future research using cross-national data could develop a coding of unions’ involvement in pension schemes to determine the degree to which this conditions the effect of stock market development on labor outcomes. It might be possible to study the effect of ownership by private equity firms and hedge funds in linked employer–employee datasets, which are increasingly available across advanced democracies. Linked employer–employee data could also help address questions of whether financialization has contributed to labor market dualism. While there has been a substantial amount of work on how past technological change has affected the employment prospects for different skill groups of workers, the effect of technological change on job tasks continues to change. Until recently, technological change had the greatest adverse effect on workers in middle skill, routine task occupations. But recent research shows that future technological change will adversely affect workers in the lowest-skill occupations. Frey and Osborne (2017) predict that 47% of jobs in the USA, mostly in lower-skills occupations, will be susceptible to automation in the coming decades. This threatens to further undermine cross-skill group workplace solidarity. But it also raises an even more important issue: if the least-educated, lowest-skill individuals become largely unemployable at any acceptable wage, how will we deal with this as a society, particularly when pro-worker institutions, such as trade unions have been greatly weakened? One policy, which has received a great deal of attention is the basic minimum income. It has even received substantial support from libertarian-leaning members of the tech community (Gordon, 2014). But this will be a political challenge as the basic minimum income failed to pass in a 2015 Swiss referendum 77–23%. Moreover, there is reason to be skeptical that most people would be happy to trade the opportunity to work a meaningful job for a basic minimum income. Acknowledgements I would like to give special thanks to Katherine Jackson for her comments on numerous drafts of this article and many discussions on law and finance. I would also like to thank Noam Gidron, Gregory Jackson, Stefan Thewissen and participants at the 2014 EPSA and SASE annual meetings for helpful written and verbal comments. Supplementary material Supplementary material is available at Socio-Economic Review online. Footnotes 1 This is particularly relevant in the Nordic countries, where union membership is voluntary and unions rely entirely on industrial action to pressure employers to sign collective agreements. 2 Katzenstein (1985) argues, however, that trade openness was not incompatible with worker-friendly democratic corporatism in small European states. Democratic corporatism helped reduce class conflict, which enabled more stable production relations and economic growth. 3 I do not address the origins of financialization here. Knafo and Dutta (2016) argue that financialization in the USA was actually driven by managers in the 1960s, who wanted to raise funds to build large conglomerates. 4 Knafo and Dutta (2016) argue that the norm of shareholder value was developed by managers not as a way of prioritizing shareholders, but as a way to value performance in financial markets. They also argue, however, that once the shareholder value norm took over, managers faced pressure (due to the possibility of takeovers) to increase share prices. While their account of how financialization affects management’s behavior is somewhat different from that in this article, managers’ response to stock prices is similar in both accounts. 5 In several countries, pension funds, which are often subject to substantial union influence, constitute a substantial percentage of equities investment. In countries where pension funds are major equities investors and unions have a great deal of influence over these, the pressure on managers to deliver shareholder value at the expense of labor may be substantially reduced. See McCarthy et al. (2016). 6 Jackson and Thelen (2015) find that despite liberalization of finance in Germany, there has been a dual trend in industrial relations. Higher-skill workers in the core manufacturing sectors have retained coordinated labor relations while lower-skill workers in manufacturing and service sectors have been subject to outsourcing and downward wage pressure. 7 This latter mechanism may be less significant in countries with Ghent unemployment schemes, such as Sweden, Denmark and Finland, where the unemployment benefits system is administered by unions and union membership is required for participation in it. 8 According to the authors, a task ‘is routine if it can be accomplished by machines following explicit programmed rules’. This includes ‘many manual tasks … such as monitoring the temperature of a steel finishing line or moving a windshield into place on an assembly line’, but also cognitive tasks, such as ‘calculating, coordinating, and communicating functions of bookkeepers, cashiers, telephone operators, and other handlers of repetitive information-processing tasks’. (1283–1284). 9 These data, however, vary widely in completeness across countries, with some countries, such as Australia, Germany, Canada and the USA having data for almost the entire period and others such as France, The UK and Switzerland having data for relatively few years. 10 The only variables for which panels had missing variables within otherwise complete blocks are RTI and inward migration, thus both of these contain linearly interpolated values. I did not extrapolate either before or after the first/last year observation on the most incomplete variable, resulting in an unbalanced panel. 11 I ran a Wooldridge test for autocorrelation and a Fisher-type test for panel stationarity. The former rejected the null hypothesis of no autocorrelation and the latter rejected the null hypothesis of panel stationarity. 12 The ECM is based on the Autoregressive Distributed Lag model: Yt = α0 + α1Yt−1 + β0Xt + β1Xt−1 + εt. It is generated by subtracting Yt−1 from both sides and adding and subtracting β0Xt from the right-hand side (DeBoef and Keele, 2008). 13 Nickell (1981) demonstrates bias in the fixed effects model in the presence of a lagged dependent variable; however, the bias is of the order 1/T, meaning that it should be fairly minimal in TSCS settings with relatively large T, such as this one. Wilson and Butler (2007) find that the bias is minimal when T > 20 and that fixed effects performs as well as more complicated estimators. 14 To give a better sense of the meaning of 1 SD of routine task intensity, it is approximately the difference in routine task intensity of employment in the USA (1986–2007), Germany (1992–2004) and Denmark (1984–1997). See the Supplementary Appendix for the full descriptive statistics. 15 These null findings for the effect of trade union institutions on occupational employment are similar to those of Oesch (2013) for the effects of the introduction of the minimum wage in UK and labor market deregulation in Germany on occupational employment. We would have expected low-wage employment to grow more in Germany, where there was deregulation than in Great Britain, which increased low-wage regulation. The opposite, however, was the case; low-wage employment grew more in Great Britain than in Germany. 16 Collective agreement coverage is above 90% in Denmark, Finland and Sweden (Visser, 2013). Jackson and Thelen (2015) argue that the predominant ownership structure in Denmark, whereby family foundations hold controlling stakes in firms, has provided patient capital and allowed for solidarity and stability in the industrial relations system. References Acemoglu D., Aghion P., Violante G. ( 2001) ‘Deunionization, Technical Change, and Inequality’, Carnegie-Rochester Conference Series on Public Policy , 55, 229– 264. Google Scholar CrossRef Search ADS   Ahlberg K, Bruun N. ( 2005) ‘Sweden: Transition through Collective Bargaining’. In Blainpain R. (ed) Collective Bargaining and Wages in Comparative Perspective: Germany, France, The Netherlands, Sweden, and the United Kingdom , The Netherlands, Kluwer Law International, pp. 117– 145. Ahlquist J. ( 2010) ‘Building Strategic Capacity: The Political Underpinnings of Coordinated Wage Bargaining’, American Political Science Review , 104, 171– 188. Google Scholar CrossRef Search ADS   Amable B., Ernst E., Palombarini S. ( 2005) ‘How do Financial Markets Affect Industrial Relations: An Institutional Complementarity Approach’, Socio-Economic Review , 3, 311– 330. Google Scholar CrossRef Search ADS   Aspara J., Pajunen K., Tikkanen H., Tainio R. ( 2014) ‘Explaining Corporate Short-termism: Self-Reinforcing Processes and Biases among Investors, the Media and Corporate Managers’, Socio-Economic Review , 12, 667– 693. Google Scholar CrossRef Search ADS   Autor D., Levy F., Murnane R. ( 2003) ‘The Skill-Content of Recent Technological Change: An Empirical Investigation’, Quarterly Journal of Economics , 118, 1279– 1333. Google Scholar CrossRef Search ADS   Autor D., Katz L., Kearney M. ( 2008) ‘Trends in U.S. Wage Inequality: Revising the Revisionists’, Review of Economics and Statistics , 90, 300– 323. Google Scholar CrossRef Search ADS   Baccaro L., Howell C. ( 2011). ‘A Common Neoliberal Trajectory: The Transformation of Industrial Relations in Advanced Capitalism’, Politics and Society , 39, 521– 563. Google Scholar CrossRef Search ADS   Beck T., Demirgüç-Kunt A., Levine R. ( 2000) ‘A New Database on Financial Development and Structure’, World Bank Economic Review , 14, 597– 605. Google Scholar CrossRef Search ADS   Beck N., Katz J. ( 1995) ‘What to Do (and Not to Do) with Time-Series Cross-Section Data’, American Political Science Review , 89, 634– 647. Google Scholar CrossRef Search ADS   Bertrand M., Schoar A., Thesmar D. ( 2007) ‘Banking Deregulation and Industry Structure: Evidence from the French Banking Reforms of 1985’, Journal of Finance , 62, 597– 628. Google Scholar CrossRef Search ADS   Beyer J., Hassel A. ( 2002) ‘The Effects of Convergence: Internationalization and the Changing Distribution of Net Value Added in Large German Firms’, Economy and Society , 31, 309– 332. Google Scholar CrossRef Search ADS   Black B., Gospel H., Pendleton A. ( 2007) ‘Finance, Corporate Governance, and the Employment Relationship’, Industrial Relations , 46, 643– 650. Bond P., Edmans A, Goldstein I. ( 2012) ‘The Real Effects of Financial Markets’, Annual Review of Financial Economics , 4, 339– 360. Google Scholar CrossRef Search ADS   Brady D. ( 2007) ‘Institutional, Economic, or Solidaristic? Assessing Explanations for Unionization across Affluent Democracies’, Work and Occupations , 34, 67– 101. Google Scholar CrossRef Search ADS   Brady D., Huber E., Stephens J. ( 2014) Comparative Welfare States Data Set . University of North Carolina and WZB Berlin Social Science Center. Accessed at http://huberandstephens.web.unc.edu/common-works/data/ on May 15, 2014. Capoccia G., Kelemen R.D. ( 2007) ‘The Study of Critical Junctures: Theory, Narrative, and Counterfactuals in Historical Institutionalism’, World Politics , 59, 341– 369. Google Scholar CrossRef Search ADS   Choi M. ( 2001) Threat Effect of Foreign Direct Investment on Labor Union Wage Premium , PERI Working Paper No. 27. Accessed at SSRN: https://ssrn.com/abstract=335480 on August 5, 2013. Claessens S., Klingebiel D., Schmukler S. ( 2006) ‘Stock Market Development and Internationalization: Do Economic Fundamentals Spur Both Similarly?’ Journal of Empirical Finance , 13, 316– 350. Google Scholar CrossRef Search ADS   Dallery T. ( 2009) ‘Post-Keynesian Theories of the Firm under Financialization’, Review of Radical Political Economics , 41, 492– 515. Google Scholar CrossRef Search ADS   Darcillon T. ( 2015) ‘How Does Finance Affect Labor Market Institutions? An Empirical Analysis in 16 OECD Countries’, Socio-Economic Review , 13, 477– 504. Google Scholar CrossRef Search ADS   DeBoef S., Keele L. ( 2008) ‘Taking Time Seriously’, American Journal of Political Science  52, 184– 200. Google Scholar CrossRef Search ADS   Deeg R., Hardie I. ( 2016) ‘What is Patient Capital and Who Supplies It?’ Socio-Economic Review , 14, 627– 645. Google Scholar CrossRef Search ADS   Dinlersoz E., Greenwood J. ( 2016) ‘The Rise and Fall of Unions in the United States’, Journal of Monetary Economics , 83, 129– 146. Du Caju P., Gautier E., Momferatou D., Ward-Warmedinger M.E. ( 2008) Institutional Features of Wage Bargaining in 23 European Countries, the US and Japan (December 1, 2008). Banque de France Working Paper No. 228. doi:10.2139/ssrn.1677920. Frey C.B., Osborne M. ( 2017) ‘The Future of Employment: How Susceptible are Jobs to Computerization’, Technological Forecasting and Social Change , 114, 254– 280. Google Scholar CrossRef Search ADS   Gautie J., Schmitt J. ( 2010) Low-Wage Work in the Wealthy World , New York, NY, Russell Sage Foundation. Goldin C., Katz L. ( 2008) The Race between Education and Technology , Cambridge, MA, Harvard University Press. Goos M., Manning A., Salomons A. ( 2014) ‘Explaining Job Polarization: Routinization and Offshoring’, American Economic Review , 104, 2509– 2526. Google Scholar CrossRef Search ADS   Gordon N. ( 2014, August 5) ‘The Conservative Case for a Guaranteed Basic Income.’ The Atlantic , accessed at https://www.theatlantic.com/politics/archive/2014/08/why-arent-reformicons-pushing-a-guaranteed-basic-income/375600/ on June 13, 2016. Gospel H., Pendleton A. ( 2003) ‘Finance, Corporate Governance and the Management of Labour: A Conceptual and Comparative Analysis.’ British Journal of Industrial Relations , 41, 557– 582. Google Scholar CrossRef Search ADS   Gospel H., Pendleton A., Vitols S., Wilke P. ( 2011) ‘New Investment Funds, Restructuring, and Labor Outcomes: A European Perspective’, Corporate Governance: An International Review , 91, 276– 289. Google Scholar CrossRef Search ADS   Graham J.R., Harvey C.R., Shiva R. ( 2005). ‘The Economic Implications of Corporate Financial Reporting’, Journal of Accounting and Economics , 40, 3– 73. Google Scholar CrossRef Search ADS   Hall P., Soskice D. (eds) ( 2001) Varieties of Capitalism: The Institutional Foundations of Comparative Advantage , New York, NY, Oxford University Press. Google Scholar CrossRef Search ADS   Hardie I., Howarth D., Maxfield S, Verdun A. ( 2013) ‘ Banks and the False Dichotomy in the Comparative Political Economy of Finance’, World Politics , 65, 691– 728. Google Scholar CrossRef Search ADS   Hirsch B. ( 2008) ‘Sluggish Institutions in a Dynamic World: Can Unions and Industrial Competition Coexist?’ Journal of Economic Perspectives , 22, 153– 176. Google Scholar CrossRef Search ADS   Höpner M. ( 2005) ‘What Connects Industrial Relations with Corporate Governance? Explaining Institutional Complementarity’, Socio-Economic Review , 3, 331– 358. Google Scholar CrossRef Search ADS   Iversen T. ( 1996) ‘Power, Flexibility, and the Breakdown of Centralized Wage Bargaining: Denmark and Sweden in Comparative Perspective’, Comparative Politics , 28, 399– 436. Google Scholar CrossRef Search ADS   Jackson G., Petraki A. ( 2011) Understanding Short-termism: the Role of Corporate Governance , Stockholm, Glasshouse Forum. Jackson G., Thelen K. ( 2015). ‘Stability and Change in CMEs: Corporate Governance and Industrial Relations in Germany and Denmark.’ In Beramendi P., Häusermann S., Kitschelt H., Kriesi H (eds) The Politics of Advanced Capitalism , New York, NY, Cambridge University Press, pp. 305– 329. Google Scholar CrossRef Search ADS   Jaimovich N., Siu H. ( 2014) The Trend is the Cycle: Job Polarization and Jobless Recoveries, accessed at http://www.nirjaimovich.com/assets/jpjr.pdf on April 5, 2015. Jung J. ( 2015) ‘Shareholder Value and Workforce Downsizing 1981-2006’, Social Forces , 93, 1335– 1368. Google Scholar CrossRef Search ADS   Katzenstein P. ( 1985) Small States in World Markets: Industrial Policy in Europe , Ithaca, NY, Cornell University Press. Knafo S., Dutta S.J. ( 2016) ‘Patient Capital in the Age of Financialized Managerialism’, Socio-Economic Review , 14, 771– 788. Google Scholar CrossRef Search ADS   Kristal T., Cohen Y. (2017) ‘The Causes of Rising Wage Inequality: The Race between Institutions and Technology’, Socio-Economic Review , 15, 187– 212. Lazonick W., O'Sullivan M. ( 2000) ‘Maximizing Shareholder Value: A New Ideology for Corporate Governance’, Economy and Society , 29, 13– 35. Google Scholar CrossRef Search ADS   Lee C-S. ( 2005) ‘International Migration, Deindustrialization and Union Decline in 16 Affluent OECD Countries, 1962-1997’, Social Forces , 84, 71– 88. Google Scholar CrossRef Search ADS   Lin K-H., Tomaskovic-Devey D. ( 2013) ‘Financialization and U.S. Income Inequality, 1970-2008’, American Journal of Sociology , 118, 1284– 1329. Google Scholar CrossRef Search ADS   Mahoney J., Thelen K. ( 2010) ‘A Theory of Gradual Institutional Change.’ In Mahoney J., Thelen K. (eds) Explaining Institutional Change: Ambiguity, Agency, and Power , New York, NY, Cambridge University Press, 1– 38. McCarthy M., Sorsa V-P., van der Zwan N ( 2016) ‘Investment Preferences and Patient Capital: Financing, Governance, and Regulation in Pension Fund Capitalism.’ Socio-Economic Review , 14, 751– 769. Google Scholar CrossRef Search ADS   Meyer B ( 2016) ‘Learning to Love the Government: Trade Unions and Late Adoption of the Minimum Wage’, World Politics , 68, 538– 575. Google Scholar CrossRef Search ADS   Nickell S. ( 1981) ‘Biases in Dynamic Models with Fixed Effects’, Econometrica , 49, 1417– 1426. Google Scholar CrossRef Search ADS   OECD ( 2014). OECD.Stat, accessed at stats.oecd.org on April 15, 2014. Oesch D. ( 2013) Occupational Change in Europe: How Technology and Education Transform the Job Structure , Oxford, UK, Oxford University Press. Google Scholar CrossRef Search ADS   Oesch D., Menes J.R. ( 2011). ‘Upgrading or Polarization? Occupational Change in Britain, Germany, Spain and Switzerland, 1990-2008’, Socio-Economic Review , 9, 503– 531. Google Scholar CrossRef Search ADS   Pontusson J., Swenson P. ( 1996) ‘Labor Markets, Production Strategies, and Wage Bargaining Institutions: The Swedish Employer Offensive in Comparative Perspective’, Comparative Political Studies , 29, 223– 250. Google Scholar CrossRef Search ADS   Scheve K., Stasavage D. ( 2009). ‘Institutions, Partisanship, and Inequality in the Long Run’, World Politics , 61, 215– 253. Google Scholar CrossRef Search ADS   Schnabel C. ( 2013) ‘Union Membership and Density: Some (Not So) Stylized Facts and Challenges’, European Journal of Industrial Relations , 19, 255– 272. Google Scholar CrossRef Search ADS   Scruggs L., Lange P. ( 2002) ‘Where Have All the Members Gone? Globalization, Institutions, and Union Density’, Journal of Politics , 64, 126– 153. Google Scholar CrossRef Search ADS   Slaughter M. ( 2007) ‘Globalization and Declining Unionization in the United States.’ Industrial Relations , 46, 329– 346. Spitz-Oener A. ( 2006) ‘Technical Change, Job Tasks, and Rising Educational Demands: Looking Outside the Wage Structure’, Journal of Labor Economics , 24, 235– 270. Google Scholar CrossRef Search ADS   Thelen K. ( 2014) Varieties of Liberalization and the New Politics of Social Solidarity . New York, NY, Cambridge University Press. Google Scholar CrossRef Search ADS   United Nations. ( 1978). 1977 Demographic Yearbook, accessed at https://unstats.un.org/unsd/demographic/products/dyb/dybsets/1977%20DYB.pdf on February 1, 2014. United Nations. ( 1987). 1985 Demographic Yearbook, accessed at https://unstats.un.org/unsd/demographic/products/dyb/dybsets/1977%20DYB.pdf on February 1, 2014. van der Zwan N. ( 2014) ‘Making Sense of Financialization’, Socio-Economic Review , 76, 538– 559. Visser J. ( 2013). Data Base on Institutional Characterictics of Trade Unions , Wage Setting, State Intervention, and Social Pacts, 1960-2012. Amsterdam Institute for Advanced Labor Studies, University of Amsterdam. Version 4.0. Wallerstein M. ( 1990). ‘Centralized Bargaining and Wage Restraint’, American Journal of Political Science , 34, 982– 1004. Google Scholar CrossRef Search ADS   Western B. ( 1997). Between Class and Market: Postwar Unionization in Capitalist Democracies , Princeton, NJ, Princeton University Press. Western B., Rosenfeld J. ( 2011). ‘Unions, Norms, and the Rise in U.S. Wage Inequality’, American Sociological Review , 76, 513– 537. Google Scholar CrossRef Search ADS   Wilson S., Butler D. ( 2007). ‘A Lot More to Do: The Sensitivity of Time-Series Cross-Section Analyses to Simple Alternative Specifications’, Political Analysis , 15, 101– 123. Google Scholar CrossRef Search ADS   © The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Socio-Economic Review Oxford University Press

Financialization, Technological Change, and Trade Union Decline

Socio-Economic Review , Volume Advance Article – Aug 8, 2017

Loading next page...
 
/lp/ou_press/financialization-technological-change-and-trade-union-decline-0QMVO41LG6
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
1475-1461
eISSN
1475-147X
D.O.I.
10.1093/ser/mwx022
Publisher site
See Article on Publisher Site

Abstract

Abstract Recent research finds that financialization and technological change have had a variety of negative effects on labor, including reducing low-skill workers’ wages and increasing income inequality. In this article, I examine the effect on trade unions of one type of financialization, equity market development and one type of technological change, routine-biased technological change. I argue that we should conceptualize trade union strength in two dimensions: (a) the strength of their institutional structures, such as the degree of wage bargaining coordination and the degree to which firms can deviate from collective agreements; (b) the strength of their membership. Using data for 21 OECD countries from 1970 to 2010, I find a negative effect of equity market development on unions’ institutional structures, but not on union membership. Contrarily, I find that routine-biased technological change has a negative effect on union density, but an inconsistent relationship with the strength of unions’ institutional structures. 1. Introduction Labor markets in advanced democracies have undergone dramatic changes in recent decades. There has been a shift from industrial to service employment, the Internet has revolutionized service delivery, and economic inequality has increased. One of the major contributions to the latter phenomenon has been the decline of trade unions. While trade unions remain quite strong in several advanced democracies, they have been weakened in most countries in several ways, with decreasing coverage of collective bargaining agreements, reduction in the scope of collective agreements and declining membership (Baccaro and Howell, 2011). Work on trade union decline has devoted great attention to how structural economic changes, such as technological change, offshoring and global capital flows have adversely affected trade unions. One of these structural changes, which has been the focus of much recent work in economic sociology, is ‘financialization’, an increased role for financial actors in firms’ management and an increased importance of revenue from financial activities to firms’ bottom lines (van der Zwan, 2014). While there is a substantial literature on how financialization adversely impacts a variety of labor outcomes, including raising the probability of firm downsizing (Jung, 2015) and increasing income inequality (Lin and Tomaskovic-Devey, 2013), there has been little work on how financialization affects unions. Another important structural change in Western democracies has been declining employment in ‘routine task’ jobs: jobs in which employees performed conceptually simple, repetitive tasks. As computing power has increased, these jobs, which can be programed as algorithms, can be performed by machines (Autor et al., 2003). Recent work in labor economics has found that this ‘routine-biased’ technological change has largely affected jobs that were in the middle of the wage distribution, such as assembly line and clerical work (Spitz-Oener 2006). As a result of this, there has been labor market ‘polarization,’ an increase in employment in low- and high-skill jobs, across advanced democracies (Goos et al., 2014). In this article, I examine how financialization and routine-biased technological change have affected various measures of trade union strength. While there are several different definitions of financialization in the economic sociology literature, I focus on the role of equities markets. I argue that stock market development, in particular, should have a negative effect on trade union strength because equity investors gain more from reduced labor costs than holders of other financial instruments, such as bonds. Investors will press managers to seek greater flexibility in their employment practices, including in collective agreements. Because much management compensation comes in the form of stock options and managers fear firm takeovers if their stock price remains low, they have personal incentives to pursue such policies. I also argue that routine task employment will be positively associated with trade union strength, i.e. when there is high employment in routine task jobs, trade unions will be stronger. Before increases in computing power, industrial production and many types of clerical work were very labor intensive, requiring large numbers of similarly skilled individuals. Because these individuals had similar interests and were concentrated within workplaces, they had a great deal of power to demand strong unions. When routine task employment declines, this source of collective power is lost. New jobs are less likely to be unionized as they are typically in smaller firms, which do not have a history or culture of unionization. In contrast to much work on union strength, which focuses on union density or wage bargaining coordination separately, I focus on multiple measures: union density, wage bargaining coordination, wage bargaining centralization and the presence of ‘opening clauses’ in collective agreements, which allow firms to opt out of collective agreement provisions when facing economic hardship. I conceptualize these variables into two categories: institutional structure variables and membership. Recent work on institutional change argues that there is typically wide latitude for flexibility within institutions and that there may be institutional change without change in the formal structure of an institution (Mahoney and Thelen, 2010). I argue that wage bargaining coordination, centralization and opening clauses capture different aspects of institutional structure while union density captures membership. Furthermore, stock market development and routine-biased technological change may differentially affect these variables. In order to assess these arguments empirically, I analyze data for 21 OECD countries from 1970 to 2010. Using error correction models (ECM), which include both changes and levels of the independent variables, I find that routine task employment and stock market development have differential effects on the institutional structure and membership variables. I find that changes in stock market development adversely affect the institutional structure variables but have little consistent effect on union density. I find that union density is more likely to grow when routine task employment is growing and that the rate of growth is highest at higher levels of routine task employment. I find, however, that the level of routine task employment is actually a negative predictor of both wage bargaining coordination and centralization, suggesting that while routine task employment has a positive effect on union membership, institutional structures do not become stronger when routine task employment is high. I run two types of robustness checks to help rule out reverse causality: (a) models where I lag changes in the independent variables and (b) models where I regress stock market development and routine task employment on the various measures of union strength. I find that the original results are largely robust to the first type of robustness check. Regressions of stock market development and routine task employment on the four measures of union strength show no consistent relationships, suggesting that reverse causality is not driving the results. I proceed as follows: in Section 2, I present my conceptualization of the trade union variables. In Section 3, I review the literature on trade union decline. In Section 4, I both review the literature on how financialization and routine-biased technological change have affected labor and develop my arguments for how these should affect trade unions. This is followed in Section 5 by a discussion of data, methods and predictions for each of the main independent variables. In Section 6, I present the main results and the two robustness checks. Section 7 follows with a discussion of the results in broader context and concludes. 2. Conceptualizing union strength: institutional structures and union membership There have been a variety of explanations for both union density decline and decentralization of collective bargaining, but there has been little work which analyzes these two phenomena together. It is important to do so, however, because their timing differs and the factors which affect them may differ as well. Union density decline has been gradual and consistent in most countries over time, while usage of opening clauses increased starting in the 1990s and decentralization of wage bargaining occurred largely in the 1980s and early 1990s. Structural changes, such as technological change and offshoring, may affect union membership, but national institutions, such as wage bargaining coordination, may experience little change due to path dependence or institutional complementarities. In addition to differential timing in how they have changed, I argue that these union variables can be conceptualized into two categories: institutional structure, including wage bargaining coordination, centralization, and opening clauses, and union membership, including union density. While institutional structures can certainly help explain union membership (and vice versa), institutional structures can change without necessarily affecting membership. If, for example, a high degree of cross-sector coordination in wage bargaining creates conflict between different skill groups of workers, individual unions may be able to better grow their membership if cross-sector coordination is reduced. Job losses due to the decline of heavily unionized occupations may cause a decline in union membership without causing a decline in the institutional structure. At the same time, within levels of wage bargaining coordination and centralization, the institutional structure can be more or less rigid. This is in part a function of the presence of opening clauses, which can increase the scope of bargaining at the local level and allow employers to deviate from the conditions of centralized collective agreements when facing economic hardship. Recent work has shown that the broadest structures of collective bargaining, wage bargaining coordination and centralization, have not greatly changed in recent years (Du Caju et al., 2008; Thelen, 2014). Collective bargaining has occurred primarily at the industry level in most European countries since the 1990s. But despite stability in the broadest level of institutional structure, there has been a substantial degree of within-structure weakening. In the Nordic countries, the content of industry-level collective agreements has changed, with collective agreements primarily specifying minimum wages, rather than full wage scales as in the post-war decades (Ahlberg and Bruun, 2005). In Germany, industry-level collective agreements have increasingly included ‘opening clauses’, which allow participating employers to deviate from the provisions in the agreement under certain economic conditions. These opening clauses represent both a weakening at the central level, in that national unions and employers’ associations allow individual employers to deviate from them, and at the local level, in that their usage means that the agreement’s conditions do not always govern employment. Union density has declined in almost all advanced democracies. Although this does not necessarily prevent unions from setting high wages through collective agreements, it may make it more difficult for unions to rally enough people to effectively strike.1 While much work of historical institutionalism has focused on abrupt change at ‘critical junctures’, unsettled times at which social norms have become weakened and there is a possibility to redesign institutions (Capoccia and Kelemen, 2007), recent work has argued that institutional change comes largely through gradual but continual change. Mahoney and Thelen (2010) argue that rule interpretation and enforcement autonomy can fundamentally change the functioning of an institution without changing its structure. Regarding industrial relations institutions, Thelen (2014) argues that the gradual hollowing out of centralized collective agreements and their replacement by less redistributive local bargaining, a process which she calls ‘embedded flexibilization,’ has allowed unions in these countries to maintain a high degree of solidarity. Centralized collective agreements no longer strictly link wage increases in different occupations and regions to each other, which reduces the between-skill group conflict that existed in these countries in the 1980s. Yet they are still responsible for setting high minimum wages, which has allowed Scandinavian countries to largely avoid working poverty. So it is still important to study institutional structures, both because they continue to matter for important outcomes, and because it is an open question whether they are amenable to the same types of explanations as within-institution strength and union membership. 3. Previous explanations for trade union decline While existing scholarship tends to separately address institutional structure and within-institutional structure change in wage bargaining institutions separately, scholars have given similar types of explanations for both. Explanations for trade union decline can be grouped into four broad categories: (a) political (b) national institutions (c) globalization (d) deindustrialization. One of the foremost explanations for union decline generally has been that right-wing politicians have become more anti-union and have actively worked to weaken unions (Brady, 2007). The most famous examples are the USA, where Ronald Reagan fired striking air-traffic controllers in 1981, which began an anti-union turn in American politics and the UK, where Margaret Thatcher’s Conservative Party passed far-reaching union reforms in the 1980s, removing much of unions’ strike immunity and implementing more stringent conditions on union votes. Another type of explanation emphasizes the role of national institutions in strengthening or weakening collective bargaining. This approach is associated with Varieties of Capitalism, according to which employers assent to strong unions in coordinated market economies (CME), such as Germany and Sweden, because they produce products that require specific skills and more cooperative labor relations (Hall and Soskice, 2001). Other scholars have found that organizations like coordinated wage bargaining and works councils help strengthen union membership (Scruggs and Lange, 2002). Western (1997) found that union membership is higher in countries with a Ghent unemployment system, where union membership is required to participate in the unemployment insurance system. Ahlquist (2010) found that coordinated wage bargaining is more likely to occur in countries in which unions have established a centralized strike fund. Perhaps the foremost types of explanations have been based on various aspects of globalization, such as increased trade, capital mobility and immigration, and on deindustrialization. As countries reduce barriers to trade, manufacturers in highly developed countries can begin to take advantage of lower labor costs in developing countries. Offshoring and trade impact largely lower skill workers in manufacturing and industry, who were likely to be union members.2 Increased capital mobility allows firms to locate production in countries with lower labor costs and repatriate profits to their home countries. Along similar lines, scholars have found that increased foreign direct investment decreases unions’ wage premia and union density (Choi, 2001; Slaughter, 2007). Finally, Lee (2005) found that inward immigration reduced union density. Deindustrialization is another prominent recent explanation for trade union decline, as industrial jobs were among those most likely to be unionized (Hirsch, 2008). There is a simple explanation based on deindustrialization, in which unions decline due to attrition when factories are offshored or jobs replaced by new technology. But there is also a more sophisticated version, in which technological change differentially shapes the bargaining power of different skill groups of workers, who then have differing preferences over unionization. With such skill-biased technological change (SBTC), both employment and wages in high-skill occupations increase due to their complementarity with new technology. These decrease, however, in lower-skill occupations. Because of this increased differential in individual bargaining leverage, high-skill workers, who have more bargaining leverage, are less willing to join low-skill workers to support unions (Acemoglu et al., 2001). A similar explanation has been given for the decline of coordinated wage bargaining in Europe in the 1980s and 1990s. New technology gave rise to ‘diversified quality production’, which increased the global competitiveness of high-skill manufacturing workers (Iversen, 1996; Pontusson and Swenson, 1996). These workers demanded higher wage differentials and unions representing them withdrew from multi-sector wage bargaining institutions. 4. Financialization, technological change, and labor market outcomes 4.1 Finance and labor While there has been a substantial amount of work on globalization, deindustrialization and union decline, recent developments in economic sociology on financialization and in labor economics on routine-biased technological change suggest new mechanisms through which to advance the study of trade union decline. In foundational work on the relationship between finance and labor, Hall and Soskice (2001) argued that ‘patient capital’ provided by banks in long-term relationships with firms is a central institution of CMEs. It enabled long-term, stable relationships with unions, in which banks promised to finance skill investments and protect workers’ specific skills during market downturns in return for success in niche, specific skill-intensive markets. But Hall and Soskice also recognized that financial markets had become more global, that financial actors had become more heterogeneous, and that finance providers might become less willing to underwrite such relationships. Indeed Hardie et al. (2013) find that even large banks in CMEs have become less willing to provide patient capital because they are dependent on international markets for funding, which makes them less able to have such long-term commitments. Empirical work on finance and labor has found that liberalized finance results in lower shares of firm revenue going to labor over a variety of outcome variables. Bertrand et al. (2007) found that after reforms reducing government intervention into bank lending in France in the mid-1980s, average wage increases were substantially lower in more bank-dependent sectors. They also found that worse-performing firms were more likely to outsource. Scholars have found that adoption of ‘shareholder value’ practices increased dividend payouts to shareholders and decreased the share of revenue going to workers (Lazonick and O’Sullivan, 2000; Beyer and Hassel, 2002). Lin and Tomaskovic-Devey (2013) found that American firms’ increasing dependence on financial income in recent decades was associated with a lower labor share of income, increased top executive share of compensation, and increasing earnings dispersion among workers. Jung (2015) found that increased reliance on institutional investors was associated with a greater probability of workforce downsizing for a sample of American firms from 1981 to 2006. Similar to my own analysis, Black et al. (2007), using cross-sectional data from the 1990s, found that wage bargaining centralization is lower in countries with higher equity market development. Also closely related, Darcillon (2015) found that higher shares of employment and total value added in finance are associated with lower levels of a composite measure of workers’ bargaining power. In this article, I focus specifically on how equity market development should affect trade unions. While previous work has focused on broader conceptualizations of financialization, such as employment in finance and share of value added in financial activities (Darcillon, 2015), I focus on stock market development. This is both because equity investors, despite their heterogeneity, should have certain common objectives with respect to firms and because firm managers will be influenced by investor preferences and the price of shares, both as components of their compensation and as performance metrics.3 One shortcoming of previous work on financialization and labor outcomes is that different types of investors may have very different preferences regarding firm performance and that measures combining different types of financial investments may obscure more than they reveal. Equity investors, for example, have different incentives than bond investors. The former make money when share prices rise while the latter earn a set rate of return that can only be wiped out through bankruptcy or firm restructuring. Because of this, bondholders should not be as concerned with firms’ profitability and should be more patient investors than equity holders (Deeg and Hardie, 2016). How should we expect equities and equity prices to matter for management of the firm? First, it is important to distinguish between how they will matter for financial actors and how they will matter for management. Equities entitle holders to a share of the firms’ earnings and they make money on their shares when the share prices increases. Their primary motive should be to achieve higher share prices, which comes through increased profitability. A firm may increase its profitability through firm growth, but it may also achieve this by limiting itself to a set of core, profitable activities and by reducing the share of revenue going to other stakeholders. In recent decades, the focus has shifted from growing the firm to focusing on a narrower set of competencies and aiming for high profitability (Dallery, 2009). Dallery (2009) argues the preference for high profitability has resulted in lower investment and growth rates than would be preferred by both management and labor. Shareholder value, the idea that the primary purpose of corporations is to make money for their shareholders (owners), has become a powerful rallying cry for investors (Lazonick and O’Sullivan, 2000; van der Zwan, 2014). Two common demands have been for managers to increase the size of dividends paid on shares and to buy back the firm’s stock, which reduces the number of outstanding shares, thus raising share prices. Money spent on share buybacks could have otherwise been spent on capital upgrades or increased employment/wages (Gospel and Pendleton, 2003). Because institutions are under constant pressure to deliver returns, fund managers are under high pressure to deliver constant performance. They face a great deal of competition from other fund managers and evaluate their analysts’ performance frequently. There are also longstanding concerns that equities investors, while exhibiting heterogeneity, are often biased toward short-term performance, preferring measures that enhance stock performance in the short term at the expense of longer-term commitments (Jackson and Petraki, 2011). Dallery (2009) argues that investors’ preference for increasing profitability by lowering investment and firm growth will be especially strong if these investors have short-term time horizons. Amable et al. (2005) argue that time-horizons are important for behavior toward labor because when investors have long-time horizons, a cooperative relationship with labor will better maximize firm performance. Unless stock-based compensation requires an extended holding period, this can encourage management to take short-term measures, such as cutting labor costs or under-investment in capital to boost the stock price and their compensation. There are several ways in which these investor preferences can directly influence management. One, due to past efforts of equity investors, a substantial percentage of manager pay now comes from stock options, which incentivize managers to take actions that will ensure high stock prices (Bond et al., 2012). As mentioned above, there is a great deal of pressure from investors to buy back the firm’s shares in order to boost its stock price and to increase the dividends paid per share. Additionally, even though equities markets are secondary markets and do not directly fund the firm’s daily activities, lenders make decisions upon these prices and if the stock price underperforms, the firm may become a target for takeovers by activist investors (Höpner, 2005). These investors would likely replace existing management to realize latent profitability potential. From a behavioral finance perspective, managers derive a great deal of personal prestige from stock prices and may be willing to forego investment projects in order to make expected quarterly numbers (Graham et al., 2005). Even if they are not directly pressured by investors to do so, stock prices are commonly seen as important metrics of performance and managers will try to boost them to signal both to markets and the public that they are competently managing the firm (Aspara et al., 2014).4 How do these preferences matter specifically for trade unions? One way to increase profitability is to ensure a high degree of flexibility in labor contracts. Generally, investors will want greater management flexibility with respect to setting labor policy.5 They should be more opposed to institutions that impose rigidity on management’s policy setting. Examples of institutions which investors should oppose for this reason are centralized and coordinated wage bargaining, where decisions for all firms within an industry or region are made at a central level. This is problematic for the creation of shareholder value because these wages may be too high for certain firms, given their economic performance. With respect to unions, this means that the growth of the shareholder value paradigm should weaken institutions like centralized and coordinated wage bargaining, which will allow more localized union/management discretion and greater response to local conditions. It should also increase investors’ and managers’ interest in including opening clauses in collective agreements. These would allow managers greater autonomy to reduce wages and labor costs to limit losses during economic downturns. Investors’ preferences may not, however, equally affect all union variables equally. While a preference for firm-level policy autonomy suggests an aversion to centralized wage bargaining, the implications for union density, which measures union membership as a percentage of the workforce, are less clear. As long as union membership does not translate into greater policy rigidity, investors are less likely to be concerned with this. On the other hand, specific tactics that management may use to weaken unions and reduce the share of firm revenue going to labor, such as outsourcing non-core tasks to work agencies, offshoring aspects of production to lower labor cost countries, or relying more heavily on part-time and temporary workers, could either prevent workers from unionizing or weaken their incentives to do so.6 Then stock market development could have an indirect negative effect on union density. And generally if investors and management push for more decentralized wage setting and this in turn lead to lower wage increases for union members, they may be less likely to join unions and pay union dues because they do not see the costs of union membership as outweighing the benefits.7 4.2 Technological change and labor One of the predominant explanations for trade union decline has been deindustrialization, a decline in employment in industry and manufacturing (Lee, 2005; Hirsch, 2008). Economists have incorporated the heterogeneous effect of technological change across skill groups into models of workers’ preferences for trade union representation and trade union decline (Acemoglu et al., 2001; Dinlersoz and Greenwood, 2016). These theories have been based on SBTC, which holds that under recent technological change, there is a positive correlation between a worker’s skill level and their wages/employment opportunities (Goldin and Katz, 2008). In other words, the highest skill workers have the best employment opportunities and make the highest wages, whereas the lowest skilled workers have the poorest employment opportunities and have seen their wages either stagnate or decline. The models of SBTC and union decline posit that because technological change has increased the individual bargaining power of high-skill workers, they will be less likely to join lower-skill workers to support unions than they were in the past. Recent work in labor economics has, however, called the SBTC explanation for changes in occupational employment into question. Autor et al. (2003) argued that technological change, specifically improvements in computing power over the past few decades, has allowed for the replacement of routine task jobs.8 They found the greatest employment decline for US occupations from 1960 to 1998 in occupations that were routine task intensive. Further research has found decline in routine task employment to be prevalent across advanced democracies and that these jobs were most heavily concentrated in the middle of the wage distribution, leading to labor market ‘polarization’ (Spitz-Oener, 2006; Goos et al., 2014). How would we expect the decline of routine task employment and labor market polarization to affect trade unions? The occupations with the highest routine task content were those in industry/manufacturing and clerical work. These occupations were likely to be highly unionized because they required many workers in large workplaces performing repetitive tasks in order to produce finished goods or clerical tasks. Workers’ similar levels of skills and individual bargaining ability gave them a great deal of common interest in supporting unions and the collective power to do so. With the decline in routine task employment however, employment has sorted into high and low-skill occupations, increasing between-skill group heterogeneity in individual bargaining power. At the same time, the competitive pool for low-skill jobs has increased due to loss of employment in routine task occupations and these workers’ lower suitability for high-skill occupations (Jaimovich and Siu, 2014). This gives low-skilled workers less leverage to pressure employers to allow them to unionize. Therefore, we would expect routine task-biased technological change to cause union decline through three mechanisms: (a) that routine task occupations were among the most likely to be unionized and job loss in these occupations would cause a corresponding decline in union density; (b) that routine-biased technological change increases between-worker skill heterogeneity, which reduces workers demand to join together to support unions; (c) that formerly routine task workers can only substitute for workers in lower-skill jobs, which creates more competition for these jobs and gives employers greater leverage to resist workers’ attempts at unionization. In the previous section, I argued that we would expect investors to want an increase in firm-level flexibility over wage-setting policy and that higher stock market development should be associated with reductions in coordinated and centralized wage bargaining as well as increased usage of opening clauses, but not necessarily a decrease in union density. In contrast to stock market development, the clearest prediction for routine-biased technological change is for union density. Workers were particularly likely to be unionized in factories during the Fordist era because production required large numbers of similarly skilled workers, giving them a common interest in union representation. Research on plant-level unionization finds that newer and smaller workplaces are less likely to be unionized (Schnabel, 2013). If union membership is voluntary, workers in these workplaces would be less likely to want to become union members. There is no legacy of union relationships as in older firms and no norm among the workers of union membership at that firm. The relationship between routine task employment and unions’ institutional structures is less clear. As routine task employment declines, it may be more difficult to maintain coordinated wage bargaining because the workforce becomes more heterogeneous. But if unions and employers allow increased local and firm-level bargaining and greater possibility of deviation through opening clauses, formal coordination may continue to exist at the sector level. It might also be the case that through high minimum wages, high levels of wage bargaining coordination price out lower-skill jobs, which would mean that the occupational structure becomes less polarized when coordination and centralization are high. In any case, the prediction for routine-biased technological change is clearer for union membership than for unions’ institutional structures. 5. Data and methods In order to empirically test these theories, I have assembled a dataset for 21 OECD countries from 1970 to 2010. My main variable StMkt is an average of stock market capitalization/GDP and stock market value traded/GDP compiled largely from data for 1975–2004 from Claessens et al. (2006) complimented by data from the Financial Development and Structure Dataset (Beck et al., 2000). These two elements capture the amount and flow of equity investment, both of which should affect firms’ decision-making. I take this measure of stock market development as a proxy for the pressure that equity prices place on investors’ preferences and managements’ decision-making. When a country’s level of stock market development is higher, the stock market plays a greater role in the economy and the importance of equity investors’ preferences for delivering shareholder value will be higher, as will managements’ responsiveness to equity prices. Data on occupational employment come from LABORSTA, which contains ISCO one-digit occupational employment for OECD countries from 1970 to 2010.9 Codings of occupational routine task intensity come from Autor et al. (2003). I generate a variable RTI, a country-year measure of the routine task intensity of overall employment, by weighting employment in each occupational category as a percentage of total employment by its routine task intensity score from Autor et al. Data for union variables, including union density, coordination, centralization, and presence of opening clauses, works councils and a strike fund come from Visser (2013). Data on economic variables come from the Comparative Welfare States Dataset and various sources within (Brady et al., 2014). Data on migration come from OECD Stats Extracts and the 1977 and 1985 UN Demographic Yearbooks (United Nations, 1978, 1987). I linearly interpolated missing observations within an otherwise complete block of observations within country panels.10 Due to the presence of both autocorrelation and panel non-stationarity, I use ECM, as recommended by DeBoef and Keele (2008).11 The basic ECM takes the following form:   ΔYt= α0+α1*Yt−1+ β0*ΔXt+ β1*Xt−1+ εt where:   α1*= (α1– 1)β0*= β0β1*= β0+β112 I include four types of models for all dependent variables: models with random effects, with country fixed effects, and with country and either 5-year or year fixed effects, all accounting for AR(1) autocorrelation and with panel corrected standard errors (Beck and Katz, 1995).13 While I present models both without and with country fixed effects, I prefer the country fixed effects models because in addition to the possibility of unobserved, relatively fixed differences between countries, equities ownership differs in ways which may have implication for labor relations. In the Netherlands and Denmark, for example, unions play an important role in the governance of pension funds, which have large stakes in equities (Jackson and Thelen, 2015; McCarthy et al., 2016). As a result, we would expect stock market development to have a less adverse effect on labor relations in these countries. In absence of a measure of the degree of union involvement in pension funds, country fixed effects help address these differences between countries and give us the effect of within-country changes in stock market development on the various trade union outcomes. In addition to helping address the issue of panel non-stationarity, the ECM is a dynamic model which gives estimates for the effects of changes versus lagged levels in the independent variables (β0*ΔXt and β1*Xt−1, respectively) on changes in the dependent variable (ΔYt). This raises further interesting and substantively important questions about what we would expect the effect of changes versus levels of stock market development and routine task employment on the various dependent variables to be. We should expect to see a negative relationship between changes in stock market development and changes in the institutional structure variables because rigid institutional structures entail reduced flexibility in wages and working conditions. As stock market development increases, the presence of investors increases and the pressure of the demand to reduce labor costs increases. This should result in the decline of centralized wage bargaining institutions, which remove decision-making power over labor from the hands of management, and increased usage of opening clauses, which allow greater scope for local bargaining and flexibility in response to hardship. The expected relationship between the lagged level of stock market development Xt−1 and change in wage bargaining coordination and centralization is not, however, so clear. A negative relationship between high levels of stock market development and bargaining coordination would mean that at high levels of stock market development, there is greater decline in bargaining coordination than at low levels of stock market development. It is unclear that this would be the case. We would expect institutional structures to settle after undergoing change in response to initial changes in stock market development, which would mean that reduction in coordination would be achieved before stock markets reached high levels of development. There will also likely be a limit to how much collective bargaining institutional structures can change. We would not expect a country like Sweden, which had highly centralized bargaining, to converge to a US-style industrial relations system. Furthermore, investors may be willing to accept formally centralized wage setting if it moderates wage outcomes (Wallerstein, 1990). Because of this, we might expect to see a negative relationship between stock market development levels and change in institutional structures—in other words, that institutional structures decline when stock market development is still relatively low, but then stabilize. The relationship between stock market development and union membership should generally be negative. On one hand, if investors are primarily concerned with flexibility, attacks on union membership may not be necessary, as long as collective agreements allow flexibility. But we would expect investors to support many types of measures that would make union membership less attractive to workers. For example, we would expect investors to approve of firms outsourcing non-core jobs when it is possible to do so. If these outsourcing firms are not unionized, this would reduce union density. It would also put downward pressure on wages in unionized firms, as employers would be able to threaten unions with outsourcing. And in countries in which workers vote on union certification for their establishment, we would expect investors to support management efforts to fight this. As with unions’ institutional structures, however, it is not clear that union density decline should be greatest when stock market development is high. We might expect that union density declines are highest when stock market development begins to increase but levels are still relatively low. Unlike with institutional structures, however, the lower bound for union density is much lower and decline could continue at high levels of stock market development. Therefore, the predicted relationship between stock market levels and union density is unclear. How should levels versus changes in routine task employment affect unions’ institutional structures and membership? The effect on membership is relatively clear; as routine task employment was heavily concentrated in industry and workers in industry were heavily unionized, we should see a decline in union density with a decline in routine task employment. Union density decline may also occur through a second mechanism: when routine task employment declines, there is greater competition for lower-wage jobs in the service sector, which allows management to place greater pressure on unions not to unionize. Moreover, we should see variation in the degree of union density decline with levels of routine task employment. When routine task employment is high, we should expect union density to grow. But when it is low, union density should decline. As explained in the previous section, the relationship between routine task employment and unions’ institutional structures is somewhat ambiguous. Unlike union density, which can undergo continuous change, institutional structures like wage bargaining coordination are less likely to undergo gradual change and more likely to rapidly change at critical junctures. When routine task employment is high and union membership is strong, unions should preserve their institutional structures. They should also be able to retain this institutional strength for some time after routine task employment begins to decline. But eventually, unions will come under pressure to weaken collective agreements in order to preserve jobs. As a result, we should expect a decline in routine task employment to be associated with decline in wage bargaining coordination/centralization and with increased usage of opening clauses. It is somewhat less clear that there is a level of routine task employment at which would which expect coordination to decrease and opening clause usage to increase. Unions should be able to avoid institutional structure weakening for a while after routine task employment begins to decline, but will eventually face a trade-off between preserving centralized collective agreements without opening clauses and job security. Nevertheless, this should occur well before routine task employment has declined to its lowest levels and therefore we should expect a decline in coordination/centralization and an increase in opening clause usage at relatively high levels of routine task employment. Table 1 summarizes the predictions for the effect of increases in and levels of stock market development and routine task employment on changes in institutional structure (wage bargaining coordination/centralization), within-institutional-structure strength (opening clauses), and union membership (union density): Table 1. Predicted Effects of Increases and Levels of Stock Market Development and Routine Task Employment   Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable    Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable  Table 1. Predicted Effects of Increases and Levels of Stock Market Development and Routine Task Employment   Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable    Change in union density  Change in wage bargaining coordination/ centralization  Change in opening clauses usage  Stock market development  Increase  Decreasing  Decreasing  Increasing  High level  Unclear  Stable  Stable  Routine task employment  Increase  Increasing  Increasing  Decreasing  High level  Increasing  Increasing/Stable  Decreasing/Stable  6. Results and interpretation 6.1 Main results Due to the number of parameters in the models, I present simplified regression tables including only the main variables of interest in Tables 2–5. The dependent variables in these tables are union density, opening clauses, wage bargaining coordination and wage bargaining centralization, respectively. Full regression results with discussion are in the Supplementary Appendix. Columns 1 and 2 of each table are simple models, with the four main variables without and with country fixed effects. Columns 3–6 add controls, then two types of time fixed effects: for 5-year blocks in column 5 and for year in column 6. The 5-year fixed effects help address common period shocks while year fixed effects help address common year shocks. Table 2. Change in union density Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 2. Change in union density Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.009***  −0.010***  −0.002  −0.002  0.004  0.01***  (−2.81)  (−3.42)  (−0.88)  (−0.68)  (1.20)  (3.63)  ΔRTI  5.75***  6.95***  4.65***  5.92***  6.71***  8.02***  (5.28)  (4.55)  (3.17)  (3.37)  (3.70)  (4.48)  StMktt−1  −0.001  −0.005***  −0.004***  −0.002  0.003  0.004  (−1.03)  (−4.11)  (−3.88)  (−0.84)  (1.55)  (1.51)  RTIt−1  0.81***  6.79***  0.56  5.69***  6.30***  6.26***  (2.91)  (6.93)  (0.91)  (4.56)  (4.28)  (4.33)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.01  0.14  0.11  0.18  0.23  0.30  n  514  514  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 3. Change in opening clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 3. Change in opening clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  0.002***  0.002***  0.002***  0.002**  0.002**  0.001  (2.83)  (3.16)  (2.80)  (2.38)  (2.38)  (1.17)  ΔRTI  −0.12  −0.50  0.22  −0.64  −0.45  −0.49  (−0.31)  (−1.25)  (0.51)  (−1.59)  (−1.09)  (−1.19)  StMktt−1  0.000  −0.000  0.000  0.000  0.000  −0.000  (0.04)  (−0.06)  (1.13)  (0.83)  (−0.71)  (−0.41)  RTIt−1  −0.22***  −1.36***  0.19  −1.26***  −1.05***  −0.86***  (−2.38)  (−6.53)  (1.31)  (−4.03)  (−3.58)  (−3.05)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.03  0.19  0.08  0.23  0.23  0.24  n  503  503  479  479  479  479  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 4. Change in wage bargaining coordination Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 4. Change in wage bargaining coordination Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.004***  −0.004***  −0.003*  −0.004**  −0.003*  −0.005**  (−2.67)  − 3.06)  (−1.75)  (−2.09)  (−1.74)  (−2.00)  ΔRTI  1.12**  0.75  1.00*  −0.14  −0.37  −0.68  (2.27)  (1.47)  (1.70)  (−0.26)  (−0.72)  (− 1.18)  StMktt−1  0.000  −0.001  0.002***  −0.001  0.000  0.001  (0.93)  (−1.38)  (2.72)  (− 0.87)  (0.12)  (0.64)  RTIt−1  0.44***  0.07  1.24***  −1.07***  −1.29***  −1.55***  (4.33)  (0.37)  (6.76)  (−2.97)  (−3.52)  (−4.08)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.08  0.28  0.17  0.38  0.38  0.41  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 5. Change in wage bargaining centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Table 5. Change in wage bargaining centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt  −0.005***  −0.005***  −0.004**  −0.005***  −0.006***  0.008***  (− 3.61)  (−3.13)  (−2.18)  (−3.03)  (− 2.73)  (−3.38)  ΔRTI  1.28**  1.34**  1.15  0.52  0.26  −0.08  (2.27)  (2.41)  (1.44)  (0.98)  (0.51)  (-0.14)  StMktt−1  −0.000  −0.000  0.001  −.003**  −0.002*  −0.002  (−1.07)  (−0.54)  (1.23)  (−2.10)  (− 1.76)  (−1.57)  RTIt−1  −0.02  1.18***  0.68***  −0.26  −0.52  −0.89**  (−0.18)  (3.75)  (4.48)  (−0.65)  (− 1.36)  (−2.41)  Controls  No  No  Yes  Yes  Yes  Yes  Country FE  No  Yes  No  Yes  Yes  Yes  Time FE  No  No  No  No  5-year  Year  R2  0.10  0.42  0.22  0.49  0.49  0.47  n  503  503  488  488  488  488  Note: t-statistics in parentheses. * P < 0.1. ** P < 0.05. *** P < 0.01. Contrary to the predictions, the stock market development variables have a fairly inconsistent relationship with union density. While the signs on both ΔStmkt and Stmktt−1 are mostly negative they are rarely significant. The results for both changes and levels of RTI are consistent with the predictions. They are positive in all of the regressions, meaning that a decline in routine task employment is associated with a decline in union density and that union density growth is higher at high levels of routine task employment. The coefficients are significant across all models for ΔRTI and are significant across all fixed effects models for RTIt−1. The dependent variable is measured as the change in the percentage of the workforce that is unionized and the independent variable is a standardized measure of the routine task intensity of employment, with mean of zero and SD ∼0.14.14 To give some sense of the substantive meaning, Model 4 would predict that for 1 SD increase in routine task employment, roughly the difference between the USA in 2007 and 1986, union density would be 5.92 percentage points higher. Table 3 presents the results for opening clauses. The opening clauses variable is coded 1–6, where 1 represents a complete lack of room for deviation from the conditions of sector-level collective agreements, while 5 indicates extensive presence and use of exceptions and 6 indicates lack of presence of sector-level collective agreements. Changes in stock market development are associated with increased usage of opening clauses across all models. Holding everything else constant, a 50 percentage point increase in stock market development, which is about the difference in Germany from 1980 to 2000, would be associated with a 0.1 unit increase (6-point scale) in opening clauses. This is a small effect, but it is important to remember that several countries, like Canada and the USA, have no variation in the dependent variable because they are always at the maximum value of the variable. The coefficients are significant in all models, except for the model with both country and year fixed effects. This is consistent with the argument that investors and managers under the influence of equity prices will want to increase flexibility in collective agreements. As with union density, there is no consistent relationship between Stmktt−1 and use of opening clauses. The signs on both RTI variables are almost always negative, consistent with the predictions. This suggests that as routine task employment decreases, employers are more able to use opening clauses, perhaps because workers are less able to prevent the internal weakening of institutional structures. The coefficients for ΔRTI are not, however, significant. Unlike routine task employment changes, it was more difficult to predict at which levels of routine task employment we would expect opening clause usage to increase. The negative coefficients suggest that opening clause usage is more likely to increase at lower levels of routine task employment, consistent with the idea that unions are able to avoid the trade-off between collective agreement strength and job protection until routine task employment has already substantially declined. Table 4 presents the results for wage bargaining coordination. The results for stock market development are consistent with those from the opening clauses regressions. The signs on ΔStmkt are all negative and significant while the signs on Stmktt−1 change and are almost entirely insignificant. This suggests that wage bargaining coordination declines quickly in response to increases in stock market development, but that it is not as sensitive to the level of stock market development. To give a sense of the magnitude, a change in German stock market development from the level in 1980 to the level in 2000 would be associated with about a 0.2 unit decrease (5-point scale) in wage bargaining coordination. ΔRTI has little consistent relationship with wage bargaining coordination and the sign on RTIt−1 depends on whether the regressions include fixed effects. Without fixed effects, RTIt−1 has a positive and significant relationship with wage bargaining coordination. But when we include fixed effects, this relationship is always negative. This is consistent with the idea that wage bargaining coordination will begin to decline when routine task employment is still relatively high. Note, however, the difference between these results and the opening clauses results; opening clauses were more likely to increase at lower levels of routine task employment. It is not surprising that there is a difference between the random effects and fixed effects results for RTI because we would expect that relatively constant country-level differences, such as culture, would affect the degree of change in institutional structures. In addition to reasons mentioned above, the negative correlation between RTI levels and change in wage bargaining coordination could be due to reverse causality. Under highly coordinated wage bargaining, it could be more difficult for employers to lay off workers. Oesch and Menes (2011) find that employment growth in low-wage sectors was higher in Spain and the UK, countries with lower wage floors, than in Germany and Switzerland. I further examine the possibility of reverse causality below in Section 6.2. As we can see in Table 5, the results for wage bargaining centralization are similar to those for wage bargaining coordination. Changes in stock market development again have a negative effect, while neither changes in RTI nor levels of stock market development show a consistent relationship. Again, RTIt−1 switches signs from positive in the random effects models to negative in the fixed effects models with full controls, although these coefficients are not significant as they were for wage bargaining coordination. 6.2 Robustness checks As with all observational data work not based on a strong natural experiment, it is not clear that we can reasonably interpret these estimates as causal. A causal interpretation would require at least two things: (a) that unobserved variables, such as countries’ legal institutions, do not explain stock market development or employment composition; (b) that there is not reverse causality, where wage bargaining institutions cause changes in stock market development or the composition of employment. That unobserved variables could explain stock market development or employment composition is a cause for concern. It could be, for example, that a country’s legal institutions or its culture help explain why some countries have higher stock market development or routine task employment than others. While not exhaustive, the inclusion of various types of fixed effects should help address the issue of selection effects as there are stark differences between countries in stock market development, which are likely based on unobserved country-level factors. Country fixed effects should, for example, help address concerns about labor and corporate governance legal institutions, which might explain stock market development, but are relatively constant within countries over time. As we saw in the preceding regressions, the inclusion of country fixed effects matters for the results. This is likely due in large part to the fact that countries have differences dating back to before the period of study. The 5-year and year fixed effects can help deal with common underlying factors that can explain stock market development and decline in routine task employment across these countries. Of these concerns, reverse causality is likely the largest problem for interpreting the coefficients. This is especially the case for the ΔXs, where simultaneity bias is a strong possibility. We would expect investors to shy away from countries with strong, and especially strong centralized union movements, as these reduce management autonomy in firm-level decision-making and would want to keep the ratio of firm revenues divided between workers and investors high. It could also be that decreases in wage bargaining coordination/centralization or introduction of opening clauses would cause stock markets to rise. Nevertheless, countries with very strong unions, such as Sweden, have had these institutions for the entire postwar period and still have many strong, internationally active firms. Investors may be willing to invest in these firms, despite union strength. As long as union actions are predictable, they can be factored in as one among many costs of doing business. Although there is no method to perfectly address this concern, one way which may help address it is to use a different lag structure in the ΔXs. In order to address this, I lagged the ΔXs by 1 year, resulting in the following model:   ΔYt= α0+α1*Yt−1+ β0*ΔXt−1+ β1*ΔXt−1+ εt If the contemporaneous correlations in the ΔXs are due to largely to reverse causality, we should find that the relationship which existed between ΔY and ΔX is either weaker or entirely disappears between ΔY and ΔXt−1. I present four regressions with full controls and the different types of fixed effects for each of the dependent variables in Tables 6,7. The results are similar to the original regressions, although there are a few exceptions. The coefficient for RTI in the union density regressions is now negative, which weakens the support for the argument although the coefficient on RTIt−1 remains positive and consistently significant. The relationship between ΔStmkt and opening clauses disappears, but the negative relationship between RTIt−1 and opening clauses remains. Generally, these results support those in the original regressions that decline in routine task employment reduces union density and that stock market development often has a negative effect, but that it is not highly robust. Table 6. Robustness Check: Change in Union Density, Opening Clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 6. Robustness Check: Change in Union Density, Opening Clauses Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.009***  −.010***  .001  .001  .001  .000  (−3.82)  (−3.71)  (0.20)  (0.30)  (0.37)  (0.18)  ΔRTI t-1  −4.33***  −4.27**  −1.48  −0.93*  −0.64  −0.97  (−2.76)  (−2.53)  (−0.96)  (−1.67)  (−1.12)  (−1.64)  StMktt-1  −.003***  −.003  −.004  .001  .001  −.000  (−3.87)  (−1.35)  (−1.60)  (0.91)  (0.80)  (−0.16)  RTIt-1  7.24***  4.62***  2.20*  −0.91***  −0.98**  −0.75*  (8.32)  (3.66)  (1.76)  (−3.81)  (−2.16)  (−1.83)  Dependent Variable  Union Density  Union Density  Union Density  Opening Clauses  Opening Clauses  Opening Clauses  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .14  .26  .32  .14  .26  .32  N  516  479  479  516  479  479  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 7. Robustness Check: Change in Wage Bargaining Coordination, Centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 7. Robustness Check: Change in Wage Bargaining Coordination, Centralization Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  ΔStMkt t-1  −.004**  −.001*  −.004*  −.005***  −.004**  −.004*  (−2.36)  (−1.80)  (−1.70)  (−3.59)  (−2.50)  (−1.76)  ΔRTI t-1  −1.134**  −1.89***  −2.18***  −1.09**  −0.85*  −0.91*  (−2.16)  (−3.21)  (−3.52)  (−2.06)  (−1.65)  (−1.71)  StMktt-1  .000  −.000  −.000  .001  −.001  −.000  (0.66)  (−0.45)  (0.44)  (1.19)  (−1.08)  (−0.19)  RTIt-1  0.25  0.33  .03  1.45***  0.36  −0.31  (1.41)  (1.03)  (0.10)  (4.41)  (1.11)  (−0.91)  Dependent Variable  Coordination  Coordination  Coordination  Centralization  Centralization  Centralization  Controls  No  Yes  Yes  No  Yes  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  No  No  Yes  No  No  Yes  R2  .26  .31  .35  .42  .43  .44  N  505  488  488  505  488  488  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 8. Robustness Check: Stock Market Development, Routine Task Employment as the Dependent Variable Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 Table 8. Robustness Check: Stock Market Development, Routine Task Employment as the Dependent Variable Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Covariates  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  ΔUnDenst-1  .02        −.09  0.00        0.00  (0.07)        (−0.26)  (0.12)        (0.82)  ΔOpClt-1    −2.88      −1.97    −0.00      0.00    (1.63)      (−1.00)    (−0.12)      (0.53)  ΔCoort-1      −1.26    −1.41      −0.00    0.00      (−0.96)    (−1.11)      (−1.08)    (−0.87)  ΔCentt-1        −0.22  0.24        −0.00  0.00        (0.18)  (0.16)        (−0.46)  (0.36)  UnDenst-1  −0.15        0.24  −0.00        −0.00  (−0.88)        (1.11)  (−0.37)        (−0.83)  OpClt-1    1.40      0.29    −.00      −0.00**    (1.02)      (0.28)    (−1.61)      (-2.09)  Coort-1      0.75    0.81      0.00    0.00      (0.50)    (0.54)      (1.00)    (0.89)  Centt-1        −0.47  0.09        0.00  −0.00        (−0.27)  (0.04)        (0.44)  (−0.77)  Dependent Variable  StMkt  StMkt  StMkt  StMkt  StMkt  RTI  RTI  RTI  RTI  RTI  Controls  No  No  No  No  Yes  No  No  No  No  Yes  Country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Time FE  No  No  No  No  No  No  No  No  No  No  R2  .07  .08  .07  .07  .22  .05  .07  .07  .05  .25  N  675  667  668  667  487  564  553  553  553  465  Note: t-statistics in parentheses. * p<.01 **p<.05 ***p<.01 The results for wage bargaining coordination and centralization in Table 7 are very similar to the original results. The coefficient on stock market development changes remains negative and significant across all specifications. In the previous regressions, RTIt−1 had a negative and significant relationship in many of the fixed effects models. This relationship disappears, but now ΔRTI has a consistently negative coefficient across models, significant in all but one of six models. The substantive conclusion is again fairly similar: the effect of stock market development on union institutions occurs is found in short-term changes, not in levels and the negative relationship is more robust for coordination/centralization than it is for either union density or opening clauses. Contrarily, higher routine task employment has a positive effect on union density and decreases usage of opening clauses, but it also appears to be consistent with lower levels of wage bargaining coordination and centralization. As an additional robustness check for reverse causality I run these same regressions, but with stock market development and routine task employment as dependent variables and the four union variables as predictors. In columns 1–4 and 6–9 of Table 8, I include the change and level parameters for each of the union variables separately without controls but with country fixed effects. Columns 5 and 10 add the control variables. If reverse causality were a concern and strong unions adversely affected stock market development, we would expect to see negative and significant coefficients on union density, bargaining coordination and bargaining centralization and a positive and significant coefficient on opening clauses. As we can see, the various union institutional variables all display inconsistent signs across the models and are seldom significant predictors of either stock market development or routine task employment. This can increase our confidence that the results are not due to reverse causality.15 7. Discussion and conclusion What conclusions can we draw from this analysis about the effect of stock market development and the decline of routine task employment on union strength? One conclusion is that these two variables affect unions in different ways. Perhaps the most surprising result is that decline in routine task employment does not have a negative effect on wage bargaining coordination and centralization. Many of the coefficients on RTI levels are negative and significant, showing that decline of wage bargaining coordination/centralization is stronger at low levels of routine task employment. One interesting possibility is that coordination and centralization are affected by between-firm differences within sectors. If many firms in a sector are similarly affected by technological change, there may not be any pressure to rethink multi-firm bargaining institutions. Stock market development, which has a consistent, negative relationship with wage bargaining coordination and centralization, may differentially affect firms within a sector, increasing differences between firms in whether management is willing to participate in such agreements and causing institutional decline. Future research should investigate the possibility that technological change, stock market development, or other conceptualizations of financialization have differential effects on within-institutional structure and institutional structure change because they have differential effects across firms within a sector. These findings also have important implications for debates on how different dimensions of unions’ strength mediate economic outcomes in Western democracies. While unions have been internally weakened, with both declining union density and collective agreements becoming both less comprehensive in their conditions and more flexible in their application, centralized institutional structures can still matter a great deal for substantive outcomes. While lower-skill workers in manufacturing and service sectors in Germany have borne the brunt of liberalization, higher-skill workers in the manufacturing sector continue to benefit from industry-level collective agreements and strong plant-level representation through works councils (Jackson and Thelen, 2015). In the Nordic countries, industry-level collective agreements usually set only minimum wages, but because these are present in all sectors and unions are still quite powerful, these countries have some of the highest minimum wages in the world, despite the fact that none of them has a statutory minimum wage (Meyer, 2016).16 For strong firms, these agreements are of minimal importance, as workers tend to make above the minimum rates. But in weaker firms, and especially for firms in sectors which are low-wage sectors in other countries (such as fast food or transportation), the sector-level collective agreements guarantee relatively high wages for all. As a result, there is almost no phenomenon of ‘the working poor’ in Sweden and other Scandinavian countries, unlike the USA or Germany (Gautie and Schmitt, 2010). One of the most notable trends across Western political economies has been the increase in economic inequality, which is broadly attributed to the decline of trade unions, among other factors (Scheve and Stasavage, 2009; Western and Rosenfeld, 2011). While recent work has attributed the growth in American inequality to financialization (Lin and Tomaskovic-Devey, 2013), the link between technological change and inequality has been more contested (Kristal and Cohen, 2017; cf. Autor et al., 2008). Regardless of how stock market development and technological change affect inequality directly, they may have an indirect effect on inequality through their effect on trade unions. My results suggest that while these affect unions in different ways, they both contribute to their decline. Future research should further investigate the mediating effect of trade unions on the relationship between financialization, technological change and inequality. This article is only one small contribution toward the broader project of explaining how finance and technological change matter for labor relations. One possibility, which I did not explore, is that the characteristics of the equities holders, such as whether they are pension funds, private equity firms or hedge funds may matter for labor relations. Pension funds tend to be large holders of equities and when unions have a strong say in how these funds invest, they may insist on worker-friendly employment practices. Contrarily, if private equity firms or hedge funds are large equities holders, we might expect an even more aggressive stance toward labor. Gospel et al. (2011) argue that private equity firms have mid-term time horizons and are most likely to take an activist role in corporate governance, while hedge funds are less likely to take an activist role, but have shorter time horizons and have as a primary goal pressuring managers to increase returns to shareholders. Nevertheless, in case studies of select firms in Spain, Germany and the UK, they find little evidence of a substantial change in labor relations after assumption of ownership by private equity funds or increased ownership by hedge funds. Future research using cross-national data could develop a coding of unions’ involvement in pension schemes to determine the degree to which this conditions the effect of stock market development on labor outcomes. It might be possible to study the effect of ownership by private equity firms and hedge funds in linked employer–employee datasets, which are increasingly available across advanced democracies. Linked employer–employee data could also help address questions of whether financialization has contributed to labor market dualism. While there has been a substantial amount of work on how past technological change has affected the employment prospects for different skill groups of workers, the effect of technological change on job tasks continues to change. Until recently, technological change had the greatest adverse effect on workers in middle skill, routine task occupations. But recent research shows that future technological change will adversely affect workers in the lowest-skill occupations. Frey and Osborne (2017) predict that 47% of jobs in the USA, mostly in lower-skills occupations, will be susceptible to automation in the coming decades. This threatens to further undermine cross-skill group workplace solidarity. But it also raises an even more important issue: if the least-educated, lowest-skill individuals become largely unemployable at any acceptable wage, how will we deal with this as a society, particularly when pro-worker institutions, such as trade unions have been greatly weakened? One policy, which has received a great deal of attention is the basic minimum income. It has even received substantial support from libertarian-leaning members of the tech community (Gordon, 2014). But this will be a political challenge as the basic minimum income failed to pass in a 2015 Swiss referendum 77–23%. Moreover, there is reason to be skeptical that most people would be happy to trade the opportunity to work a meaningful job for a basic minimum income. Acknowledgements I would like to give special thanks to Katherine Jackson for her comments on numerous drafts of this article and many discussions on law and finance. I would also like to thank Noam Gidron, Gregory Jackson, Stefan Thewissen and participants at the 2014 EPSA and SASE annual meetings for helpful written and verbal comments. Supplementary material Supplementary material is available at Socio-Economic Review online. Footnotes 1 This is particularly relevant in the Nordic countries, where union membership is voluntary and unions rely entirely on industrial action to pressure employers to sign collective agreements. 2 Katzenstein (1985) argues, however, that trade openness was not incompatible with worker-friendly democratic corporatism in small European states. Democratic corporatism helped reduce class conflict, which enabled more stable production relations and economic growth. 3 I do not address the origins of financialization here. Knafo and Dutta (2016) argue that financialization in the USA was actually driven by managers in the 1960s, who wanted to raise funds to build large conglomerates. 4 Knafo and Dutta (2016) argue that the norm of shareholder value was developed by managers not as a way of prioritizing shareholders, but as a way to value performance in financial markets. They also argue, however, that once the shareholder value norm took over, managers faced pressure (due to the possibility of takeovers) to increase share prices. While their account of how financialization affects management’s behavior is somewhat different from that in this article, managers’ response to stock prices is similar in both accounts. 5 In several countries, pension funds, which are often subject to substantial union influence, constitute a substantial percentage of equities investment. In countries where pension funds are major equities investors and unions have a great deal of influence over these, the pressure on managers to deliver shareholder value at the expense of labor may be substantially reduced. See McCarthy et al. (2016). 6 Jackson and Thelen (2015) find that despite liberalization of finance in Germany, there has been a dual trend in industrial relations. Higher-skill workers in the core manufacturing sectors have retained coordinated labor relations while lower-skill workers in manufacturing and service sectors have been subject to outsourcing and downward wage pressure. 7 This latter mechanism may be less significant in countries with Ghent unemployment schemes, such as Sweden, Denmark and Finland, where the unemployment benefits system is administered by unions and union membership is required for participation in it. 8 According to the authors, a task ‘is routine if it can be accomplished by machines following explicit programmed rules’. This includes ‘many manual tasks … such as monitoring the temperature of a steel finishing line or moving a windshield into place on an assembly line’, but also cognitive tasks, such as ‘calculating, coordinating, and communicating functions of bookkeepers, cashiers, telephone operators, and other handlers of repetitive information-processing tasks’. (1283–1284). 9 These data, however, vary widely in completeness across countries, with some countries, such as Australia, Germany, Canada and the USA having data for almost the entire period and others such as France, The UK and Switzerland having data for relatively few years. 10 The only variables for which panels had missing variables within otherwise complete blocks are RTI and inward migration, thus both of these contain linearly interpolated values. I did not extrapolate either before or after the first/last year observation on the most incomplete variable, resulting in an unbalanced panel. 11 I ran a Wooldridge test for autocorrelation and a Fisher-type test for panel stationarity. The former rejected the null hypothesis of no autocorrelation and the latter rejected the null hypothesis of panel stationarity. 12 The ECM is based on the Autoregressive Distributed Lag model: Yt = α0 + α1Yt−1 + β0Xt + β1Xt−1 + εt. It is generated by subtracting Yt−1 from both sides and adding and subtracting β0Xt from the right-hand side (DeBoef and Keele, 2008). 13 Nickell (1981) demonstrates bias in the fixed effects model in the presence of a lagged dependent variable; however, the bias is of the order 1/T, meaning that it should be fairly minimal in TSCS settings with relatively large T, such as this one. Wilson and Butler (2007) find that the bias is minimal when T > 20 and that fixed effects performs as well as more complicated estimators. 14 To give a better sense of the meaning of 1 SD of routine task intensity, it is approximately the difference in routine task intensity of employment in the USA (1986–2007), Germany (1992–2004) and Denmark (1984–1997). See the Supplementary Appendix for the full descriptive statistics. 15 These null findings for the effect of trade union institutions on occupational employment are similar to those of Oesch (2013) for the effects of the introduction of the minimum wage in UK and labor market deregulation in Germany on occupational employment. We would have expected low-wage employment to grow more in Germany, where there was deregulation than in Great Britain, which increased low-wage regulation. The opposite, however, was the case; low-wage employment grew more in Great Britain than in Germany. 16 Collective agreement coverage is above 90% in Denmark, Finland and Sweden (Visser, 2013). Jackson and Thelen (2015) argue that the predominant ownership structure in Denmark, whereby family foundations hold controlling stakes in firms, has provided patient capital and allowed for solidarity and stability in the industrial relations system. References Acemoglu D., Aghion P., Violante G. ( 2001) ‘Deunionization, Technical Change, and Inequality’, Carnegie-Rochester Conference Series on Public Policy , 55, 229– 264. Google Scholar CrossRef Search ADS   Ahlberg K, Bruun N. ( 2005) ‘Sweden: Transition through Collective Bargaining’. In Blainpain R. (ed) Collective Bargaining and Wages in Comparative Perspective: Germany, France, The Netherlands, Sweden, and the United Kingdom , The Netherlands, Kluwer Law International, pp. 117– 145. Ahlquist J. ( 2010) ‘Building Strategic Capacity: The Political Underpinnings of Coordinated Wage Bargaining’, American Political Science Review , 104, 171– 188. Google Scholar CrossRef Search ADS   Amable B., Ernst E., Palombarini S. ( 2005) ‘How do Financial Markets Affect Industrial Relations: An Institutional Complementarity Approach’, Socio-Economic Review , 3, 311– 330. Google Scholar CrossRef Search ADS   Aspara J., Pajunen K., Tikkanen H., Tainio R. ( 2014) ‘Explaining Corporate Short-termism: Self-Reinforcing Processes and Biases among Investors, the Media and Corporate Managers’, Socio-Economic Review , 12, 667– 693. Google Scholar CrossRef Search ADS   Autor D., Levy F., Murnane R. ( 2003) ‘The Skill-Content of Recent Technological Change: An Empirical Investigation’, Quarterly Journal of Economics , 118, 1279– 1333. Google Scholar CrossRef Search ADS   Autor D., Katz L., Kearney M. ( 2008) ‘Trends in U.S. Wage Inequality: Revising the Revisionists’, Review of Economics and Statistics , 90, 300– 323. Google Scholar CrossRef Search ADS   Baccaro L., Howell C. ( 2011). ‘A Common Neoliberal Trajectory: The Transformation of Industrial Relations in Advanced Capitalism’, Politics and Society , 39, 521– 563. Google Scholar CrossRef Search ADS   Beck T., Demirgüç-Kunt A., Levine R. ( 2000) ‘A New Database on Financial Development and Structure’, World Bank Economic Review , 14, 597– 605. Google Scholar CrossRef Search ADS   Beck N., Katz J. ( 1995) ‘What to Do (and Not to Do) with Time-Series Cross-Section Data’, American Political Science Review , 89, 634– 647. Google Scholar CrossRef Search ADS   Bertrand M., Schoar A., Thesmar D. ( 2007) ‘Banking Deregulation and Industry Structure: Evidence from the French Banking Reforms of 1985’, Journal of Finance , 62, 597– 628. Google Scholar CrossRef Search ADS   Beyer J., Hassel A. ( 2002) ‘The Effects of Convergence: Internationalization and the Changing Distribution of Net Value Added in Large German Firms’, Economy and Society , 31, 309– 332. Google Scholar CrossRef Search ADS   Black B., Gospel H., Pendleton A. ( 2007) ‘Finance, Corporate Governance, and the Employment Relationship’, Industrial Relations , 46, 643– 650. Bond P., Edmans A, Goldstein I. ( 2012) ‘The Real Effects of Financial Markets’, Annual Review of Financial Economics , 4, 339– 360. Google Scholar CrossRef Search ADS   Brady D. ( 2007) ‘Institutional, Economic, or Solidaristic? Assessing Explanations for Unionization across Affluent Democracies’, Work and Occupations , 34, 67– 101. Google Scholar CrossRef Search ADS   Brady D., Huber E., Stephens J. ( 2014) Comparative Welfare States Data Set . University of North Carolina and WZB Berlin Social Science Center. Accessed at http://huberandstephens.web.unc.edu/common-works/data/ on May 15, 2014. Capoccia G., Kelemen R.D. ( 2007) ‘The Study of Critical Junctures: Theory, Narrative, and Counterfactuals in Historical Institutionalism’, World Politics , 59, 341– 369. Google Scholar CrossRef Search ADS   Choi M. ( 2001) Threat Effect of Foreign Direct Investment on Labor Union Wage Premium , PERI Working Paper No. 27. Accessed at SSRN: https://ssrn.com/abstract=335480 on August 5, 2013. Claessens S., Klingebiel D., Schmukler S. ( 2006) ‘Stock Market Development and Internationalization: Do Economic Fundamentals Spur Both Similarly?’ Journal of Empirical Finance , 13, 316– 350. Google Scholar CrossRef Search ADS   Dallery T. ( 2009) ‘Post-Keynesian Theories of the Firm under Financialization’, Review of Radical Political Economics , 41, 492– 515. Google Scholar CrossRef Search ADS   Darcillon T. ( 2015) ‘How Does Finance Affect Labor Market Institutions? An Empirical Analysis in 16 OECD Countries’, Socio-Economic Review , 13, 477– 504. Google Scholar CrossRef Search ADS   DeBoef S., Keele L. ( 2008) ‘Taking Time Seriously’, American Journal of Political Science  52, 184– 200. Google Scholar CrossRef Search ADS   Deeg R., Hardie I. ( 2016) ‘What is Patient Capital and Who Supplies It?’ Socio-Economic Review , 14, 627– 645. Google Scholar CrossRef Search ADS   Dinlersoz E., Greenwood J. ( 2016) ‘The Rise and Fall of Unions in the United States’, Journal of Monetary Economics , 83, 129– 146. Du Caju P., Gautier E., Momferatou D., Ward-Warmedinger M.E. ( 2008) Institutional Features of Wage Bargaining in 23 European Countries, the US and Japan (December 1, 2008). Banque de France Working Paper No. 228. doi:10.2139/ssrn.1677920. Frey C.B., Osborne M. ( 2017) ‘The Future of Employment: How Susceptible are Jobs to Computerization’, Technological Forecasting and Social Change , 114, 254– 280. Google Scholar CrossRef Search ADS   Gautie J., Schmitt J. ( 2010) Low-Wage Work in the Wealthy World , New York, NY, Russell Sage Foundation. Goldin C., Katz L. ( 2008) The Race between Education and Technology , Cambridge, MA, Harvard University Press. Goos M., Manning A., Salomons A. ( 2014) ‘Explaining Job Polarization: Routinization and Offshoring’, American Economic Review , 104, 2509– 2526. Google Scholar CrossRef Search ADS   Gordon N. ( 2014, August 5) ‘The Conservative Case for a Guaranteed Basic Income.’ The Atlantic , accessed at https://www.theatlantic.com/politics/archive/2014/08/why-arent-reformicons-pushing-a-guaranteed-basic-income/375600/ on June 13, 2016. Gospel H., Pendleton A. ( 2003) ‘Finance, Corporate Governance and the Management of Labour: A Conceptual and Comparative Analysis.’ British Journal of Industrial Relations , 41, 557– 582. Google Scholar CrossRef Search ADS   Gospel H., Pendleton A., Vitols S., Wilke P. ( 2011) ‘New Investment Funds, Restructuring, and Labor Outcomes: A European Perspective’, Corporate Governance: An International Review , 91, 276– 289. Google Scholar CrossRef Search ADS   Graham J.R., Harvey C.R., Shiva R. ( 2005). ‘The Economic Implications of Corporate Financial Reporting’, Journal of Accounting and Economics , 40, 3– 73. Google Scholar CrossRef Search ADS   Hall P., Soskice D. (eds) ( 2001) Varieties of Capitalism: The Institutional Foundations of Comparative Advantage , New York, NY, Oxford University Press. Google Scholar CrossRef Search ADS   Hardie I., Howarth D., Maxfield S, Verdun A. ( 2013) ‘ Banks and the False Dichotomy in the Comparative Political Economy of Finance’, World Politics , 65, 691– 728. Google Scholar CrossRef Search ADS   Hirsch B. ( 2008) ‘Sluggish Institutions in a Dynamic World: Can Unions and Industrial Competition Coexist?’ Journal of Economic Perspectives , 22, 153– 176. Google Scholar CrossRef Search ADS   Höpner M. ( 2005) ‘What Connects Industrial Relations with Corporate Governance? Explaining Institutional Complementarity’, Socio-Economic Review , 3, 331– 358. Google Scholar CrossRef Search ADS   Iversen T. ( 1996) ‘Power, Flexibility, and the Breakdown of Centralized Wage Bargaining: Denmark and Sweden in Comparative Perspective’, Comparative Politics , 28, 399– 436. Google Scholar CrossRef Search ADS   Jackson G., Petraki A. ( 2011) Understanding Short-termism: the Role of Corporate Governance , Stockholm, Glasshouse Forum. Jackson G., Thelen K. ( 2015). ‘Stability and Change in CMEs: Corporate Governance and Industrial Relations in Germany and Denmark.’ In Beramendi P., Häusermann S., Kitschelt H., Kriesi H (eds) The Politics of Advanced Capitalism , New York, NY, Cambridge University Press, pp. 305– 329. Google Scholar CrossRef Search ADS   Jaimovich N., Siu H. ( 2014) The Trend is the Cycle: Job Polarization and Jobless Recoveries, accessed at http://www.nirjaimovich.com/assets/jpjr.pdf on April 5, 2015. Jung J. ( 2015) ‘Shareholder Value and Workforce Downsizing 1981-2006’, Social Forces , 93, 1335– 1368. Google Scholar CrossRef Search ADS   Katzenstein P. ( 1985) Small States in World Markets: Industrial Policy in Europe , Ithaca, NY, Cornell University Press. Knafo S., Dutta S.J. ( 2016) ‘Patient Capital in the Age of Financialized Managerialism’, Socio-Economic Review , 14, 771– 788. Google Scholar CrossRef Search ADS   Kristal T., Cohen Y. (2017) ‘The Causes of Rising Wage Inequality: The Race between Institutions and Technology’, Socio-Economic Review , 15, 187– 212. Lazonick W., O'Sullivan M. ( 2000) ‘Maximizing Shareholder Value: A New Ideology for Corporate Governance’, Economy and Society , 29, 13– 35. Google Scholar CrossRef Search ADS   Lee C-S. ( 2005) ‘International Migration, Deindustrialization and Union Decline in 16 Affluent OECD Countries, 1962-1997’, Social Forces , 84, 71– 88. Google Scholar CrossRef Search ADS   Lin K-H., Tomaskovic-Devey D. ( 2013) ‘Financialization and U.S. Income Inequality, 1970-2008’, American Journal of Sociology , 118, 1284– 1329. Google Scholar CrossRef Search ADS   Mahoney J., Thelen K. ( 2010) ‘A Theory of Gradual Institutional Change.’ In Mahoney J., Thelen K. (eds) Explaining Institutional Change: Ambiguity, Agency, and Power , New York, NY, Cambridge University Press, 1– 38. McCarthy M., Sorsa V-P., van der Zwan N ( 2016) ‘Investment Preferences and Patient Capital: Financing, Governance, and Regulation in Pension Fund Capitalism.’ Socio-Economic Review , 14, 751– 769. Google Scholar CrossRef Search ADS   Meyer B ( 2016) ‘Learning to Love the Government: Trade Unions and Late Adoption of the Minimum Wage’, World Politics , 68, 538– 575. Google Scholar CrossRef Search ADS   Nickell S. ( 1981) ‘Biases in Dynamic Models with Fixed Effects’, Econometrica , 49, 1417– 1426. Google Scholar CrossRef Search ADS   OECD ( 2014). OECD.Stat, accessed at stats.oecd.org on April 15, 2014. Oesch D. ( 2013) Occupational Change in Europe: How Technology and Education Transform the Job Structure , Oxford, UK, Oxford University Press. Google Scholar CrossRef Search ADS   Oesch D., Menes J.R. ( 2011). ‘Upgrading or Polarization? Occupational Change in Britain, Germany, Spain and Switzerland, 1990-2008’, Socio-Economic Review , 9, 503– 531. Google Scholar CrossRef Search ADS   Pontusson J., Swenson P. ( 1996) ‘Labor Markets, Production Strategies, and Wage Bargaining Institutions: The Swedish Employer Offensive in Comparative Perspective’, Comparative Political Studies , 29, 223– 250. Google Scholar CrossRef Search ADS   Scheve K., Stasavage D. ( 2009). ‘Institutions, Partisanship, and Inequality in the Long Run’, World Politics , 61, 215– 253. Google Scholar CrossRef Search ADS   Schnabel C. ( 2013) ‘Union Membership and Density: Some (Not So) Stylized Facts and Challenges’, European Journal of Industrial Relations , 19, 255– 272. Google Scholar CrossRef Search ADS   Scruggs L., Lange P. ( 2002) ‘Where Have All the Members Gone? Globalization, Institutions, and Union Density’, Journal of Politics , 64, 126– 153. Google Scholar CrossRef Search ADS   Slaughter M. ( 2007) ‘Globalization and Declining Unionization in the United States.’ Industrial Relations , 46, 329– 346. Spitz-Oener A. ( 2006) ‘Technical Change, Job Tasks, and Rising Educational Demands: Looking Outside the Wage Structure’, Journal of Labor Economics , 24, 235– 270. Google Scholar CrossRef Search ADS   Thelen K. ( 2014) Varieties of Liberalization and the New Politics of Social Solidarity . New York, NY, Cambridge University Press. Google Scholar CrossRef Search ADS   United Nations. ( 1978). 1977 Demographic Yearbook, accessed at https://unstats.un.org/unsd/demographic/products/dyb/dybsets/1977%20DYB.pdf on February 1, 2014. United Nations. ( 1987). 1985 Demographic Yearbook, accessed at https://unstats.un.org/unsd/demographic/products/dyb/dybsets/1977%20DYB.pdf on February 1, 2014. van der Zwan N. ( 2014) ‘Making Sense of Financialization’, Socio-Economic Review , 76, 538– 559. Visser J. ( 2013). Data Base on Institutional Characterictics of Trade Unions , Wage Setting, State Intervention, and Social Pacts, 1960-2012. Amsterdam Institute for Advanced Labor Studies, University of Amsterdam. Version 4.0. Wallerstein M. ( 1990). ‘Centralized Bargaining and Wage Restraint’, American Journal of Political Science , 34, 982– 1004. Google Scholar CrossRef Search ADS   Western B. ( 1997). Between Class and Market: Postwar Unionization in Capitalist Democracies , Princeton, NJ, Princeton University Press. Western B., Rosenfeld J. ( 2011). ‘Unions, Norms, and the Rise in U.S. Wage Inequality’, American Sociological Review , 76, 513– 537. Google Scholar CrossRef Search ADS   Wilson S., Butler D. ( 2007). ‘A Lot More to Do: The Sensitivity of Time-Series Cross-Section Analyses to Simple Alternative Specifications’, Political Analysis , 15, 101– 123. Google Scholar CrossRef Search ADS   © The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

Socio-Economic ReviewOxford University Press

Published: Aug 8, 2017

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