TY - JOUR
AU1 - Born,, Alexandra
AU2 - Enders,, Zeno
AB - Abstract We employ a dynamic stochastic general equilibrium model to investigate the transmission of the global financial crisis via the collapse of export demand (trade channel) and through losses on cross-border asset holdings (financial channel). Calibrated to German data, the model predicts the trade channel to be twice as important as the financial channel. In the United Kingdom, the latter dominates due to higher foreign-asset holdings, which, at the same time, serve as an automatic stabiliser in case of plummeting foreign demand. The financial channel leads to much longer-lasting effects. Stricter enforcement of bank capital requirements would have frontloaded the recession. The financial crisis of 2007–9 started in a small segment of the U.S. financial market and spread rapidly around the world, infecting in particular the large and globalised banking systems of advanced economies. It soon spilled over to the real economy, leading to a global recession. Alternative narratives attribute important roles for its international transmission to the observed collapse in international trade (the trade channel), or to losses in cross-border asset holdings (the financial channel, as defined in this article). It is, however, neither empirically nor theoretically clear which of the two dominated the global transmission of the crisis. As the nature and the importance of the financial channel have changed during the financial globalisation of recent decades, this question is particularly relevant for forming an understanding of international linkages now and in the future. In this article, we therefore investigate the effects of the trade and financial channels in a unified framework and assess which of the two was more important in the transmission of the crisis. For this purpose, we put forward a model that features both channels in order to assess their isolated relevance. Their correlation and mutual interdependence complicate purely empirical analyses. A deeper investigation of the transmission channels by means of a structural model therefore seems necessary and worthwhile. Our international business cycle model features a small open economy integrated with the rest of the world through trade in goods and holdings of international financial assets in the banking sector. The assumption of a small open economy allows us to treat the specific origin of the last crisis as exogenous to the economy in question. To put that differently, we are interested in foreign developments only to the extent that they arrive at the border of the domestic country. Without the need to take a stand on how the financial crisis originated, caused massive losses and led to a collapse in global trade, the applicability of the theoretical results is broadened to generic financial and trade crises. In the model, the transmission through the financial channel works via losses on foreign assets that destroy a part of banks’ capital.1 The bank uses some of its own funds to finance loans and consequently reacts by restricting lending, leading to long-lasting declines in investment and output. Transmission via the trade channel entails that foreign demand for home goods falters, triggering a reduction in exports and output. Calibrating the sizes of the two shocks to the recent financial crisis, we can hence analyse how the economy reacted to each shock and assess which of them had the greater share in the output decline. Furthermore, the model allows us to investigate how policy measures affect the shock transmission. We calibrate the model to German and UK data. We take Germany and the UK as insightful cases on the receiving side of the crisis transmission as they are well integrated with the rest of the world. In particular, trade and financial links between Germany and the UK on one side and the USA on the other side are strong, where Germany is a traditionally strong exporter, while the UK has established London as a global financial centre. They hence lend themselves to investigating the relative importance of both channels. Furthermore, and in line with the model, the financial crisis was arguably exogenous to Germany and the UK, both of which had featured robust growth before the crisis. Germany did not experience a housing or financial asset bubble, while falling house prices in the UK had only a minor autonomous impact on real activity.2 In order to obtain empirical counterparts to the model predictions, we estimate a structural vector autoregression (VAR) featuring international trade and financial data as well as domestic variables from either Germany or the UK. We identify the two shocks that have the highest contribution to the forecast error variance of external demand and financial stress, respectively. We find that shocks to external demand have a significant impact on financial stress and vice versa, which underlines the aforementioned difficulties in disentangling the two channels empirically. We hence look at the crisis impact of both simultaneous shocks, obtaining an approximation for the joint effect of the financial and trade channels. We find that they played a major role in the recessions in Germany and the UK during the financial crisis episode. Comparing the model predictions with our empirical VAR results, we observe that the model can explain half of the empirically estimated maximum GDP decline for Germany, and four-fifths for the UK over a four-quarter horizon following the onset of the crisis.3 Regarding the relative contributions, it turns out that the trade channel is twice as important as the financial channel for the GDP decline in Germany. For the UK, the relative contributions almost reverse, with the financial channel explaining 1.7 times as much as the trade channel. This difference can be explained to a large degree by the higher presence of foreign assets in the UK banking sector, making the British economy more vulnerable to transmission via the financial channel. At the same time, however, the depreciation of the terms of trade that follows a reduction in external demand increases the value of foreign assets in domestic currency, improving banks’ balance sheets. Banks are hence relatively more able to create loans, such that foreign-asset holdings serve as an automatic stabiliser as regards the trade channel. Transmission via the trade channel triggered a relatively short recession in both countries. The financial channel, in contrast, had longer-lasting effects. This channel is thus crucial in accounting for the fact that German output in the last quarter of 2010 was still below its level from two years before, and even more so in the UK. This pattern is in line with general results established by Reinhart and Rogoff (2009) and others, showing that financial crises generally lead to protracted recessions. We additionally explore the effects of a stricter banking regulation in the forms of either higher costs for violating the capital requirement or a higher capital requirement. The former policy turns out to have procyclical effects by frontloading the recession, i.e., the GDP drop is simultaneously deeper and shorter. The same is true for higher capital requirements if banks leave the size of their balance sheets unchanged. Frontloading the recession, however, has a negative impact on the welfare of workers, who form the largest group of agents in the model. If banks react to higher capital requirements by shrinking the balance sheet, the recession is shorter and flatter, with a lower steady-state level of GDP. Within the theoretical literature, our analysis is particularly related to studies analysing the international transmission of financial shocks via a global banking sector. Using a one-good two-country model, Kollmann et al. (2011) show how a banking sector subject to a bank capital requirement can transmit a loan default shock originating in one country. Kollmann (2013) estimates this model on U.S. and euro-area data and finds that a version with a bank capital constraint outperforms a version without it. Attributing a prominent role to the banking sector fits well to the UK and Germany, as firms there rely more heavily on bank lending as in, e.g., the USA. In comparison with Kollmann et al. (2011), we study a two-good model, which enables us to analyse transmission via the trade channel in more detail, in addition to transmission via the banking sector. We furthermore abstract from the foreign economy and instead model a small open economy, thereby avoiding a detailed specification of the origins of the crisis. Lastly, while Kollmann et al. (2011) focus on the effects of loan losses in one of the two countries in the model, we consider losses on foreign assets held by domestic banks. These are much more volatile than loan losses, giving them a larger role in shaping business cycles.4 Related to our research question, several papers in the empirical literature investigate the transmission of the 2007–9 financial crisis, including the trade and financial channels. Abate et al. (2016) study the transmission of U.S. financial shocks to a set of advanced economies, using a factor-augmented VAR. They find that the recent negative shock was large compared to previous financial shocks. While they are not able to clearly disentangle how the financial shock was transmitted via the different channels, they can show that both the trade and financial channels contributed to its transmission. Other studies analyse whether the cross-country variation in the crisis incidence—measured by severity and duration of the output decline as well as business cycle correlations—can be attributed to pre-crisis indicators. Several empirical papers find that advanced economies were hit harder by the crisis (Lane and Milesi-Ferretti, 2011; Rose and Spiegel, 2011) and that financial variables, such as credit growth, are linked to the crisis intensity. The results of Olafsson and Pétursson (2011) show that relatively large banking sectors and strong global financial linkages—together with macro variables such as inflation, current account deficits and a leveraged private sector—played an important role in the propagation of the U.S. shock, whereas there is little evidence for transmission via trade. On the other hand, Rose and Spiegel (2011) identify few consistent results linking pre-crisis indicators and crisis intensity. Furthermore, considering the transmission to financial variables such as credit default swap premia, bank stock prices or equity portfolios, there is little direct evidence that U.S. exposure or external exposure via trade or financial openness led to higher contagion (Bekaert et al., 2014). Given the relatively inconclusive results of the empirical literature, and the need for counterfactual simulations clearly to disentangle the two channels, we think that our analysis through the lens of an appropriate international dynamic stochastic general equilibrium (DSGE) model is worthwhile. 1. The Trade and the Financial Channel Before investigating the transmission of the financial crisis in more detail, we first define the specific channels that we seek to quantify. Taking the financial channel as an example, unforeseen developments may emerge in the foreign financial sector, transmit via international financial linkages and/or spread via the domestic financial sector. Each of these elements can by itself be called a ‘financial channel’. Depending on where the relevant origin (from the perspective of the receiving country) of these developments is located, the distinction between shocks and channels is blurred as well. When investigating the domestic economy, it may suffice to treat all unexpected developments arriving at the border as shocks, although they have their origin in structural shocks hitting foreign economies. Figure 1 illustrates alternative definitions of the trade and financial channels. Our understanding of the transmission in a setting in which two countries are linked via trade in goods and financial assets is the following. In the foreign country, structural shocks affect the trade and/or financial sectors (arrow A), where a single shock can affect both sectors simultaneously. These shocks may change export demand and the valuation of foreign assets of home banks (arrow B), which, in turn, affect the home country via the domestic financial and export sectors (arrow C). These sectors also interact with each other and influence the real economy. Fig. 1. Open in new tabDownload slide Schematic Representation of Specific Transmission Processes. Fig. 1. Open in new tabDownload slide Schematic Representation of Specific Transmission Processes. In this article, we want to measure the effect of foreign developments via a specific financial channel (the valuation of foreign assets of home banks) and foreign export demand (the trade channel), i.e., arrow B. That is, we aim to measure not the effects of individual foreign shocks, but the effects of changes in these two variables. They summarise the effects of foreign developments as they arrive at the border of the receiving country via the two transmission channels. We capture these developments by introducing the corresponding ‘channel shocks’ in our model: a financial-channel shock reduces the valuation of foreign assets of home banks, while a trade-channel shock lowers export demand. Both are exogenous to the economy of the domestic country. As we discuss in more detail in Section 2, these channels are not independent in the data, which makes an empirical investigation of their isolated effects very difficult. We therefore use a structural model to disentangle the two. While arrows A and C deliver valid, alternative definitions of the transmission channels, we think that our approach has the following advantages: (i) it answers our research question: to what extent was the financial crisis transmitted to Germany and the UK via these two narrowly and well-defined channels? Analysing, for example, the transmission via domestic sectors (arrow C) answers the question of the extent to which the financial and export sectors were responsible for transmitting the crisis within the domestic economy. As both variables in B, in addition to further influences, affect both sectors, this is a different question.5 (ii) Changes in the valuation of foreign assets and export demand are easier to measure in the data and hence to calibrate. If we were to model the shocks (arrow A) in the foreign country, which would require additional and potentially controversial assumptions, we would have to make sure that they affect the variables at arrow B as empirically observed. As we aim to isolate the effects of the transmission channels instead of the impact of different foreign structural shocks, we can skip this step and move directly to arrow B. (iii) Our approach has a broad application. Without the need to model foreign structural shocks, the model is applicable to many different situations in which the value of foreign assets and/or export demand change for various reasons. 2. Empirical Responses to the Trade and Financial Channel In this section we identify reactions of the German and UK economies to adverse developments transmitted via the trade and financial channels (arrow B in Figure 1). As laid out in Section 1, these developments can be considered external shocks from the perspective of the receiving countries. However, an empirical identification of shocks that activate a single channel, leaving the other inactive, is inherently complicated since such shocks are unlikely to exist in reality. As we will discuss below, shifts in one channel are indeed accompanied by changes in the other. This is not surprising, given that the reasons that activate one channel, i.e., shocks in foreign economies, are likely to activate the other channel as well. As we aim to investigate each channel in isolation, however, the other channel should ideally remain constant, i.e., inactive. Hence, because of their correlation, we cannot accurately disentangle the transmission channels empirically, which is one motivation behind the development of our theoretical model. Despite these difficulties, we propose an empirical procedure that allows us to gauge the joint impact of the transmission of the crisis via the trade and financial channels on the German and UK economies. We identify ‘crisis shocks’ that are linked to these two channels and estimate their contribution to the recession in both countries. Given the interdependence of the channels, we use a set of relatively unrestrictive identification assumptions. Trade-channel shocks are defined as those that are responsible for the bulk of unexpected changes in external demand: that is, they have the largest contribution to the corresponding one-period ahead forecast error of the below-specified VAR. Unfortunately, we are missing the data equivalent for the financial channel. We will use write-downs of German and UK banks during the crisis as a direct measure of the financial contagion for the calibration of our theoretical model. This variable most closely matches our definition of the financial transmission channel. As no sufficient time series are available for this variable, however, we use the U.S. excess bond premium as provided by Gilchrist and Zakrajšek (2012) in the empirical exercise. This variable has proven to be highly correlated with financial stress, which itself is tightly linked to losses of financial assets by internationally operating banks (e.g., the excess bond premium and write-downs both increased significantly during the financial crisis). We hence estimate separate VARs for Germany and the UK, including the log of domestic GDP, the log of a global demand measure (overall imports of a large country sample, including the respective ten largest trading partners: see Subsection 3.5) and the U.S. excess bond premium (details on the data can be found in the Online Appendix). The VAR takes the form: $$\begin{equation} \bf{\tilde{Y}_t =\sum _{i=1}^L A_i \tilde{Y}_{t-i}+B \nu _t, } \end{equation}$$ where |$\mathbf{ \tilde{Y}_t}$| is a vector containing the mentioned three variables and νt represents i.i.d. shocks with unitary variances. L denotes the number of lags. Based on the Akaike Criterion, we choose three lags. We employ the sample 1991Q1–2012Q4 for Germany and starting after the UK's exit from the European Exchange Rate Mechanism 1993Q1–2012Q4 for the UK. We also include a constant and a linear time trend. To identify the shocks, we employ the strategy that was outlined above. We find the rotation of the impact matrix B that maximises the sum of the contribution of the trade-channel shock to the forecast error variance of our global demand measure plus the contribution of the financial-channel shock to the forecast error variance of the excess bond premium. With this procedure, we identify two shocks that are closely related to international trade and financial developments. The forecast error variance is calculated at horizon 1; i.e., we consider the effects of the shocks on impact.6 Additionally, we impose that the remaining, unidentified shock to domestic GDP (or other domestic variables added below) has no contemporaneous impact on the foreign variables global demand and the U.S. excess bond premium. Analysing the impact of each shock in isolation confirms our conjecture that the channels are not activated individually: external demand moves significantly after a shock to the excess bond premium and vice versa.7 For this reason, we will analyse simultaneous occurrences of both shocks as observed in the financial crisis, yielding a measure of the joint contribution of both channels. In this way we do not need to isolate ‘pure’ individual channel shocks. We will use our theoretical model to do so. In the following, we estimate the VAR once for Germany and a second time for the UK, and feed in both the identified trade- and financial-channel shocks of 2008Q4 and 2009Q1 in the initial period (with no more shocks occurring afterwards). These periods featured the largest shocks in absolute value according to our VAR, such that they come closest to representing the ‘crisis shocks’. We also add other domestic variables of interest to the VAR, one at a time. We display empirical impulse-response functions to these shocks for Germany and the UK in Figures 2 and 3, respectively.8 Shaded areas depict 90% confidence intervals, generated by standard bootstrapping. We observe that the two shocks had a considerable impact on all variables under consideration. Most real variables fell in reaction to the shocks, while the interest rate spread increased. The terms of trade depreciated on impact. Figure 2 shows the initial positive reaction of loans (and deposits) in Germany, potentially due to pre-negotiated credit lines. Net exports fell quickly and significantly in the case of Germany, but only with a delay in the UK. Fig. 2. Open in new tabDownload slide Germany: Impulse Responses to Identified ‘Crisis Shocks’. Notes: Solid lines indicate point estimates, shaded areas 90% confidence bounds obtained by bootstrap sampling (1,000 repetitions). Horizontal axes: years. Vertical axes: percentage points for spread and net exports, percentage deviations from trend otherwise. Fig. 2. Open in new tabDownload slide Germany: Impulse Responses to Identified ‘Crisis Shocks’. Notes: Solid lines indicate point estimates, shaded areas 90% confidence bounds obtained by bootstrap sampling (1,000 repetitions). Horizontal axes: years. Vertical axes: percentage points for spread and net exports, percentage deviations from trend otherwise. Fig. 3. Open in new tabDownload slide UK: Impulse Responses to Identified ‘Crisis Shocks’. Notes: See Figure 2 for description. Fig. 3. Open in new tabDownload slide UK: Impulse Responses to Identified ‘Crisis Shocks’. Notes: See Figure 2 for description. We also note that the trade- and financial-channel shocks were indeed important for the transmission of the global financial crisis. Performing a historical decomposition, we find that all shocks together triggered declines in GDP in 2008Q4 and 2009Q1 (i.e., changes in percentage deviations from trend) of 2.24 and 4.96 percentage points, respectively, for Germany. If we feed only the identified trade- and financial-channel shocks into the VAR, we obtain drops of 1.80 and 3.04 percentage points.9 The corresponding statistics for the UK are GDP declines of 2.55 and 1.94 percentage points in 2008Q4 and 2009Q1, with 1.42 and 1.41 percentage points explained by the two shocks. This shows that our two identified shocks were important, if not the main drivers of economic activity during the financial crisis episode in Germany and the UK. 3. The Model We use a small open economy variant of the model in Kollmann et al. (2011). The economy is inhabited by a representative worker, an entrepreneur and a bank.10 There are two goods: a home intermediate good produced by the entrepreneur and a foreign intermediate good produced in the rest of the world. Both intermediate goods are combined into a final good that is used for consumption by the three agents and for investment by the entrepreneur. The economy is connected to the rest of the world through trade in intermediate goods representing the trade channel and through trade in foreign assets representing the financial channel. 3.1. The Worker The worker's utility depends on consumption of the final good Ct, bank deposits Dt + 1 and hours worked Nt: $$\begin{equation} U_t=E_t\sum ^\infty _{s=0}\beta ^s\left[\frac{\left(C_{t+s}-\psi _wC_{t+s-1}\right)^{1-\sigma _w}-1}{1-\sigma _w}+\Psi ^D\frac{\left(D_{t+1+s}\right)^{1-\sigma _w}-1}{1-\sigma _w}-\Psi ^N N_{t+s}\right], \end{equation}$$ where β is the subjective discount factor, σw > 0 governs the worker's intertemporal elasticity of substitution, and |$\Psi ^D,\, \Psi ^N\gt 0$| are preference parameters. Workers have habits in consumption, where ψw measures the degree of internal habit persistence.11 In addition to paying interest, deposits provide liquidity services to the worker. That way the worker can have the same subjective discount factor as the entrepreneur (and the banker) and still hold positive deposits. The budget constraint of the representative worker in terms of the final good, which is used as the numéraire, is: $$\begin{equation} C_t+p^a_{t}D_{t+1}=p^a_{t}W_tN_t+p^a_{t}D_tR_{t-1}^D. \end{equation}$$ The household earns income from supplying labour to the entrepreneur and from interest payments on deposits held with the bank. The wage rate Wt is measured in terms of the home intermediate good. Thus, labour income in terms of the final good is |$p^a_{t}W_tN_t$|⁠, where |$p^a_t$| is the relative price of the home intermediate good. |$R_{t-1}^D$| is the gross interest rate on deposits made last period, Dt, measured in terms of the home intermediate good as well. The worker either consumes her income or saves in new deposits Dt + 1. 3.2. The Entrepreneur and Final Good Production The entrepreneur produces the home intermediate good at by combining capital and labour provided by the worker via a Cobb-Douglas production function: $$\begin{equation} Y_t=z_t K_t^{\alpha }N_t^{1-\alpha }, \qquad \log (z_t)=\rho _z \log (z_{t-1})+\varepsilon _{z,t}. \end{equation}$$ where α is the capital share and zt is total factor productivity (TFP) following an AR(1) process. The capital stock, owned by the entrepreneur, depreciates with rate δ and increases through gross investment It. The entrepreneur uses the final good for investment. However, it cannot be transformed costlessly into capital. Instead, to produce investment It, the amount ξ(It) of final goods is needed: $$\begin{equation} K_{t+1}=\left(1-\delta \right)K_t+I_t, \qquad \xi (I_t)=I_t+0.5 \Xi \left(\frac{I_t}{I_{t-1}}-1\right)^2, \,\,\Xi \gt 0. \end{equation}$$ To finance parts of her operations, the entrepreneur borrows from the bank one-period loans Lt, on which she has to pay the gross loan rate |$R_{t-1}^L$|⁠. The entrepreneur's budget constraint is: $$\begin{equation} p^a_{t}L_tR_{t-1}^L+\xi (I_t)+p^a_{t}W_tN_t+d^E_t=p^a_{t}L_{t+1}+p^a_{t}Y_t, \end{equation}$$ where |$d_t^E$| is the entrepreneur's dividend income. Her utility is determined by: $$\begin{equation} U_t=E_t\sum ^\infty _{s=0}\beta ^s\left[\frac{\left(d^E_{t+s}-\psi _Ed^E_{t+s-1}\right)^{1-\sigma _E}-1}{1-\sigma _E}\right]. \end{equation}$$ Below we will fix σE to be lower than σw, making the entrepreneur less risk averse than the worker (implying that less risk-averse people are more likely to become entrepreneurs). The subjective discount factor β is the same for all agents. Like the worker, the entrepreneur has habits in consumption, with the parameter ψE. The final good Ft used for consumption and investment is bundled from home and foreign intermediate goods, at and bt, via the following CES-aggregator: $$\begin{equation} F_t=\left(\omega ^{\frac{1}{\theta }}(a_{t})^{\frac{\theta -1}{\theta }}+(1-\omega )^{\frac{1}{\theta }}(b_{t})^{\frac{\theta -1}{\theta }}\right)^{\frac{\theta }{\theta -1}}, \end{equation}$$ where θ is the elasticity of substitution between home and foreign goods and 0 ≤ ω ≤ 1 is the share of the home intermediate good used for the final good in case of equal prices. A cost-minimisation argument yields the demand functions for at and bt with |$p^b_t$| denoting the relative price of foreign intermediate goods in terms of the final good: $$\begin{eqnarray} a_{t}&=\omega \left(p^a_{t}\right)^{-\theta }F_t, \qquad &b_{t}=(1-\omega )\left(p^b_{t}\right)^{-\theta }F_t. \end{eqnarray}$$ 3.3. The Bank The bank collects deposits from the worker, makes loans to the entrepreneur, and trades foreign assets with the rest of the world. Foreign assets At + 1 are measured in terms of foreign intermediate goods.12 The value of foreign assets in terms of the home intermediate good is therefore ptAt + 1, where |$p_t=p^b_t/p^a_t$| are the terms of trade defined as the ratio of import to export prices. |$p^b_t$| is set constant due to our small open economy assumption. As in Kollmann et al. (2011), the bank faces a capital requirement. The capital in period t, Lt + 1 + ptAt + 1 − Dt + 1, should not fall below a fraction γ of the bank's assets Lt + 1 + ptAt + 1. When the bank does not meet the capital requirement, i.e., excess capital xt = (1 − γ)(Lt + 1 + ptAt + 1) − Dt + 1 is negative, it incurs a cost. These costs might be imposed by the regulators or by market discipline and depend in an increasing manner on the amount of capital falling short of the requirement. The cost function ϕ(xt) has the following convex form:13 $$\begin{equation} \phi (x_t)=\phi _1x_t+\frac{\phi _2}{2}(x_t)^2. \end{equation}$$ If capital is below the requirement the bank faces positive costs. Holding more capital than required reduces the chance of falling below the constraint, thereby easing market operations and reducing operation costs by adding ϕ(xt) > 0.14 All bank operations—collecting deposits from workers, handing out loans to entrepreneurs, and holding foreign assets—lead to linear operation costs Γ. The bank's budget constraint is: $$\begin{eqnarray} p_t^a\left(L_{t+1} + D_tR_{t-1}^D+\Gamma\left(D_{t+1}+L_{t+1}+A_{t+1}\right)+\phi(x_t)+\frac{\chi_A}{2}{\left(A_{t+1}-\bar{A}\right)}^2\right)\nonumber\\ +\,p_t^bA_{t+1}+d_t^B=p_t^a\left(L_tR_{t-1}^L+D_{t+1}\right)+p_t^bA_tR_t^AQ_t, \end{eqnarray}$$ where |$d^B_t$| is the banker's dividend income. To induce stationarity we assume that the foreign assets are subject to quadratic portfolio adjustment costs. Specifically, holding foreign assets that are different from their steady-state value |$\bar{A}$| is costly. The foreign asset pays a risky return. We differentiate between two components that determine this return. In normal times, the return is autocorrelated and subject to a shock with a comparable low variance. That is, the expected gross return of foreign assets accumulated in period t − 1, denoted by |$R^A_{t}$|⁠, is exogenous and follows an AR(1) process: $$\begin{equation} \log (R^A_{t})=(1-\rho _R)\log (\bar{R}^A)+\rho _R \log (R^A_{t-1})+\varepsilon _{R,t}. \end{equation}$$ In addition, the stock of foreign assets AtQt with Qt = 1 + εF, t may be subject to an unpredictable i.i.d. shock εF, t, which represents fundamental re-evaluations, such as write-downs. In normal times, εF, t = 0, while a large negative value represents the financial-channel shock, as defined in Section 1. We make this distinction due to the unprecedented nature of the last financial crisis and the high write-downs on loans and assets (as discussed in Subsection 3.5). The results are identical whether we use the valuation shock as the crisis shock or calibrate |$R^A_{t}$| to the size of the valuation shock and assume ρR to be zero during the crisis. The missing autocorrelation corresponds to the singularity of the crisis and allows us to study the impact of the realised, unexpected write-downs. The longer-lasting effects of the financial-channel shock (see Section 4) are hence not mechanically induced by an assumed positive autocorrelation. The banker consumes her dividend income and maximises her utility function $$\begin{equation} U_t=E_t\sum ^\infty _{s=0}\beta ^s\left[\frac{\left(d^B_{t+s}\right)^{1-\sigma _B}-1}{1-\sigma _B}\right], \end{equation}$$ by choosing Lt + 1, At + 1, Dt + 1, subject to the budget constraint (7). 3.4. Market Clearing and Definitions We assume that the costs incurred by the bank are paid in terms of the home intermediate good. The bank has to buy these resources from the entrepreneur. Thus, market clearing for the home intermediate good requires : $$\begin{equation} Y_t=a_{t}+a^*_{t}+\phi (x_t)+\Gamma (D_{t+1}+ L_{t+1}+ A_{t+1})+\frac{\chi _A}{2}\left(A_{t+1}-\bar{A}\right)^2, \end{equation}$$ where |$a^*_{t}$| are exports of the home intermediate good, determined by : $$\begin{equation} a^*_{t}=(1-\omega )\left(p^{*a}_{t}\right)^{-\theta }Y^*_{t}, \qquad \log (Y^*_{t})=\rho _Y \log (Y^*_{t-1})+\varepsilon _{T,t}. \end{equation}$$ Exports depend on the relative price for the home intermediate good in the rest of the world |$p^{*a}_t$|⁠, which is inversely related to the terms of trade pt, and on foreign demand |$Y^*_t$| that follows an AR(1) process, where εT, t represents our trade-channel shock. Net exports scaled by GDP and the interest rate spread are denoted by X and S, respectively, $$\begin{equation} X_t=\frac{a^*_{t}-p_tb_{t}}{Y_t}, \qquad S_t=R^L_t-R^D_t. \end{equation}$$ Finally, market clearing for the final good requires that its production equals aggregate consumption, which is the sum of worker, entrepreneur and banker consumption, plus goods used for investment: $$\begin{equation} F_t=C_t+d^E_t+d^B_t+\xi (I_t). \end{equation}$$ 3.5. Calibration The model is calibrated to match properties of the German and UK economies. Our aim is to make the calibration as specific to each country as possible. That is, we try to capture the different economic structures by setting those parameters to country-specific values for which sufficient data exist. In the following sections, we will then assess the role of crucial differences in economic structures across the two countries for the transmission of the financial crisis. An overview of the calibration exercise is shown in Table 1. Where possible, the sample period runs from 1991Q1 to 2012Q4.15 A detailed description of all data sources and their specific sample can be found in the Online Appendix. Table 1. Calibration. Parameter GE UK Target/Source Trade price elast. θ 1.5 1.5 Backus et al. (1994) Depreciation rate δ 0.025 0.025 Annual δ Portfolio adj. cost χA 0.005 0.005 Davis (2010) IEOS σB = σw 1 1 Log utility Entrepreneur IEOS σE 0.01 0.01 Risk neutrality Utility parameter ψw = ψE 0.85 0.85 Gerali et al. (2010) Convexity of costs ϕ2 0.25/Y 0.25/Y Kollmann et al. (2011) St. st. slope of costs ϕ1 −0.0037 −0.0027 rD = 2.69% (GE), 3.73% (UK) Labour supply ΨN 2.46 3.18 L/Y= 33% (GE), 25% (UK) Preference for deposits ΨD 0.021 0.019 x = 0 Discount factor β 0.986 0.985 rL = 5.6% (GE), 5.91% (UK) St. st. A/L A/L 0.58 1.37 GE/UK data Home bias ω 0.63 0.71 ” Capital share α 0.30 0.35 ” St. st. capital/asset ratio γ 0.0435 0.065 ” Investment adj. costs Ξ 0.0324 0.0221 σI/σY = 2.16 (GE), 3.32 (UK) Operation costs Γ 0.0035 0.0026 Spread = 2.91% (GE), 2.18% (UK) Autocorr. trade |$\rho _{Y^*}$| 0.95 0.96 SUR estimation SD trade 1.53% 1.51% ” Autocorr. TFP ρz 0.76 0.93 ” SD TFP 1.05% 0.64% ” Autocorr. for. return ρR 0.097 0.076 ” SD for. return 4.37% 4.46% ” Corr. TFP and trade Corr(⁠|$\varepsilon _{z},\varepsilon _{Y^{\!*}}\!\!$|⁠) 0.39 0.46 ” Corr. TFP and for. return Corr(εz, εR) 0 0.28 ” Corr. trade and for. return Corr(⁠|$\varepsilon _{Y^{\!*}},\varepsilon _{R}$|⁠) 0.28 0.27 ” Crisis Shocks Financial-channel shock εF −10% −7.25% Foreign write-downs xsTrade-channel shock εT −10.76% −10.35% Foreign demand Autocorr. trade-ch. shock |$\rho _{Y^*}$| 0.53 0.53 yoy ΔY∗ Parameter GE UK Target/Source Trade price elast. θ 1.5 1.5 Backus et al. (1994) Depreciation rate δ 0.025 0.025 Annual δ Portfolio adj. cost χA 0.005 0.005 Davis (2010) IEOS σB = σw 1 1 Log utility Entrepreneur IEOS σE 0.01 0.01 Risk neutrality Utility parameter ψw = ψE 0.85 0.85 Gerali et al. (2010) Convexity of costs ϕ2 0.25/Y 0.25/Y Kollmann et al. (2011) St. st. slope of costs ϕ1 −0.0037 −0.0027 rD = 2.69% (GE), 3.73% (UK) Labour supply ΨN 2.46 3.18 L/Y= 33% (GE), 25% (UK) Preference for deposits ΨD 0.021 0.019 x = 0 Discount factor β 0.986 0.985 rL = 5.6% (GE), 5.91% (UK) St. st. A/L A/L 0.58 1.37 GE/UK data Home bias ω 0.63 0.71 ” Capital share α 0.30 0.35 ” St. st. capital/asset ratio γ 0.0435 0.065 ” Investment adj. costs Ξ 0.0324 0.0221 σI/σY = 2.16 (GE), 3.32 (UK) Operation costs Γ 0.0035 0.0026 Spread = 2.91% (GE), 2.18% (UK) Autocorr. trade |$\rho _{Y^*}$| 0.95 0.96 SUR estimation SD trade 1.53% 1.51% ” Autocorr. TFP ρz 0.76 0.93 ” SD TFP 1.05% 0.64% ” Autocorr. for. return ρR 0.097 0.076 ” SD for. return 4.37% 4.46% ” Corr. TFP and trade Corr(⁠|$\varepsilon _{z},\varepsilon _{Y^{\!*}}\!\!$|⁠) 0.39 0.46 ” Corr. TFP and for. return Corr(εz, εR) 0 0.28 ” Corr. trade and for. return Corr(⁠|$\varepsilon _{Y^{\!*}},\varepsilon _{R}$|⁠) 0.28 0.27 ” Crisis Shocks Financial-channel shock εF −10% −7.25% Foreign write-downs xsTrade-channel shock εT −10.76% −10.35% Foreign demand Autocorr. trade-ch. shock |$\rho _{Y^*}$| 0.53 0.53 yoy ΔY∗ Notes: IEOS stands for Intertemporal Elasticity of Substitution, A/L for the Asset-to-loan ratio, and SUR for Seemingly Unrelated Regression. Open in new tab Table 1. Calibration. Parameter GE UK Target/Source Trade price elast. θ 1.5 1.5 Backus et al. (1994) Depreciation rate δ 0.025 0.025 Annual δ Portfolio adj. cost χA 0.005 0.005 Davis (2010) IEOS σB = σw 1 1 Log utility Entrepreneur IEOS σE 0.01 0.01 Risk neutrality Utility parameter ψw = ψE 0.85 0.85 Gerali et al. (2010) Convexity of costs ϕ2 0.25/Y 0.25/Y Kollmann et al. (2011) St. st. slope of costs ϕ1 −0.0037 −0.0027 rD = 2.69% (GE), 3.73% (UK) Labour supply ΨN 2.46 3.18 L/Y= 33% (GE), 25% (UK) Preference for deposits ΨD 0.021 0.019 x = 0 Discount factor β 0.986 0.985 rL = 5.6% (GE), 5.91% (UK) St. st. A/L A/L 0.58 1.37 GE/UK data Home bias ω 0.63 0.71 ” Capital share α 0.30 0.35 ” St. st. capital/asset ratio γ 0.0435 0.065 ” Investment adj. costs Ξ 0.0324 0.0221 σI/σY = 2.16 (GE), 3.32 (UK) Operation costs Γ 0.0035 0.0026 Spread = 2.91% (GE), 2.18% (UK) Autocorr. trade |$\rho _{Y^*}$| 0.95 0.96 SUR estimation SD trade 1.53% 1.51% ” Autocorr. TFP ρz 0.76 0.93 ” SD TFP 1.05% 0.64% ” Autocorr. for. return ρR 0.097 0.076 ” SD for. return 4.37% 4.46% ” Corr. TFP and trade Corr(⁠|$\varepsilon _{z},\varepsilon _{Y^{\!*}}\!\!$|⁠) 0.39 0.46 ” Corr. TFP and for. return Corr(εz, εR) 0 0.28 ” Corr. trade and for. return Corr(⁠|$\varepsilon _{Y^{\!*}},\varepsilon _{R}$|⁠) 0.28 0.27 ” Crisis Shocks Financial-channel shock εF −10% −7.25% Foreign write-downs xsTrade-channel shock εT −10.76% −10.35% Foreign demand Autocorr. trade-ch. shock |$\rho _{Y^*}$| 0.53 0.53 yoy ΔY∗ Parameter GE UK Target/Source Trade price elast. θ 1.5 1.5 Backus et al. (1994) Depreciation rate δ 0.025 0.025 Annual δ Portfolio adj. cost χA 0.005 0.005 Davis (2010) IEOS σB = σw 1 1 Log utility Entrepreneur IEOS σE 0.01 0.01 Risk neutrality Utility parameter ψw = ψE 0.85 0.85 Gerali et al. (2010) Convexity of costs ϕ2 0.25/Y 0.25/Y Kollmann et al. (2011) St. st. slope of costs ϕ1 −0.0037 −0.0027 rD = 2.69% (GE), 3.73% (UK) Labour supply ΨN 2.46 3.18 L/Y= 33% (GE), 25% (UK) Preference for deposits ΨD 0.021 0.019 x = 0 Discount factor β 0.986 0.985 rL = 5.6% (GE), 5.91% (UK) St. st. A/L A/L 0.58 1.37 GE/UK data Home bias ω 0.63 0.71 ” Capital share α 0.30 0.35 ” St. st. capital/asset ratio γ 0.0435 0.065 ” Investment adj. costs Ξ 0.0324 0.0221 σI/σY = 2.16 (GE), 3.32 (UK) Operation costs Γ 0.0035 0.0026 Spread = 2.91% (GE), 2.18% (UK) Autocorr. trade |$\rho _{Y^*}$| 0.95 0.96 SUR estimation SD trade 1.53% 1.51% ” Autocorr. TFP ρz 0.76 0.93 ” SD TFP 1.05% 0.64% ” Autocorr. for. return ρR 0.097 0.076 ” SD for. return 4.37% 4.46% ” Corr. TFP and trade Corr(⁠|$\varepsilon _{z},\varepsilon _{Y^{\!*}}\!\!$|⁠) 0.39 0.46 ” Corr. TFP and for. return Corr(εz, εR) 0 0.28 ” Corr. trade and for. return Corr(⁠|$\varepsilon _{Y^{\!*}},\varepsilon _{R}$|⁠) 0.28 0.27 ” Crisis Shocks Financial-channel shock εF −10% −7.25% Foreign write-downs xsTrade-channel shock εT −10.76% −10.35% Foreign demand Autocorr. trade-ch. shock |$\rho _{Y^*}$| 0.53 0.53 yoy ΔY∗ Notes: IEOS stands for Intertemporal Elasticity of Substitution, A/L for the Asset-to-loan ratio, and SUR for Seemingly Unrelated Regression. Open in new tab For the first set of parameters in Table 1, we draw on existing studies and set equal values for Germany and the UK. As Kollmann et al. (2011) do, we assume log utility for the worker and the banker and almost risk-neutral entrepreneurs.16 Following Gerali et al. (2010), entrepreneurs and workers have the same consumption habits. For the curvature of the excess cost function ϕ2 we choose a value that is consistent with the empirical relation between loan-to-deposit ratios and interest rate spreads as demonstrated in detail in Kollmann et al. (2011) and supported by micro evidence cited therein. The second set of parameters is chosen to match the averages over the sample period for UK and German data. For example, we set ω equal to the average trade openness, the capital ratio (γ) equal to the ratio of bank equity to total non-risk-weighted bank assets, and the asset-to-loan (A/L) ratio equal to U.S. securities over domestic loans. We use the investment adjustment cost parameter Ξ to match the relative volatility of investment of the model, i.e., the standard deviation of investment relative to the standard deviation of GDP (see Tables 2 and 3). To calculate the relative investment volatility, we simulate the model including all shocks simultaneously. Together with the bank's costs Γ for handling deposits, loans and foreign assets, ϕ1 determines the deposit and loan rates, and hence the spread. These parameters are therefore set to match both average interest rates. The loan interest rate also determines the subjective discount factor, which results from the Euler equation of the entrepreneur.17 The loans to physical capital ratio, which, together with excess capital in steady state being zero, determines ΨN and ΨD, is set such that the ratio of loans to annual GDP in steady state matches the mean of the empirical counterparts for Germany and the UK. Table 2. Business Cycle Statistics of German Data and the Model. Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.56 1.81 1.71 0.04 0.61 1.66 Relative standard deviations Consumption 0.46 0.70 0.58 2.79 1.35 0.56 Investment 2.16 2.16 1.96 9.76 2.82 2.16 Hours 0.43 0.63 0.45 1.49 1.41 0.42 Deposits 0.95 0.81 0.43 7.36 1.84 0.52 Loans 1.53 1.64 1.03 8.89 4.19 0.95 Interest rate spread 0.38 0.12 0.03 0.86 0.38 0.00 Terms of trade 0.76 0.67 0.63 22.42 0.68 0.66 Net exports 0.48 0.35 0.20 7.15 0.74 0.26 Correlation with GDP Consumption 0.42 0.74 0.81 0.51 0.80 0.79 Investment 0.88 0.92 0.96 −0.95 0.75 0.96 Hours 0.68 0.83 0.92 0.95 1.00 0.91 Deposits 0.09 0.10 0.17 −0.82 −0.02 0.21 Loans 0.00 −0.11 −0.13 0.26 −0.16 −0.08 Interest rate spread −0.39 −0.44 −0.96 −0.82 −0.97 −0.65 Terms of trade 0.31 0.56 0.95 0.81 0.33 0.61 Net exports 0.24 0.21 0.24 −0.69 −0.46 0.41 Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.56 1.81 1.71 0.04 0.61 1.66 Relative standard deviations Consumption 0.46 0.70 0.58 2.79 1.35 0.56 Investment 2.16 2.16 1.96 9.76 2.82 2.16 Hours 0.43 0.63 0.45 1.49 1.41 0.42 Deposits 0.95 0.81 0.43 7.36 1.84 0.52 Loans 1.53 1.64 1.03 8.89 4.19 0.95 Interest rate spread 0.38 0.12 0.03 0.86 0.38 0.00 Terms of trade 0.76 0.67 0.63 22.42 0.68 0.66 Net exports 0.48 0.35 0.20 7.15 0.74 0.26 Correlation with GDP Consumption 0.42 0.74 0.81 0.51 0.80 0.79 Investment 0.88 0.92 0.96 −0.95 0.75 0.96 Hours 0.68 0.83 0.92 0.95 1.00 0.91 Deposits 0.09 0.10 0.17 −0.82 −0.02 0.21 Loans 0.00 −0.11 −0.13 0.26 −0.16 −0.08 Interest rate spread −0.39 −0.44 −0.96 −0.82 −0.97 −0.65 Terms of trade 0.31 0.56 0.95 0.81 0.33 0.61 Net exports 0.24 0.21 0.24 −0.69 −0.46 0.41 Notes: SD stands for standard deviation, FA ret. for Foreign asset returns, and No fin. fric. for the specification without financial frictions. Open in new tab Table 2. Business Cycle Statistics of German Data and the Model. Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.56 1.81 1.71 0.04 0.61 1.66 Relative standard deviations Consumption 0.46 0.70 0.58 2.79 1.35 0.56 Investment 2.16 2.16 1.96 9.76 2.82 2.16 Hours 0.43 0.63 0.45 1.49 1.41 0.42 Deposits 0.95 0.81 0.43 7.36 1.84 0.52 Loans 1.53 1.64 1.03 8.89 4.19 0.95 Interest rate spread 0.38 0.12 0.03 0.86 0.38 0.00 Terms of trade 0.76 0.67 0.63 22.42 0.68 0.66 Net exports 0.48 0.35 0.20 7.15 0.74 0.26 Correlation with GDP Consumption 0.42 0.74 0.81 0.51 0.80 0.79 Investment 0.88 0.92 0.96 −0.95 0.75 0.96 Hours 0.68 0.83 0.92 0.95 1.00 0.91 Deposits 0.09 0.10 0.17 −0.82 −0.02 0.21 Loans 0.00 −0.11 −0.13 0.26 −0.16 −0.08 Interest rate spread −0.39 −0.44 −0.96 −0.82 −0.97 −0.65 Terms of trade 0.31 0.56 0.95 0.81 0.33 0.61 Net exports 0.24 0.21 0.24 −0.69 −0.46 0.41 Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.56 1.81 1.71 0.04 0.61 1.66 Relative standard deviations Consumption 0.46 0.70 0.58 2.79 1.35 0.56 Investment 2.16 2.16 1.96 9.76 2.82 2.16 Hours 0.43 0.63 0.45 1.49 1.41 0.42 Deposits 0.95 0.81 0.43 7.36 1.84 0.52 Loans 1.53 1.64 1.03 8.89 4.19 0.95 Interest rate spread 0.38 0.12 0.03 0.86 0.38 0.00 Terms of trade 0.76 0.67 0.63 22.42 0.68 0.66 Net exports 0.48 0.35 0.20 7.15 0.74 0.26 Correlation with GDP Consumption 0.42 0.74 0.81 0.51 0.80 0.79 Investment 0.88 0.92 0.96 −0.95 0.75 0.96 Hours 0.68 0.83 0.92 0.95 1.00 0.91 Deposits 0.09 0.10 0.17 −0.82 −0.02 0.21 Loans 0.00 −0.11 −0.13 0.26 −0.16 −0.08 Interest rate spread −0.39 −0.44 −0.96 −0.82 −0.97 −0.65 Terms of trade 0.31 0.56 0.95 0.81 0.33 0.61 Net exports 0.24 0.21 0.24 −0.69 −0.46 0.41 Notes: SD stands for standard deviation, FA ret. for Foreign asset returns, and No fin. fric. for the specification without financial frictions. Open in new tab Table 3. Business Cycle Statistics of UK Data and the Model. Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.15 2.17 1.27 0.18 1.53 1.22 Relative standard deviations Consumption 0.83 0.92 0.62 1.25 1.11 0.56 Investment 3.32 3.32 2.73 1.80 3.57 3.32 Hours 0.58 1.13 0.53 1.44 1.48 0.49 Deposits 2.05 1.38 0.82 0.86 1.69 0.89 Loans 3.73 3.95 2.38 5.25 4.95 2.30 Interest rate spread 0.46 0.26 0.05 0.43 0.36 0.00 Terms of trade 1.21 0.42 0.46 4.67 0.29 0.57 Net exports 0.87 0.35 0.13 1.21 0.48 0.18 Correlation with GDP Consumption 0.78 0.84 0.82 0.95 0.88 0.72 Investment 0.76 0.91 0.93 −0.03 0.90 0.93 Hours 0.68 0.94 0.96 1.00 0.99 0.96 Deposits −0.18 0.03 0.16 −0.72 −0.03 0.22 Loans 0.42 0.01 0.11 −0.07 −0.03 0.17 Interest rate spread −0.66 −0.84 −0.96 −0.96 −0.97 −0.25 Terms of trade 0.16 0.02 0.94 0.95 −0.22 0.19 Net exports −0.25 −0.72 −0.56 −0.98 −0.91 −0.07 Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.15 2.17 1.27 0.18 1.53 1.22 Relative standard deviations Consumption 0.83 0.92 0.62 1.25 1.11 0.56 Investment 3.32 3.32 2.73 1.80 3.57 3.32 Hours 0.58 1.13 0.53 1.44 1.48 0.49 Deposits 2.05 1.38 0.82 0.86 1.69 0.89 Loans 3.73 3.95 2.38 5.25 4.95 2.30 Interest rate spread 0.46 0.26 0.05 0.43 0.36 0.00 Terms of trade 1.21 0.42 0.46 4.67 0.29 0.57 Net exports 0.87 0.35 0.13 1.21 0.48 0.18 Correlation with GDP Consumption 0.78 0.84 0.82 0.95 0.88 0.72 Investment 0.76 0.91 0.93 −0.03 0.90 0.93 Hours 0.68 0.94 0.96 1.00 0.99 0.96 Deposits −0.18 0.03 0.16 −0.72 −0.03 0.22 Loans 0.42 0.01 0.11 −0.07 −0.03 0.17 Interest rate spread −0.66 −0.84 −0.96 −0.96 −0.97 −0.25 Terms of trade 0.16 0.02 0.94 0.95 −0.22 0.19 Net exports −0.25 −0.72 −0.56 −0.98 −0.91 −0.07 Notes:SD stands for standard deviation, FA ret. for Foreign asset returns, and No fin. fric. for the specification without financial frictions. Open in new tab Table 3. Business Cycle Statistics of UK Data and the Model. Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.15 2.17 1.27 0.18 1.53 1.22 Relative standard deviations Consumption 0.83 0.92 0.62 1.25 1.11 0.56 Investment 3.32 3.32 2.73 1.80 3.57 3.32 Hours 0.58 1.13 0.53 1.44 1.48 0.49 Deposits 2.05 1.38 0.82 0.86 1.69 0.89 Loans 3.73 3.95 2.38 5.25 4.95 2.30 Interest rate spread 0.46 0.26 0.05 0.43 0.36 0.00 Terms of trade 1.21 0.42 0.46 4.67 0.29 0.57 Net exports 0.87 0.35 0.13 1.21 0.48 0.18 Correlation with GDP Consumption 0.78 0.84 0.82 0.95 0.88 0.72 Investment 0.76 0.91 0.93 −0.03 0.90 0.93 Hours 0.68 0.94 0.96 1.00 0.99 0.96 Deposits −0.18 0.03 0.16 −0.72 −0.03 0.22 Loans 0.42 0.01 0.11 −0.07 −0.03 0.17 Interest rate spread −0.66 −0.84 −0.96 −0.96 −0.97 −0.25 Terms of trade 0.16 0.02 0.94 0.95 −0.22 0.19 Net exports −0.25 −0.72 −0.56 −0.98 −0.91 −0.07 Model Data All TFP Trade FA ret. No fin. fric. (1) (2) (3) (4) (5) (6) SD output 1.15 2.17 1.27 0.18 1.53 1.22 Relative standard deviations Consumption 0.83 0.92 0.62 1.25 1.11 0.56 Investment 3.32 3.32 2.73 1.80 3.57 3.32 Hours 0.58 1.13 0.53 1.44 1.48 0.49 Deposits 2.05 1.38 0.82 0.86 1.69 0.89 Loans 3.73 3.95 2.38 5.25 4.95 2.30 Interest rate spread 0.46 0.26 0.05 0.43 0.36 0.00 Terms of trade 1.21 0.42 0.46 4.67 0.29 0.57 Net exports 0.87 0.35 0.13 1.21 0.48 0.18 Correlation with GDP Consumption 0.78 0.84 0.82 0.95 0.88 0.72 Investment 0.76 0.91 0.93 −0.03 0.90 0.93 Hours 0.68 0.94 0.96 1.00 0.99 0.96 Deposits −0.18 0.03 0.16 −0.72 −0.03 0.22 Loans 0.42 0.01 0.11 −0.07 −0.03 0.17 Interest rate spread −0.66 −0.84 −0.96 −0.96 −0.97 −0.25 Terms of trade 0.16 0.02 0.94 0.95 −0.22 0.19 Net exports −0.25 −0.72 −0.56 −0.98 −0.91 −0.07 Notes:SD stands for standard deviation, FA ret. for Foreign asset returns, and No fin. fric. for the specification without financial frictions. Open in new tab The shock processes are calibrated in the following way (see the Online Appendix for details). The foreign demand process is approximated using a series of aggregated trade-weighted total imports of a broad set of trading partners for Germany and the UK, including the respective top ten trading partners. Since the trade weights differ between Germany and the UK, we obtain two slightly different series for Germany and the UK. Because this measure consists of overall imports, it mainly reflects developments in the trading partners, instead of events in Germany or the UK. The AR(1) process for TFP is estimated on linearly detrended German or UK log TFP. For the return process of foreign assets we combine data on stock and corporate debt returns. We also match the empirical correlations between the three data series via a seemingly unrelated regression (SUR) estimation. The correlation between the returns to foreign assets and home TFP shocks for Germany is not significantly different from zero, while for the UK it is 0.28. Regarding the crisis shocks, we choose the magnitudes of the shocks to match the observed declines in trade and bank asset values. Based on our measure of external demand, the trade-channel shock is set to −10.8% for Germany and −10.4% for the UK. The autocorrelation is adjusted to 0.53 for both countries to capture the relatively shorter length of depressed demand, compared with normal times. The financial-channel shock is based on estimates of write-downs by the International Monetary Fund (2010b), which imply a cumulative loss rate of εF, t =−10% for foreign loans and securities of German banks and of 7.25% for UK banks between 2007 and 2010.18 4. Theoretical Predictions 4.1. Business Cycle Statistics Before we analyse with our model how the financial crisis was transmitted, we first examine whether it is able to capture features of normal German and UK business cycles. For this purpose, we compare second moments of HP-filtered German and UK data for the period 1991Q1–2012Q4 with unconditional HP-filtered moments of the model. For all simulations, we linearise the model around the steady state.19 For the three shock processes for TFP, trade and foreign-asset returns, we use the fitted AR(1) processes as described in the previous section. The valuation shock does not play a role under normal circumstances. Hence, we include only the shock process for ‘normal’ foreign asset returns. Tables 2 and 3 present moments of the data (Column 1), of the model with all shocks (Column 2), and of the model including TFP trade, and foreign-asset return shocks individually (Columns 3–5). The model is able to replicate many features of German and UK business cycles. GDP volatility generated by the model including all shocks is somewhat higher for the UK calibration than the one in the data. However, the volatilities of the other variables relative to GDP and their correlations with GDP are matched well. As in the data, aggregate consumption is less volatile than GDP. Loans show a higher relative volatility than deposits in the model, similar to the empirical observations. Deposits and loans are a lot more volatile in the UK than in the German data. The model successfully replicates this difference. While it is not able to generate the negative correlation between deposits and GDP for the UK and loan acyclicality in Germany, the signs for all remaining variables are matched correctly for both countries. The terms of trade—typically difficult to match for international business cycle models (see, e.g., Enders and Müller, 2009)—are also predicted to be procyclical in both countries. The most striking difference between Germany and the UK in terms of correlations is the opposite sign for the net-export-to-GDP correlation, which is correctly generated by the model. Considering the three shocks individually shows that the model including only the TFP shock underpredicts the relative volatilities of the financial variables and net exports. It also generates too large a correlation between deposits and GDP. This underlines the importance of including the trade- and financial-channel shocks to account for business cycle moments. The trade-channel shock contributes relatively little to output fluctuations. However, it is not only the main driver of relative net exports and the terms of trade but also has a large impact on investment and deposits (in Germany) and loans (in both countries). On the other hand, its predictions for the correlations of deposits and loans with GDP are quite far away from the empirical values. The return shock brings these correlations closer towards the data. It is furthermore responsible for a large part of the fluctuations in loans and deposits, and generates volatilities of the spread that are in line with the data. We also demonstrate the importance of the financial friction in this model. The introduction of a bank with a capital requirement introduces a distortionary wedge between the lending and the deposit rate. In the steady state, the effects of this wedge are well understood: the marginal product of capital does not correspond to the marginal rate of substitution, thereby lowering consumption and output. The effects of a time-varying spread, however, are less straightforward. To analyse those, we conduct simulations in which we leave the steady-state distortion as in the baseline calibration, but set ϕ2 = 0.0001.20 Up to a first-order approximation, the bank has no incentive to return to the required bank capital. Hence, the spread does not depend on the level of bank capital in the linearised model, eliminating the effect of the financial friction on the model dynamics. Column (6) in Tables 2 and 3 shows the results. They are close to those obtained if only TFP shocks occur. We conclude that the financial friction is important in generating realistic business cycle fluctuations in our model.21 One main difference between Germany and the UK is the structure of banks’ balance sheets. While German banks are traditionally very engaged in providing loans to domestic entrepreneurs (with a loan-to-GDP ratio of 33% compared with 25% in the UK), banks in the UK invested much more in foreign financial assets.22 The resulting difference in the A/L ratio is responsible for a large part of the differences between the simulated German and UK economies. We analyse the importance of financial frictions and structures for the transmission of the crisis in Section 5.23 4.2. Crisis Transmission In this section, we assess whether the model economy can replicate characteristics of the German and UK downturns during the financial crisis. In particular, we are interested in how the theoretical prediction for the effects of the trade and financial channel, as defined in Section 1 and formalised in Section 3, relates to the empirical estimate of their combined effect in Section 2. Remember that, for the reasons explained there, it is inherently difficult to disentangle the two shocks empirically. Once we are confident that the model is able to replicate key aspects of the crisis episode, we will investigate both transmission channels in isolation. Figure 4 displays the impulse responses to a negative trade- and a simultaneous negative financial-channel shock for Germany (solid line) and the UK (dashed line), where the magnitudes of the shocks are as discussed in Subsection 3.5. The model predicts reductions in output, investment, exports, banks’ foreign assets, bank capital, loans and an increase in the loan-deposit interest rate spread, all in line with our empirical observations in Section 2.24 Additionally, consumption, deposits and hours worked are predicted to fall, while the terms of trade depreciate. The main differences between Germany and the UK occur in the responses of foreign assets, loans, deposits, the spread and investment. The initial impact on GDP following the simultaneous shocks is slightly higher for Germany, while the UK suffers from a more sluggish recovery. We will provide interpretation of these results when discussing the reactions to the individual shocks in Section 5. Fig. 4. Open in new tabDownload slide Impulse Responses to Simultaneous Trade- and Financial-Channel Shocks. Notes: Solid lines represent impulse response functions for Germany, dashed lines for the UK, dotted lines for German calibration with shocks and A/L ratio of UK. Variables are expressed in percentage deviations from steady state, except spread (in percentage points p.a.), capital-asset ratio and net exports over GDP (both in percentage points). Fig. 4. Open in new tabDownload slide Impulse Responses to Simultaneous Trade- and Financial-Channel Shocks. Notes: Solid lines represent impulse response functions for Germany, dashed lines for the UK, dotted lines for German calibration with shocks and A/L ratio of UK. Variables are expressed in percentage deviations from steady state, except spread (in percentage points p.a.), capital-asset ratio and net exports over GDP (both in percentage points). In order to systematically compare the model predictions with the data, we use the estimated reactions in both countries to the ‘crisis shocks’, as identified in the VAR of Section 2. Table 4 compares the maximum responses of the empirical impulse-response functions in Figure 2 with those predicted by the model (Figure 4) for the German calibration, while Table 5 presents results for the UK calibration and the empirical counterparts of Figure 3. Column (1) shows the maximum responses of the empirical reactions and Column (2) the model predictions after the occurrence of both shocks. The maximum responses are calculated in the four quarters following the shocks in both cases. The model is able to replicate 51% of the estimated output decline following the crisis shocks for Germany. This might be an indication that our empirical estimates are correlated with further negative shocks, such as uncertainty shocks, and/or transmission channels that are not the subject of this article. Nevertheless, the model predicts the sign and value of most other variables fairly well, as discussed below. For the UK, the predictions of the model are closer to the empirical estimates, covering 79% of the estimated maximum output decline after the crisis shocks. Table 4. Crisis Responses Germany. Maximum response Cumulative response Data Both TC FC Both TC FC Output −4.79 −2.42 −1.79 −0.87 −5.37 −1.38 −3.99 Consumption −1.10 −1.29 0.26 −1.35 −4.39 0.89 −5.27 Investment −5.57 −1.60 1.75 −2.57 −5.63 5.77 −11.40 Hours −2.25 −3.46 −2.56 −1.17 −7.59 −2.13 −5.46 Deposits −0.10 −3.70 −2.17 −1.59 −13.33 −9.04 −4.29 Loans −1.01 −2.01 1.49 −3.50 −3.19 6.23 −9.42 Interest rate spread 0.94 0.29 −0.08 0.37 1.40 −0.37 1.77 Terms of trade 1.14 1.88 1.97 0.10 6.31 6.18 0.14 Net exports −1.49 −2.78 −3.05 0.58 −2.94 −5.35 2.41 Maximum response Cumulative response Data Both TC FC Both TC FC Output −4.79 −2.42 −1.79 −0.87 −5.37 −1.38 −3.99 Consumption −1.10 −1.29 0.26 −1.35 −4.39 0.89 −5.27 Investment −5.57 −1.60 1.75 −2.57 −5.63 5.77 −11.40 Hours −2.25 −3.46 −2.56 −1.17 −7.59 −2.13 −5.46 Deposits −0.10 −3.70 −2.17 −1.59 −13.33 −9.04 −4.29 Loans −1.01 −2.01 1.49 −3.50 −3.19 6.23 −9.42 Interest rate spread 0.94 0.29 −0.08 0.37 1.40 −0.37 1.77 Terms of trade 1.14 1.88 1.97 0.10 6.31 6.18 0.14 Net exports −1.49 −2.78 −3.05 0.58 −2.94 −5.35 2.41 Notes: The first column displays maximum responses in the four periods following the shock as estimated in Section 2. Other columns show results from simulation of the model, with either both shocks (second and fifth columns), only the trade-channel shock (third and sixth columns) or only the financial-channel shock (fourth and seventh columns) operating. Open in new tab Table 4. Crisis Responses Germany. Maximum response Cumulative response Data Both TC FC Both TC FC Output −4.79 −2.42 −1.79 −0.87 −5.37 −1.38 −3.99 Consumption −1.10 −1.29 0.26 −1.35 −4.39 0.89 −5.27 Investment −5.57 −1.60 1.75 −2.57 −5.63 5.77 −11.40 Hours −2.25 −3.46 −2.56 −1.17 −7.59 −2.13 −5.46 Deposits −0.10 −3.70 −2.17 −1.59 −13.33 −9.04 −4.29 Loans −1.01 −2.01 1.49 −3.50 −3.19 6.23 −9.42 Interest rate spread 0.94 0.29 −0.08 0.37 1.40 −0.37 1.77 Terms of trade 1.14 1.88 1.97 0.10 6.31 6.18 0.14 Net exports −1.49 −2.78 −3.05 0.58 −2.94 −5.35 2.41 Maximum response Cumulative response Data Both TC FC Both TC FC Output −4.79 −2.42 −1.79 −0.87 −5.37 −1.38 −3.99 Consumption −1.10 −1.29 0.26 −1.35 −4.39 0.89 −5.27 Investment −5.57 −1.60 1.75 −2.57 −5.63 5.77 −11.40 Hours −2.25 −3.46 −2.56 −1.17 −7.59 −2.13 −5.46 Deposits −0.10 −3.70 −2.17 −1.59 −13.33 −9.04 −4.29 Loans −1.01 −2.01 1.49 −3.50 −3.19 6.23 −9.42 Interest rate spread 0.94 0.29 −0.08 0.37 1.40 −0.37 1.77 Terms of trade 1.14 1.88 1.97 0.10 6.31 6.18 0.14 Net exports −1.49 −2.78 −3.05 0.58 −2.94 −5.35 2.41 Notes: The first column displays maximum responses in the four periods following the shock as estimated in Section 2. Other columns show results from simulation of the model, with either both shocks (second and fifth columns), only the trade-channel shock (third and sixth columns) or only the financial-channel shock (fourth and seventh columns) operating. Open in new tab Table 5. Crisis Responses UK. Maximum response Cumulative response Data Both TC FC Both TC FC Output −2.84 −2.23 −0.97 −1.62 −7.28 0.30 −7.58 Consumption −1.87 −1.75 0.39 −2.00 −6.18 1.65 −7.83 Investment −6.07 −3.57 3.00 −5.64 −14.92 9.01 −23.93 Hours −2.07 −3.42 −1.50 −2.36 −10.86 0.14 −11.00 Deposits −3.67 −3.69 −1.13 −2.72 −11.98 −4.60 −7.39 Loans −5.67 −5.03 3.25 −8.28 −10.80 12.24 −23.04 Interest rate spread 0.75 0.49 −0.15 0.65 2.27 −0.71 2.97 Terms of trade 0.61 1.64 1.67 0.25 6.14 5.26 0.88 Net exports −0.33 −1.93 −2.39 0.80 −0.90 −4.44 3.54 Maximum response Cumulative response Data Both TC FC Both TC FC Output −2.84 −2.23 −0.97 −1.62 −7.28 0.30 −7.58 Consumption −1.87 −1.75 0.39 −2.00 −6.18 1.65 −7.83 Investment −6.07 −3.57 3.00 −5.64 −14.92 9.01 −23.93 Hours −2.07 −3.42 −1.50 −2.36 −10.86 0.14 −11.00 Deposits −3.67 −3.69 −1.13 −2.72 −11.98 −4.60 −7.39 Loans −5.67 −5.03 3.25 −8.28 −10.80 12.24 −23.04 Interest rate spread 0.75 0.49 −0.15 0.65 2.27 −0.71 2.97 Terms of trade 0.61 1.64 1.67 0.25 6.14 5.26 0.88 Net exports −0.33 −1.93 −2.39 0.80 −0.90 −4.44 3.54 Notes: See Table 4 for description. Open in new tab Table 5. Crisis Responses UK. Maximum response Cumulative response Data Both TC FC Both TC FC Output −2.84 −2.23 −0.97 −1.62 −7.28 0.30 −7.58 Consumption −1.87 −1.75 0.39 −2.00 −6.18 1.65 −7.83 Investment −6.07 −3.57 3.00 −5.64 −14.92 9.01 −23.93 Hours −2.07 −3.42 −1.50 −2.36 −10.86 0.14 −11.00 Deposits −3.67 −3.69 −1.13 −2.72 −11.98 −4.60 −7.39 Loans −5.67 −5.03 3.25 −8.28 −10.80 12.24 −23.04 Interest rate spread 0.75 0.49 −0.15 0.65 2.27 −0.71 2.97 Terms of trade 0.61 1.64 1.67 0.25 6.14 5.26 0.88 Net exports −0.33 −1.93 −2.39 0.80 −0.90 −4.44 3.54 Maximum response Cumulative response Data Both TC FC Both TC FC Output −2.84 −2.23 −0.97 −1.62 −7.28 0.30 −7.58 Consumption −1.87 −1.75 0.39 −2.00 −6.18 1.65 −7.83 Investment −6.07 −3.57 3.00 −5.64 −14.92 9.01 −23.93 Hours −2.07 −3.42 −1.50 −2.36 −10.86 0.14 −11.00 Deposits −3.67 −3.69 −1.13 −2.72 −11.98 −4.60 −7.39 Loans −5.67 −5.03 3.25 −8.28 −10.80 12.24 −23.04 Interest rate spread 0.75 0.49 −0.15 0.65 2.27 −0.71 2.97 Terms of trade 0.61 1.64 1.67 0.25 6.14 5.26 0.88 Net exports −0.33 −1.93 −2.39 0.80 −0.90 −4.44 3.54 Notes: See Table 4 for description. Open in new tab Importantly, Tables 4 and 5  also display the reactions triggered by each channel individually. For Germany, the trade-channel shock in the model (Column 3) explains around 37% of the estimated drop of output in the VAR, while for the UK the number is 34%. The isolated financial-channel shock (Column 4) triggers reductions in output of 18% and 57% of the estimated output drop for Germany and the UK, respectively. The fraction of the total output decline predicted by the model that can be attributed to each channel is hence almost opposite for both countries. While in Germany the trade-channel shock alone generates an output drop of 74% of the decline generated by both channels together (resulting from −1.79%/−2.42%), the financial-channel shock triggers 36% of this response.25 For the UK, the proportions are 43% for the trade channel and 73% for the financial channel. In general, the trade channel is the driving force behind the developments in the external sector of both countries. The financial-channel shock, on the other hand, exerts strong negative pressure on consumption, investment, deposits, hours worked and loans. It also pushes up the spread in the crisis. The depreciation of the terms of trade is predicted relatively well for Germany, mainly attributable to the trade-channel shock. German net exports are predicted to fall substantially, although more than in the data. Because of Germany's larger trade dependency, hours worked and deposits are also influenced strongly by the trade channel. The financial-channel shock plays a smaller role for the output drop in Germany compared with that in the UK, but is essential to bring about a decline in loans and an increase in the spread. It also has a much more detrimental effect on investment than the trade-channel shock, although the model strongly underpredicts the investment slump. Arguably, risen uncertainty after the crisis, which might have been correlated with the identified empirical shocks but is not present in the model, might have contributed to this extreme response in the data. The drop in deposits is overpredicted, but note that the empirical response reaches lower levels after the initial four periods. For the UK, the drop in output is relatively similar to the empirical counterpart, which is mostly due to the financial channel. The fall in investment is again underpredicted, but less so than in the German case. Deposits drop similarly in the model and in the data. Loans, as in the data, are predicted to fall more than in Germany. The trade-channel shock is again the main driver of the terms of trade and the trade balance, whose response is, despite being too large, correctly predicted to be less strong compared with that in Germany. To summarise, the model does well in predicting the reactions of important variables in the crisis, with some exceptions. In particular, the model responses show a decline in macroeconomic activity, with output, consumption, investment and hours decreasing following the shocks. Similarly, deposits, loans and net exports fall, while the interest rate spread rises. The model hence predicts the correct signs for all of these variables if compared with our estimates from Section 2. To account for the impact of the financial crisis on Germany and the UK, it turns out to be crucial to consider the simultaneous occurrence of the trade- and financial-channel shocks. The simulation also highlights the role of each channel for specific variables, particularly the dominance of the trade channel for Germany and of the financial channel for the UK. 5. Inspecting the Mechanism To gain more insights into the dynamic responses triggered by the two shocks, we display the impulse-response functions for the financial- and the trade-channel shock separately in Figures 5 and 6 and provide interpretation. The responses in Germany are again depicted by blue solid lines, while red dashed lines represent UK responses. The shocks have the same magnitudes as before. Both lead to a decline in output and hours, where the financial-channel shock triggers hump-shaped responses. The reactions of other variables, in particular bank capital and exports, differ more starkly across shocks. As evident from Tables 4 and 5, we find that the developments in Germany and the UK can be explained only by a combination of the trade- and the financial-channel shock. Both shocks would predict the wrong sign for key variables if they occurred in isolation. Fig. 5. Open in new tabDownload slide Impulse Responses to Financial-Channel Shock. Notes: See Figure 4 for description. Fig. 5. Open in new tabDownload slide Impulse Responses to Financial-Channel Shock. Notes: See Figure 4 for description. Fig. 6. Open in new tabDownload slide Impulse Responses to Trade-Channel Shock. Notes: See Figure 4 for description. Fig. 6. Open in new tabDownload slide Impulse Responses to Trade-Channel Shock. Notes: See Figure 4 for description. In the case of the financial-channel shock, the deteriorating value of foreign assets leads to a decline in bank capital. Interestingly, the relatively lower write-downs on UK foreign asset holdings, resulting from the better portfolio choice of UK banks relative to their German counterparts, translate into a larger drop of the capital-asset ratio because of the much larger ratio of foreign assets to loans (and to GDP). Correspondingly, the increase in the spread and the output drop are more pronounced in the UK. Specifically, the shock raises the loan-deposit rate spread by about 0.4 percentage points for the German model and by about 0.7 percentage points for the UK, which translates into a fall of loans and deposits. The losses and the larger increase in the spread reduces domestic demand—that is, aggregate consumption and investment—relatively more in the UK, lowering prices of domestically produced goods. The resulting depreciated terms of trade increase exports slightly, by less than 0.5% over four years, while imports fall. Germany, on the other hand, experiences an even smaller depreciation of the terms of trade, as its higher openness implies that a larger part of the reduction in demand falls on foreign goods. We hence obtain a relatively subdued expansion of exports. In contrast, bank capital increases in the case of the trade-channel shock, putting downward pressure on the spread. This is driven by a positive response of loans and a reduction of deposits during the first few quarters. Workers reduce deposits to smooth out the impact of the shock on consumption, while loans increase due to a lower loan rate. The latter results from an expected appreciation of the terms of trade, i.e., back towards the steady state after the initial depreciation that follows the reduction in demand. Expected returns on foreign assets in domestic currency hence decrease and by arbitrage domestic loan rates fall as well. Foreign assets in domestic currency (plotted) decline despite the depreciation, as they become less attractive. Instead, the bank shifts its assets towards loans until expected returns are equalised. The initial slight increase of aggregate consumption following a trade-channel shock is mainly driven by consumption of entrepreneurs because of intertemporal consumption shifting, triggered by the reduced loan rate.26 Since they are almost risk neutral, entrepreneurs react strongly to interest-rate movements. Bankers’ consumption also increases because of rising income from lending activities and the increasing domestic value of foreign assets, but its share in aggregate consumption is very low. On the other hand, consumption of workers, which constitutes the largest part of aggregate consumption, falls. Lower interest rates and higher net exports, which turn into a surplus because of the continued depreciation of the terms of trade, lead to a quick and overshooting recovery of GDP. The main difference between the German and the UK reaction to the trade-channel shock is again due to the larger foreign-assets-to-loans ratio in the UK. The depreciation of the terms of trade increases the value of UK foreign assets in terms of domestic goods relatively more, such that bank capital rises more than in Germany.27 The resulting stronger reduction in the spread stimulates the economy more in the UK. The large exposure to foreign assets, which leads to detrimental effects in case of the financial-channel shock, hence serves as an automatic stabiliser after a trade-channel shock. The trade-channel shock explains almost all the movements of net exports and the terms of trade, while the financial-channel shock is responsible for most of the reactions of consumption and the financial variables.28 Deposits in Germany and foreign assets in both countries, however, are also driven to a non-negligible amount by the trade-channel shock. Concerning GDP, the trade-channel shock's relative influence is largest on impact, especially in Germany. Longer horizons are dominated by the financial-channel shock. Given that German—and even more so UK—output in the third quarter of 2010 was still below its level from two years before, the financial-channel shock plays an important role in explaining the prolonged recovery. This is also reflected in the cumulative responses, accumulating the quarterly values of individual responses over the course of the year following the shocks, presented in Columns (5)–(7) of Tables 4 and 5. Next, we want to highlight the importance of different structures in the UK and Germany for the transmission of the shocks, and in particular the composition of banks’ balance sheets. To do so, we conduct the following counterfactual simulation. We take the model calibrated to the German economy, but change the A/L ratio to the higher value observed in the UK. As crisis shocks, we also use those of the UK. We hence obtain a hypothetical crisis path for a German economy with parts of the financial structure of the UK and the crisis shocks as experienced in the UK. Results are plotted as green dotted lines in Figures 4–6. Any remaining differences between this scenario and the UK responses (red dashed lines) hence stem from other differences in the calibration. Consider first the financial channel, shown in Figure 5. Remember that the UK experienced lower write-downs than Germany during the financial crisis (in percentage terms, but larger in absolute terms). If we equalise the A/L ratio and the size of the financial-channel shock, foreign assets fall similarly in the hypothetical German and the actual UK economy. Because of the higher A/L ratio, most hypothetical German responses are thus much closer to their UK counterparts, compared with the baseline.29 Using the UK value for the A/L ratio hence amplifies most responses relative to the German calibration with German shocks, despite the lower reduction in foreign assets. Intuitively, as the shock is applied to a larger stock of foreign assets, bank capital falls by more than in the original German calibration. This triggers a larger increase in the spread and hence stronger reductions in loans and deposits. The largest remaining differences to the UK calibration are in the external sector. Because Germany is more engaged in international trade, the reduction in demand falls to a larger extent on foreign goods. As a result of the smaller weight on domestic goods, net exports increase more, the terms of trade depreciate less, and exports are lower compared with those in the UK. Adjusting the export share of Germany to the UK value results in more similar responses of exports, net exports and the terms of trade.30 As for the UK responses, we also obtain a quick depreciation and a simultaneous increase in exports in this case. The green dotted lines in Figure 6 represent the hypothetical German responses to a trade-channel shock, if we set the shock size and the calibrated A/L ratio to the UK value. Again, the responses are closer to the UK responses compared with the original German calibration. As discussed above, the terms-of-trade depreciation has a positive effect on the domestic value of the foreign asset, counteracting to some degree the reduced demand. A larger stock of foreign assets increases this effect, such that bank capital rises more strongly and the spread falls by more. With a higher A/L ratio, a smaller share of foreign assets has to be transformed into loans to satisfy the increased loan demand; the reduction in foreign assets is hence dampened further. Given the larger steady-state export share of Germany, the negative GDP response is nevertheless stronger for the hypothetical German response than for the UK. The largest remaining difference is the behaviour of loans. The larger steady-state stock of loans in Germany leads to a stronger wealth effect for entrepreneurs, as their debt-service payments decline with lower loan rates. Their expenditures can hence be financed to a larger degree out of own funds instead of loans. Imposing equal financial structures, i.e., setting also the German steady-state L/Y and capital-asset ratios to the UK values in addition to the A/L ratio, reduces this difference significantly.31 We conclude that the financial structures are also important for the transmission of the trade-channel shock, while trade openness plays a large role for the transmission of the financial-channel shock as well. Lastly, the green dotted lines in Figures 7 and 8 plot important responses for a model without (dynamic) financial frictions. In this case, neither regulators nor market discipline force the bank to return to a certain capital ratio after disturbances, implying a constant spread.32 Bank capital reduces due to the write-downs in the case of a financial-channel shock, but stays relatively flat afterwards. The constant spread causes the deposit rate to fall by less compared with the baseline, leading to lower worker consumption and hence higher labour supply. The resulting lower wages depreciate the terms of trade and increase exports. As a result, GDP increases slightly for a couple of periods. Fig. 7. Open in new tabDownload slide Germany: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Costs. Notes: Responses for ϕ2 = 0.25/Y (solid line), those for ϕ2 = 0.5/Y (dashed line), and for ϕ2 = 0 (dotted line). Variables are expressed in percentage deviations from steady state, except spread (in percentage points p.a.), capital-asset ratio and net exports over GDP (both in percentage points). Fig. 7. Open in new tabDownload slide Germany: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Costs. Notes: Responses for ϕ2 = 0.25/Y (solid line), those for ϕ2 = 0.5/Y (dashed line), and for ϕ2 = 0 (dotted line). Variables are expressed in percentage deviations from steady state, except spread (in percentage points p.a.), capital-asset ratio and net exports over GDP (both in percentage points). Fig. 8. Open in new tabDownload slide UK: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Costs. Notes: See Figure 7 for description. Fig. 8. Open in new tabDownload slide UK: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Costs. Notes: See Figure 7 for description. The responses to a trade-channel shock are changed less. The initial reaction of loans and deposits is very similar to the baseline. As the spread remains constant, however, there is a slightly smaller positive effect of a falling spread on GDP (and consumption). This effect is stronger for the UK since its spread falls by more in the baseline scenario. If both shocks happen simultaneously, the relative positive effects of the financial channel dominate: GDP and consumption are higher relative to the baseline. Furthermore, note the different long-run implications. Given that bank capital is free to remain at any given level, we obtain a unit root in the model. Bank capital is determined by loans and deposits, besides foreign assets. Their values represent relative wealth positions of the entrepreneur, the worker and the banker. As is usual in models with incomplete markets, they do not return to the initial steady state (in absence of a special mechanism for that purpose). The resulting aggregate long-run effects, however, are quantitatively very small. Considering the simulation results for normal business cycles (Tables 2 and 3) and those for the transmission of the crisis, we find that the financial friction is important for the dynamics of certain variables during normal business cycles. Its largest effect, however, arises in the aftermath of a negative shock that arrives via the financial channel. Here, it reduces GDP (and consumption) considerably. 6. Policy Intervention: Stricter Capital Regulations As a response to the crisis, some authors have asked for stricter capital regulations.33 In this section, we increase the costs that a bank incurs when deviating from the mandatory capital requirement. This can be interpreted as being consistent with the strengthening of supervision after the financial crisis (see, e.g., BCBS, 2015) or the observation that since the financial crisis capital markets pay more attention to banks’ compliance with capital regulation. Alternatively, we also analyse higher mandatory capital requirements, as called for in the wake of the financial crisis (see, e.g., Admati, and Hellwig, 2013). These can loosely be interpreted as being in line with the tighter capital requirements introduced with Basel III (BCBS, 2010). For example, the higher quality of capital required by Basel III would play the same role as a higher capital ratio in our model, since we do not distinguish different types of capital. Our aim, however, is not quantitatively to evaluate the consequences of specific suggestions, but to assess the tradeoffs that regulators might face in the context of adverse spillovers via the trade and financial channels. In particular, we are interested in whether these measures can contribute to a lower and/or shorter recession after the considered external shocks. So far, the parameter for the bank capital cost function was set at ϕ2 = 0.25/Y. A stricter enforcement would lead to larger punishments for violations of the requirement. We explore the implications of these larger punishments by doubling ϕ2 to = 0.5/Y. Figure 7 shows how output, bank capital and the loan-deposit spread react for the German case, while Figure 8 depicts the UK scenario. The initial response of the capital-to-asset ratio to simultaneous financial- and trade-channel shocks (Column 1) or to either one (Columns 2 and 3) is almost the same in the case of a stricter enforcement compared with the baseline calibration. However, the spread increases by more in order to revert bank capital faster back to its steady-state value. The negative GDP response is hence shifted forward, i.e., a stronger initial response is followed by a quicker recovery. Responsibility for this procyclical effect is mainly the altered reaction to the financial-channel shock. In the case of the trade-channel shock, the responses of real variables hardly change. For both shocks together, the German responses are slightly less affected by a stricter regulation, given the relatively lower importance of the financial channel for Germany. The reaction of the spread after simultaneous shocks doubles, while the initial output decline increases from 2.42% to 2.85%. The duration of the recession shortens by two quarters (16 versus 18 periods of GDP below trend). The initial output drop in the UK is 2.80% with stricter regulation, compared with 2.23% before. Again, the factor that is responsible for the stronger reaction is the spread, which more than doubles. As bank capital returns to normal values more quickly, the recession is three periods shorter (17 against 20). As a second approach to stricter banking regulations, we consider a higher capital requirement (we evaluate this measure in isolation, i.e., ϕ2 is set back to its baseline value). In the baseline calibration, we have targeted the long-run loan-to-GDP ratio in both countries. Since it is not clear a priorihow the balance sheets of banks react to higher capital requirements, we will evaluate two extreme cases: one in which the size of bank balance sheets remains unchanged and one in which bank capital is constant but the balance sheet adjusts to meet the higher required capital. In the former case, banks’ liabilities consist of more equity but fewer deposits. To demonstrate the effects intuitively, we require banks to hold a higher capital share by doubling the steady-state bank capital γ relative to its baseline value. Steady-state GDP is hardly altered under this calibration. We depict the hypothetical responses for Germany and the UK by red dashed lines in Figures 9 and 10. We observe similar effects as with the stricter enforcement discussed above: the recession is frontloaded compared with the baseline. Given the lower stock of deposits, the bank cuts the deposit rate by more in order to restore its capital after an adverse financial-channel shock. The resulting higher spread has a negative impact on GDP. The recovery, however, is again faster compared with that in an economy with lower bank capital in steady state. The intertemporally falling consumption profile of workers induced by the lower deposit rate lets wages fall below the baseline response, while hours worked rise above it. For simultaneous trade- and financial-channel shocks, the effects of the new regulation on the financial channel dominate, as the reaction of real variables to a trade-channel shock is hardly altered after the policy intervention. In terms of numbers, the initial GDP drop amounts to 2.70% for Germany and 2.52% for the UK, while the recession lasts 16 quarters in Germany and 18 in the UK. Fig. 9. Open in new tabDownload slide Germany: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Requirement. Notes: Responses for capital requirement (γ) of 4.35% (solid line), for γ = 8.7% with constant balance-sheet size (dashed line), and with constant absolute capital (dotted line). Variables are expressed in percentage deviations from steady state, except spread (in percentage points p.a.), capital-asset ratio and net exports over GDP (both in percentage points). Fig. 9. Open in new tabDownload slide Germany: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Requirement. Notes: Responses for capital requirement (γ) of 4.35% (solid line), for γ = 8.7% with constant balance-sheet size (dashed line), and with constant absolute capital (dotted line). Variables are expressed in percentage deviations from steady state, except spread (in percentage points p.a.), capital-asset ratio and net exports over GDP (both in percentage points). Fig. 10. Open in new tabDownload slide UK: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Requirement. Notes: Responses for capital requirement (γ) of 6.5% (solid line), for γ = 13% with constant balance-sheet size (dashed line), and with constant absolute capital (dotted line). See Figure 9 for further description. Fig. 10. Open in new tabDownload slide UK: Impulse Responses to Financial- and Trade-Channel Shocks, Different Bank Capital Requirement. Notes: Responses for capital requirement (γ) of 6.5% (solid line), for γ = 13% with constant balance-sheet size (dashed line), and with constant absolute capital (dotted line). See Figure 9 for further description. In the alternative experiment, we assume that banks keep the size of their capital constant and thus shrink their balance sheets in order to comply with the higher capital requirement. That is, we double the steady-state bank capital γ, but also adjust the steady-state loan-to-GDP ratio downwards.34 We also obtain a new steady-state level of GDP, around 1.7% below baseline, demonstrating the adverse effects of a lower supply of loans. The responses are shown by the green dotted lines in Figures 9 and 10. Intuitively, the smaller balance sheet features less international assets and hence reduces banks’ exposure to foreign losses. The resulting absolute reduction in bank capital after a financial-channel shock relative to the stock of deposits is comparable with the baseline case, such that the spread rises by a similar amount. The lower absolute loss translates into a quicker recovery (only seven quarters in Germany and 17 in the UK) and a smaller negative impact on GDP (2.26% in Germany and 1.95% in the UK, relative to the new steady state), compared with the baseline. Due to the higher foreign asset holdings, this positive effect is more pronounced for the UK. We conclude that the reaction of banks (and their customers) is important for the effects of imposing a higher capital ratio. While in both cases the recession is shorter, its depth can be larger or smaller relative to the baseline. Thus, the consequences of this policy can only be evaluated by taking further assumptions. To summarise, all considered policy interventions induce a shorter recession in reaction to the financial-channel shock. Note, however, that workers’ relative welfare—expected lifetime utility compared with the baseline—falls by more in the reaction to the shocks with both a stricter capital regulation and a higher equity requirement with a constant balance sheet size.35 Higher financial frictions, resulting in a higher spread, induce relative low investment and GDP. However, given that we do not consider financial stability issues (risks of bank failure), we hesitate to give policy recommendations based only on these results. 7. Conclusion In this article, we have analysed how the recent financial crisis was transmitted internationally. For this purpose, we have employed a quantitative business cycle model featuring trade with the rest of the world as well as a globally acting banking sector. Calibrated to German and UK data, the model can account for key features of empirical regular German and UK business cycles. In order to compare the model predictions for the crisis episode to the data, we estimate a VAR that identifies the effects of external trade- and financial-channel shocks on Germany and the UK. The theoretical predictions replicate important aspects of the effects of the estimated crisis shocks. In particular, the predicted signs and relative strengths of a variety of variables are as observed empirically. Analysing the relative importance of the two transmission channels shows that the trade channel was more important for the transmission of the crisis to Germany, due to its larger trade openness. The financial channel, on the other hand, played a stronger role for the UK. Specifically, the German banking system had to write down a larger share of its foreign assets, but the resulting effects were less severe due to the relatively smaller holdings of foreign assets. This underlines the greater vulnerability of the UK economy to the kind of financial-channel shocks considered here, i.e., losses on foreign assets. In the case of a trade-channel shock, however, larger holdings of foreign assets serve as an automatic hedge because of the induced terms-of-trade depreciation. Since transmission via the financial channel has longer-lasting effects than that via the trade channel, it is responsible for the prolonged recessions that followed the financial crisis, in particular in the UK. By the same token, the cumulated effects of the financial channel are more contractionary than those of the trade channel, except for the reaction of international trade. Regarding possible policy experiments, we consider the effects of two forms of stricter bank capital regulation. It turns out that a stricter implementation of existing requirements would frontload the recession, as banks increase the spread by more to rebuild capital quicker. The effects of higher capital requirements depend on the way in which banks’ balance sheets adjust to this regulation. Additional supporting information may be found in the online version of this article: Online Appendix. Replication Package. Footnotes 1 While there is no unique definition of the financial channel, we focus on transmission via losses on cross-border asset holdings on banks’ balance sheets. These played a major role in the recent financial crisis, as shown below. In Section 1 we define our concept of the financial (and trade) channels in detail. 2 In particular, Germany's house prices were flat for an extended period before the crisis. The UK experienced a larger increase in housing prices, but the construction of new units and employment in the construction sector remained modest due to strict planning laws. The following reduction in house prices, which turned into an upswing in 2009, is hence likely to have had relatively minor effects on aggregate activity, compared with, e.g., the USA, Spain or Ireland. See International Monetary Fund (2010a). 3 We attribute the remaining part to additional transmission channels and/or shocks that are captured in our empirical estimates but not in the model. 4 Other contributions on global banking sectors include Ueda (2012), who shows how financial constraints and the net worth of creditors contribute to business cycle synchronisation in a two-country model similar to that of Kollmann et al. (2011), but with a two-good setup as in the present article. Analysing a model with financial constraints, Mendoza and Quadrini (2010) show how financial contagion can spread across countries through shocks to bank equity. They do not consider business cycles, though. 5 We have also investigated transmission via the domestic financial and trade sectors, i.e., arrow C. Conclusions are similar to those of our analysis of arrow B. Results are available upon request. 6 We first decompose the variance-covariance matrix Ω via a Cholesky-decomposition into |$\mathbf{ \tilde{B}\tilde{B}^{\prime }}$|⁠. Given that any matrix |$\mathbf{ \hat{B}\!=\!\tilde{B} \hat{r}}$|⁠, where |$\mathbf{ \hat{r}}$| is an orthonormal rotation matrix, fulfills the restriction |$\mathbf{ \hat{B}\hat{B}^{\prime }\!=\!\tilde{B}\hat{r} \hat{r}^{\prime } \tilde{B}^{\prime }\!}=\!\Omega$|⁠, we can search for the rotation |$\mathbf{ B\!=\!\tilde{B}r}$| that meets the described identification assumption. We need to maximise the sum of both contributions, as maximising the contribution of each shock separately would result in two different impact matrices. This amounts to finding the maximum of B(i, i)2 + B(j, j)2, where the index i denotes the position of external demand and, simultaneously, of the trade-channel shock, while j is the index for the excess bond premium and the financial-channel shock. Both shocks remain the main drivers of external demand and the excess bond premium over longer horizons, as they together account for more than 85% of the forecast error variance for these variables at a one-year horizon for both countries. 7 The shocks account for between 8% and 25% of the one-year ahead forecast error variance of the respective other variable. 8 Net exports are measured towards the country sample that is used for the construction of the external demand series. 9 Specifically, we subtract the counterfactual level of GDP that is obtained if no shocks are fed into the estimated VAR—i.e., the level of GDP that is explained by the constant, the trend and initial conditions—from the levels of GDP that are obtained if we either feed in all identified shocks or only the two mentioned shocks. 10 We assume entrepreneurs and households to be separate in order to obtain a role for borrowing and lending (with equal discount rates). Introducing bankers as separate agents, on the other hand, is a modelling device to obtain a sensible description of bank behaviour regarding the interest rate spread and lending activities. 11 Consumption habits are often assumed in the literature, as they bring consumption volatility closer to the data. 12 Actually, more than half of the claims by German banks on non-residents are denominated in euros. However, here we focus on financial developments originating in the U.S. financial market and the vast majority of German banks’ claims on the United States are denominated in U.S. dollars, on average around 90%. The same percentage applies to overall claims on non-residents by UK banks. 13 This form guarantees that the bank has an incentive to return to the steady-state bank capital after the occurrence of shocks in the linearised model. Note that an approximation of ϕ1, which plays an important role in the bank's first-order conditions, features ϕ2 times excess capital. The parameter ϕ2 hence determines the effects of bank capital deviating from its target on the bank's costs. 14 Note the difference between our setup and models that emphasise nonlinearities, such as Akinci and Queralto (2014). These authors assume an occasionally binding leverage constraint for banks. The bank in our model faces a similar constraint that becomes stronger the further it falls below a certain capital-asset ratio. Given that banks were still paying dividends in the crisis, we think that an increasingly stricter constraint is a good description for the restrictions that banks were facing during the crisis. 15 Our following results do not change significantly if we end the sample in 2007Q4. 16 Besides the positive influence of lower risk aversion on the decision to become entrepreneurs, the implied higher volatility of entrepreneurial consumption is in line with the empirical finding of a higher consumption volatility of wealthier people: see Parker and Vissing-Jørgensen (2009) for evidence based on the U.S. Consumer Expenditure Survey. Similarly, Vissing-Jørgensen (2002) shows that the implied intertemporal elasticity of substitution of stock holders is much higher than that of non-stock holders. Ait-Sahalia et al. (2004) document that sales of high-end luxury goods are an order of magnitude more volatile than aggregate consumption. More importantly, we also obtain sensible predictions for bank behaviour. There is widespread evidence that banks were reluctant to cut dividends during the financial crisis, despite accumulating losses. See, e.g., Floyd et al. (2015) for U.S. banks and Acharya et al. (2016) for Europe. Setting σB = 1 gives predictions in line with these observations. 17 We note that the consumption share of the worker in the model is 71.3% for Germany and 67.0% for the UK in steady state. Subtracting their interest income (from deposits), we obtain 70.4% (Germany) and 65% (UK) as the labour share. Although not calibration targets, these values are close to the average observed labour shares of 63% in Germany and 66% in the UK (averages over the period 1991–2012 from the AMECO data base of the EU, corresponding values without crisis period: 64% for Germany and 66% for the UK). 18 Using write-downs instead of changes in the value of asset holdings allows us to disentangle losses from, e.g., sales. 19 We use a smoothing coefficient of 1,600 and take logs of all variables before filtering, except for net exports and the interest rate spread, as these variables are already expressed in percentage points. We also employed a second-order approximation, yielding virtually identical results for all experiments in the article. 20 We thereby avoid a unit root. Investment adjustment costs are reset to generate the observed investment volatilities. 21 Contrary to us, Kollmann et al. (2011) find a more limited role for financial frictions during normal business cycles in a distinct but related model. This can be explained by the different types of financial shocks considered. While they focus on loan-default shocks, which were at the centre of the financial crisis in the United States, we consider variable returns on foreign asset holdings, as they were an important transmission channel from the United States to financially connected countries. These returns are much more volatile than losses on loans. Furthermore, write-downs on domestic loans are equivalent to transfers to domestic entrepreneurs in Kollmann et al. (2011). In our model, the lost resources have a stronger effect as they do not appear in other agents’ budget constraints, given that they represent transfers to the foreign country. 22 We take these differences in economic structures as given. Explaining those is the subject of a large literature, see, among others, Tilly (1989) and Allen and Gale (2000). 23 In the Online Appendix we discuss the effects on business cycle statistics. 24 Here and in the following we simulate the model either for simultaneous or for isolated occurrences of the financial- and/or the trade-channel shock. We set the correlation between shocks to zero, such that the other shock (in the case of isolated shocks) and TFP remain constant. 25 The numbers add up to more than 100% for some variables, as the maximum responses following shocks through each channel can occur at different times. 26 While the positive consumption response might seem counterfactual at first sight, note that we consider the effects of a very temporary reduction in foreign demand. In the model, it is known to all agents that external demand will recover fairly quickly, such that negative wealth effects are limited. We thereby exclude elements that could reduce consumption and investment, such as uncertainty about the future development of external demand. The exclusion of this and related channels is in line with our analysis, as we are merely interested in the narrowly defined trade channel. Setting σE = 1 eliminates the positive consumption response without significantly changing the remaining responses. 27 Remember that bank capital in period t is Lt + 1 + ptAt + 1 − Dt + 1, where the terms of trade pt unexpectedly rise following a trade-channel shock. Asset returns typically fall in times of lower external demand because of deteriorating economic situations abroad, another factor that links falling external demand to negative wealth effects. This effect is deliberately not captured in the trade, but in the financial channel of our model. 28 Figures A1 and A2 in the Online Appendix visualise these findings in a comprehensive way. 29 Figures A3 and A4 in the Online Appendix show the responses for the unchanged German calibration with shocks as estimated for the UK. Considerable differences to the UK remain. 30 Figure A3 in the Online Appendix depicts the corresponding impulse-response functions for the financial-channel shock. 31 Figure A4 in the Online Appendix depicts the corresponding impulse-response functions for the trade-channel shock, which are very close to the UK responses. 32 As we simulate a one-time shock, we can set ϕ2 = 0 despite the resulting unit root, different to the business cycle statistics above. The steady-state interest-rate spread is left at its baseline value to compare scenarios with equal steady states. 33 Bank capital requirements, which generally affect the fraction of liabilities that banks can lend and the resulting credit spreads, have been analysed by a number of studies; see, e.g., Quint and Rabanal (2014) and Brzoza-Brzezina et al. (2015). The exact specification of the bank capital requirement depends on the model and the question asked. 34 We hence need to change ΨN in the utility function. Considering this parameter invariant implies that a constant size of the balance sheet after a required higher bank capital is the more likely outcome. In response to the Basel III reforms that introduced tighter capital requirements, banks have indeed increased their absolute levels of capital (see, e.g., the Basel III Monitoring Report of September 2017). 35 If the balance sheet shrinks in reaction to the regulation, workers’ welfare in reaction to the crisis shocks falls by less relative to the respective steady states if compared with the baseline. This is again due to the reduced stock of foreign assets that is hit by the financial-channel shock. However, steady-state GDP and welfare are lower. Notes We thank Benjamin Born, Gernot Müller, Johannes Pfeifer, three anonymous referees and seminar participants at the IMF, the CEPR Conference ‘The European Crisis—Causes and Consequences’ in Bonn, the annual conference of the Verein für Socialpolitik in Münster, the IMFS Conference on Monetary and Financial Stability in Frankfurt, and the DFG SPP 1578 conference in Cologne for helpful comments. Contact details:zeno.enders@uni-heidelberg.de (Bergheimer Str. 58, 69115 Heidelberg) and alexandra.born@ecb.europa.eu. Part of this research was conducted while Enders was a visiting scholar at the IMF and the Board of Governors, whose hospitality is gratefully acknowledged. Enders thanks the DFG for financial support under project number EN892/1. The views expressed in this article are solely those of the authors and do not necessarily reflect the views of the European Central Bank, the IMF, the Board of Governors or their staff. References Abate A. , Eickmeier S. , Lemke W. , Marcellino M. ( 2016 ). ‘The changing international transmission of financial shocks: evidence from a classical time-varying FAVAR’ , Journal of Money, Credit and Banking , vol. 48 ( 4 ), pp. 573 – 601 . Google Scholar Crossref Search ADS WorldCat Acharya V.V. , Pierret D. , Steffen S. ( 2016 ). ‘High time to tell European banks: no dividends’ , http://pages.stern.nyu.edu/∼sternfin/varachya/public_html/pdfs/dividends_v4Aug2016.pdf, mimeo . WorldCat Admati A.R. , Hellwig M. ( 2013) . The Bankers’ New Clothes: What's Wrong with Banking and What to Do About It , Princeton, NJ: , Princeton University Press . Ait-Sahalia Y. , Parker J. , Yogo M. ( 2004 ). ‘Luxury goods and the equity premium’ , Journal of Finance , vol. 56 ( 6 ), pp. 2959 – 3004 . Google Scholar Crossref Search ADS WorldCat Akinci O. , Queralto A. ( 2014 ). ‘Banks, capital flows and financial crises’ , Board of Governors of the Federal Reserve System, International Finance Discussion Papers Nr. 1121 . WorldCat Allen F. , Gale D. ( 2000 ). Comparing Financial Systems , Cambridge, MA :, MIT Press . Backus D.K. , Kehoe P.J. , Kydland F.E. ( 1994 ). ‘Dynamics of the trade balance and the terms of trade: the J-curve?’ , American Economic Review , vol. 84 ( 1 ), pp. 84 – 103 . WorldCat BCBS. ( 2010 ). ‘Basel III: a global regulatory framework for more resilient banks and banking systems’ , Basel Committee on Banking Supervision, https://www.bis.org/bcbs189.htm, revised June 2011 . WorldCat BCBS . ( 2015 ). Report on the Impact and Accountability of Banking Supervision , Basel Committee on Banking Supervision , https://www.bis.org/bcbs/publ/d326.htm . Bekaert G. , Ehrmann M. , Fratzscher M. , Mehl A. ( 2014 ). ‘Global crises and equity market contagion’ , Journal of Finance , vol. 69 ( 6 ), pp. 2597 – 649 . Google Scholar Crossref Search ADS WorldCat Brzoza-Brzezina M. , Kolasa M. , Makarski K. ( 2015 ). ‘Macroprudential policy and imbalances in the euro area’ , Journal of International Money and Finance , vol. 51 , pp. 137 – 54 . Google Scholar Crossref Search ADS WorldCat Davis S. , 2010 , ‘ The adverse feedback loop and the effects of risk in both the real and financial sectors . ’ Wokring Paper No. 66, Federal Reserve Bank of Dallas . WorldCat Enders Z. , Müller G.J. ( 2009 ). ‘On the international transmission of technology shocks’ , Journal of International Economics , vol. 78 : (1 ), pp. 45 – 59 . Google Scholar Crossref Search ADS WorldCat Floyd E. , Li N. , Skinner D.J. ( 2015 ). ‘Payout policy through the financial crisis: The growth of repurchases and the resilience of dividends’ , Journal of Financial Economics , vol. 118 ( 2 ), pp. 299 – 316 . Google Scholar Crossref Search ADS WorldCat Gerali A. , Neri S. , Sessa L. , Signoretti F.M. ( 2010 ). ‘Credit and banking in a DSGE model of the euro area’ , Journal of Money, Credit and Banking , vol. 42 ( s1 ), pp. 108 – 41 . WorldCat Gilchrist S. , Zakrajšek E. ( 2012 ). ‘Credit spreads and business cycle fluctuations’ , American Economic Review , vol. 102 ( 4 ), pp. 1692 – 720 . Google Scholar Crossref Search ADS WorldCat International Monetary Fund . ( 2010b ). 'Global Financial Stability Report: Meeting New Challenges to Stability and Building a Safer System' . Washington, DC, April . International Monetary Fund . ( 2010a ). ‘United Kingdom: 2010 article IV consultation—staff report’ , IMF Country Report No. 10/338 , Washington, DC . WorldCat Kollmann R. ( 2013 ). ‘Global banks, financial shocks, and international business cycles: evidence from an estimated model’ , Journal of Money, Credit and Banking , vol. 45 : (s2 ), pp. 159 – 95 . Google Scholar Crossref Search ADS WorldCat Kollmann R. , Enders Z. , Mller G. ( 2011 ). ‘Global banking and international business cycles’ , European Economic Review , vol. 55 : (3 ), pp. 307 – 442 . WorldCat Lane P.R. , Milesi-Ferretti G.M. ( 2011 ). ‘The cross-country incidence of the global crisis’ , IMF Economic Review , vol. 59 : (1) , pp. 77 – 110 . Google Scholar Crossref Search ADS WorldCat Mendoza E.G. , Quadrini V. ( 2010 ). ‘Financial globalization, financial crises and contagion’ , Journal of Monetary Economics , vol. 57 ( 1 ) pp. 24 – 39 . Google Scholar Crossref Search ADS WorldCat Olafsson T.T. , Pétursson T.G. ( 2011 ). ‘Weathering the financial storm: the importance of fundamentals and flexibility’ , in ( Beblavý M. , Cobham D. , Ódor L. , eds.), The Euro Area and the Financial Crisis , pp. 23 – 55 ., Cambridge : Cambridge University Press . WorldCat Parker J.A. , Vissing-Jørgensen A. ( 2009 ). ‘Who bears aggregate fluctuations and how?’ , American Economic Review , vol. 99 ( 2 ), pp. 399 – 405 . Google Scholar Crossref Search ADS WorldCat Quint D. , Rabanal P. ( 2014 ). ‘Monetary and macroprudential policy in an estimated DSGE model of the euro area’ , International Journal of Central Banking , vol. 10 ( 2 ), pp. 169 – 236 . WorldCat Reinhart C.M. , Rogoff K.S. ( 2009 ). ‘The aftermath of financial crises’ , American Economic Review , vol. 99 ( 2 ), pp. 466 – 72 . Google Scholar Crossref Search ADS WorldCat Rose A.K. , Spiegel M.M. ( 2011 ). ‘Cross-country causes and consequences of the crisis: an update’ , European Economic Review , vol. 55 : (3 ), pp. 309 – 24 . Google Scholar Crossref Search ADS WorldCat Tilly R.H. ( 1989 ). ‘Banking institutions in historical and comparative perspective: Germany, Great Britain and the United States in the nineteenth and early twentieth century’ , Journal of Institutional and Theoretical Economics , vol. 145 ( 1 ), pp. 189 – 209 . WorldCat Ueda K. ( 2012 ). ‘Banking globalization and international business cycles: cross-border chained credit contracts and financial accelerators’ , Journal of International Economics , vol. 86 ( 1 ), pp. 1 – 16 . Google Scholar Crossref Search ADS WorldCat Vissing-Jørgensen A. ( 2002 ). ‘Limited asset market participation and the elasticity of intertemporal substitution’ , Journal of Political Economy , vol. 110 ( 4 ), pp. 825 – 53 . Google Scholar Crossref Search ADS WorldCat © 2019 Royal Economic Society. Published by Oxford University Press. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
TI - Global Banking, Trade, and the International Transmission of the Great Recession
JF - Economic Journal
DO - 10.1093/ej/uez010
DA - 2019-10-01
UR - https://www.deepdyve.com/lp/oxford-university-press/global-banking-trade-and-the-international-transmission-of-the-great-MD818yJzCs
SP - 2691
VL - 129
IS - 623
DP - DeepDyve
ER -