BUSINESS CYCLES, Exams of Business

Download BUSINESS CYCLES and more Exams Business in PDF only on Docsity! STRUCTURAL POLICIES AND ECONOMIC RESILIENCE TO SHOCKS by Romain Duval, Jørgen Elmeskov and Lukas Vogel1 1. A trend towards more moderate business cycle fluctuations is often quoted as a stylised feature of economic developments in OECD countries over the past several decades. Among the causes frequently cited are better macroeconomic policies that have helped to anchor inflation expectations, a lower incidence and size of outside exogenous shocks, better financial market instruments for risk allocation and a reduced role for and better control of inventories.2 Reflecting the more moderate cycle within countries, cyclical divergences across countries have also tended to shrink over time. As a result, macroeconomic policy requirements have become less divergent across countries. This is obviously important when it comes to countries inside the euro area, where monetary policy settings are by definition identical. 2. Much more controversial is whether, and to what extent, business cycles have become more synchronised across countries. A factor making for synchronous business cycles are common shocks. Historically, oil price hikes have been prominent in this respect, but the oil intensity of OECD economies has tended to decline over time, implying that oil price fluctuations have become less important as a source of large common shocks. Another factor potentially making for synchronous business cycles is propagation of shocks across countries. Here, the increasing trade and financial linkages between countries are likely to have led to faster and stronger transmission of shocks across borders. A particular issue relates to business cycles across euro area countries, with the common currency potentially leading to greater integration and faster transmission, and thereby better alignment of cycles (Frankel and Rose, 1998), but also to greater specialisation and thereby a larger role for idiosyncratic shocks (Krugman, 1993). 3. Apart from idiosyncratic shocks, business cycle divergence across countries may also reflect different responses to common shocks. At issue here is the extent to which some economies are more resilient than others to various shocks. Since 2001, the experience of the large continental European economies contrasts with that of English-speaking OECD countries and many smaller European economies. Many of the shocks hitting countries appeared to be similar between the two groups or even marginally larger in some of the countries in the second group, such as with mass terrorism, the bursting of the equity bubble and corporate governance scandals. Yet, growth performance was generally better in the second group of countries (even adjusting for typically higher rates of potential growth) and even when 1. The authors are, respectively, senior economist, director and economic researcher at the OECD Economics Department. They would like to thank Christophe André, Benoît Bellone, Jean-Philippe Cotis, Boris Cournède, Sébastien Jean and Dave Rae for helpful discussions. We also thank participants to the March 2007 OECD Working Party No.1 workshop on resilience for comments and suggestions. Remaining errors and omissions are the responsibility of the authors. The views expressed do not necessarily represent those of the OECD or its member governments. 2. See e.g. Dalsgaard et al. (2002); Stock and Watson (2003). 1 recessions occurred they were usually short-lived, with economies bouncing back smartly. It seems unlikely that the differences in performance between the two country groups can be explained by different macroeconomic policy settings, even if these may in some cases have contributed.3 As a result, the hypothesis has emerged that economic resilience is stronger in some countries than in others. Interestingly, and perhaps not coincidentally, the countries seen as more resilient also appear to be the ones that have made most progress on structural reform over the past two decades. 4. Economic resilience may be loosely defined as the ability to maintain output close to potential in the aftermath of shocks. Hence, it comprises at least two dimensions: the extent to which shocks are dampened and the speed with which economies revert to normal following a shock. Structural policies are likely to affect both the strength and persistence of the effects of outside exogenous shocks. Macroeconomic stabilisation policies will also play an important role for resilience, but their effectiveness will also be conditioned by structural policy settings. For example, structural policy settings may affect the strength of the monetary policy transmission mechanism. 5. Against this background, the current paper first reviews simple evidence on business cycle volatility and convergence among OECD countries and within the euro area.4 It then examines at greater length the impact of a range of structural policies on the resilience of economies to shocks, both across countries and over time. A final section sums up the main findings and concludes. 1. Stylised features of business cycles 1.1 Business cycle indicators 6. Business cycles are unobservables and indicators of business cycles rely on the separation of economic developments into trend and cyclical components. This paper focuses exclusively on developments in the volume of GDP and considers three different procedures for decomposing them into trend and cycle.5 Two rely on purely statistical procedures: the Hodrick-Prescott and the Baxter-King filters.6 The cyclical component of GDP, or the output gap, is derived as the difference in per cent between actual and trend GDP. In addition to output gaps derived using these statistical methods, OECD estimates of output gaps are also used. These are constructed as deviations from a trend calculated using a production function, taking as given actual capital stocks and trends of total factor productivity and employment (which in turn is derived based on estimates of the trend participation rates, the NAIRU and trend working hours).7 3. See Cotis and Coppel (2005). 4. Part of this descriptive analysis is an update of prior OECD analysis in Dalsgaard et al. (2002) and Cotis and Coppel (2005). 5. Dalsgaard et al. (2002) provide evidence on cyclical behaviour of a number of demand components. 6. In line with standard practice on quarterly numbers, a HP-filter with a smoothing parameter of 1 600 is used. The Baxter-King filter is implemented so as to remove high-frequency components of less than six and low-frequency components of more than 32 quarters. In practice, in order to mitigate the usual “end- point” problem, these filters are implemented over the period 1960Q1-2008Q4 (1963Q1-2008Q4, 1966Q1- 2008Q4 and 1970Q1-2008Q4 for France, Denmark and Korea, respectively), using OECD forecasts (as published in OECD Economic Outlook 80) to extrapolate GDP data up to the fourth quarter of 2008. 7. See e.g. Cotis et al. (2005). 2 a result of these developments, idiosyncratic fluctuations appear to have become substantially smaller than the common fluctuations among virtually all euro area countries since 1990. [Table 1.2. Idiosyncratic relative to common fluctuations over time] [Table 1.3. Idiosyncratic relative to common fluctuations in the euro area] 1.3 Summing up the descriptive evidence 16. Overall, there is clear evidence that cyclical fluctuations have become smaller over the past three or four decades. By contrast, based on the very simple indicators used here there is no strong evidence to suggest that cycles have become more synchronised across OECD countries. However, among euro area countries, some signs of greater synchronisation are in evidence. The most recent empirical literature does not yield clear-cut results either. Focusing on cross-country correlations between cyclical components and other simple indicators such as those used above, Helbling and Bayoumi (2003) conclude that business cycle linkages have remained unchanged over the 1973-2001 period. Benalal et al. (2006) find evidence of increased business cycle co-movements across euro area economies, but Camacho et al. (2006) obtain no such result using industrial production series and more sophisticated measures of synchronisation -- such as VAR-based or spectral-based approaches. Bergman (2006) even finds that business cycles were more coordinated across euro area members before the implementation of the “hard” exchange rate mechanism (in 1987) than after. 17. The continued importance of idiosyncratic as compared with common cyclical fluctuations, in particular at the level of the OECD at large, may have several causes. One is that economies differ in their degree of resilience to otherwise similar shocks. The rest of the paper explores to what extent different structural policy settings lead to different degrees of resilience across countries. 2. Structural policy determinants of resilience to shocks 18. As already noted, two key dimensions of resilience are the ability of the policy and institutional framework to cushion the initial impact of shocks and to reduce the persistence of the ensuing output gap. In this respect, those policies and institutions that dampen the initial impact of a shock may actually increase its persistence, and vice versa, i.e. they may have conflicting effects on resilience. For instance, strict employment protection legislation (EPL) may deter firms from laying off workers in the short run, thereby supporting employment and private consumption. At the same time, it may slow down the wage adjustment process (Blanchard and Summers, 1986) as well as workers’ reallocation towards more productive jobs, thereby delaying the return of employment and output to their initial levels. Box 1 offers a more coherent theoretical framework for thinking about links between rigidities in labour and product markets and resilience. The upshot is that there is no simple link between policy-induced rigidities and resilience and that the net effect of structural policies in product and labour markets is essentially an empirical issue. 19. Another potential determinant of economic resilience to shocks is the strength of monetary policy transmission channels. Here the expected effect is less ambiguous: in general, the more powerful the monetary transmission mechanism, the smaller and the less persistent the monetary policy and output responses to demand shocks.12 Among the host of factors that contribute to shape monetary transmission channels, the degree of liberalisation of financial markets plays an important role, not least by facilitating 12. In principle, output could also be stabilised in response to demand shocks under a weak transmission mechanism but this would be associated with an instrument variability that might be unpalatable. In the face of supply shocks, the strength of the transmission mechanism might at worst be irrelevant to the inflation and output pattern. 5 intertemporal consumption smoothing. For example, the degree of mortgage market “completeness” (the range and variety of mortgage products available to borrowers) has been shown to amplify the transmission channel from housing wealth to consumption (Catte et al., 2004). More broadly, econometric analyses of consumption behaviour typically find larger “wealth effects” from housing and financial assets in those countries that have the most liberalised financial markets (see e.g. Deroose, 2006). That said, the use a country can make of an effective monetary transmission mechanism obviously depends on the chosen exchange rate policy.13 For example, small members of a monetary union will not be helped by a strong transmission mechanism when faced with an idiosyncratic shock because monetary policy, calibrated on the union average, will not respond. Likewise, when confronted with a common shock, members of a monetary union can have too much of a good thing in the sense that a transmission mechanism that is stronger than average, together with a monetary policy response that is calibrated on the average, could be destabilizing. The analysis below controls for the influence of exchange rate policy as a constraint on monetary policy that may make it more difficult to stabilise the economy in the face of idiosyncratic shocks and/or heterogeneous propagation mechanisms of common shocks. 20. Fiscal policy may also affect resilience patterns via two main channels. First, automatic stabilisers are expected to dampen the impact of shocks. Strong automatic stabilisers are typically associated with large public sectors, which in turn partly reflect some of the structural policies mentioned above -- such as high and long-lasting unemployment benefits. Second, discretionary fiscal policy may be stabilising or destabilising, depending on whether it is counter or pro-cyclical. Here, one might expect government size, which allows strong automatic stabilisers, to be associated with a reduced need for discretionary fiscal impulses. This may not always be the case in practice, however. Evidence in Ahrend et al. (2006) suggests that several countries with large public sectors have supplemented automatic stabilisers with sizeable discretionary (counter-cyclical) actions over the past two decades. These were on balance stabilising in Nordic countries, but destabilising in many euro-area countries. Fiscal policy was more in line with expectations in countries with smaller public sectors -- such as a number of English-speaking and Asian OECD countries --, with sizeable discretionary impulses contributing to output stabilisation. Box 1. Theoretical considerations on the link between product and labour market rigidities and resilience Modern business cycle theory offers a systematic take on the link between rigidities and resilience. In a basic New Keynesian model, greater (nominal) wage and/or price stickiness flattens the (New Keynesian) Phillips curve and increases the sacrifice ratio. In turn, under optimal monetary policy,1 an independent central bank credibly committed to medium-term price stability will react less aggressively to most shocks -- including temporary but persistent cost- push or technology shocks -- thereby engineering a smaller but more prolonged output gap response (see e.g. Altissimo et al., 2006). The intuition is that since nominal rigidities worsen the inflation-output variability trade-off, a more aggressive policy reaction to a cost-push shock would induce large output losses for a limited gain in terms of reduced inflation.2 By contrast, in the case of pure demand shocks, rigidities may be of little influence since monetary policy can readily stabilise aggregate demand without facing any trade-off between output and inflation stabilisation. Any policy or institution that increases wage and/or price stickiness would therefore be expected to lead to a smaller but more persistent output reaction to certain shocks. Among the many theoretical underpinnings of price stickiness, imperfect competition in product markets features prominently, e.g. through menu costs or coordination failure3 approaches.4 On the empirical side, there is now fairly strong evidence at the microeconomic level that firms tend to reset their prices more frequently in more competitive markets, lending some support to the view that low product market competition increases price stickiness (see e.g. the recent analysis carried out within the context of the Eurosystem Inflation Persistence Network, including inter alia Alvaréz et al., 2006, and Fabiani et al., 2006). Likewise, among the various theoretical explanations for wage stickiness, some authors have stressed the role played by labour market policies and institutions.5 For example, stringent EPL and/or high coverage of collective agreements bargained between unions and firms may slow down the adjustment of labour contracts in the face of shocks and thereby be conducive to nominal wage rigidities (Holden, 1994, 2004). Like price stickiness, real wage rigidities flatten the Phillips curve and increase the sacrifice ratio. Real wage rigidities may be strengthened, for example, by high unemployment benefit replacement rates available over long 13. The exchange rate is part of the monetary transmission mechanism, but this aspect is not explicitly covered in the current analysis. 6 periods. Ceteris paribus, rigid real wages should induce a less aggressive monetary policy response, and therefore a smaller but more persistent output reaction, to a variety of shocks. However, unlike price stickiness, real wage rigidities also increase the persistence of inflation, which should prompt monetary authorities to be more aggressive, thereby engineering a larger but less persistent output reaction to shocks. The latter effect may dominate in practice.6 To sum up, while nominal rigidities should lead to smaller but more persistent output gaps, real rigidities might go in the opposite direction.7 In both cases, the implications for resilience are expected to be ambiguous a priori.8 ________ 1. Minimising a quadratic loss function defined over both inflation and output gaps. 2. Given the convexity of the central bank’s welfare loss function, such a policy response would not be optimal. Another reason for the central bank to react less aggressively to a cost-push shock in the presence of price stickiness is that the initial impact of the shock on inflation will be smaller, for example because firms reset prices less frequently. 3. Co-ordination failure relates to the observation that in oligopolistic markets, following a shock, firms may choose not to change their prices unless their competitors move first. 4. Furthermore, Rotemberg and Woodford (1991) have suggested that during upswings, oligopolistic firms have greater incentives to free ride on other firms’ efforts to maintain collusive price behaviour, so that mark-ups should fall. This counter-cyclicality of mark-ups provides another reason to expect that low product market competition flattens the Phillips curve. 5. Other factors may also play a role. For example, in conditions of low and stable inflation, contracts may lengthen which could induce greater nominal inertia. 6. Work undertaken as part of this paper shows that when real wage rigidities à la Blanchard and Galí (2005) are introduced in an otherwise basic New Keynesian model, the optimal monetary response to a cost-push shock is more aggressive than under flexible real wages. This more aggressive policy response is associated with a larger but less persistent impact on the output gap. A similar conclusion appears to hold for technology shocks. 7. Model simulations suggest that this conclusion also holds within a monetary union. In a member country with an above-average degree of nominal rigidity, the initial impact of a common cost-push shock on prices is smaller. This results in a competitiveness gain which in turn mitigates the impact of the common shock on the output gap. However, this comes at the cost of a more persistent output gap, as more flexible countries quickly restore their competitiveness through the larger negative impact on inflation of their negative output gap. By contrast, in a member country with an above-average degree of real wage rigidity, the dominant effect is that the common cost-push shock brings about higher wages and therefore a loss of competitiveness. This in turn results in a larger initial impact of the common shock on the output gap.8. It should also be stressed that the empirical notion of output gap used in this paper differs from the theoretical concept of output gap featured in New Keynesian models. The former is essentially the difference between actual and smoothed GDP, while the latter is the difference between actual and “natural” output, where -- in line with the real business cycle literature -- natural output may be quite volatile, e.g. due to temporary technology shocks. 2.1 Modelling strategy and preliminary cross-country comparison of business cycle patterns 21. With resilience determined by both amplification and persistence mechanisms, an empirical investigation of the phenomenon has to be dynamic in nature. This sub-section explores the determinants of cross-country business cycle patterns by means of dynamic, panel data output gap equations. As the first section has shown, most features of business cycles appear to hold independently of the exact measure chosen. Therefore, the analysis undertaken below focuses on only one of these, namely the OECD output gap measure. Another obvious consideration is that the modelled dynamics should fit OECD output gap patterns. As discussed in Box 2, visual inspection and specification tests point to an AR(2) specification for describing output gaps (as well as unemployment gaps, dealt with below in Box 3). 22. As a starting point, the following dynamic (non-linear) panel regression is estimated for a sample of 20 OECD countries14 using annual OECD output gap data over the period 1982-2003:15 14. Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States. 15. The choice of the time period is driven by the availability of policy and institutional indicators, in particular the OECD EPL index which starts in 1982. 7 2.2 The impact of labour and product market regulation on business cycle patterns Correlations between policies and resilience parameters 26. As preliminary evidence of the effects of labour and product market regulation on business cycle patterns, the country-specific persistence and amplification coefficients φi and γi estimated in Table 2.1 are retrieved and regressed on the following indicators (averaged for each country over the period 1982-2003) in separate regressions: 21 • The unemployment benefit replacement rate (averaged across a variety of income levels, family situations and unemployment durations). • The stringency of employment protection legislation for regular workers (EPL). • The stringency of product market regulation (PMR) across seven non-manufacturing industries.22 • Collective bargaining coverage, i.e. the share of workers covered by a collective agreement, a measure of union influence in wage bargaining.23 • The degree of centralisation/co-ordination of wage bargaining, a proxy for the concept of “corporatism” which has received widespread attention in the comparative political economy literature. In practice, the variable entered in the econometric estimates is a “low corporatism” dummy variable, which equals 1 if bargaining is decentralised and uncoordinated and zero otherwise. 27. Cross-country correlation coefficients between persistence and amplification coefficients φi and γi on the one hand, and the above policy indicators on the other, are presented in Table 2.2. Given that persistence and amplification coefficients are estimated rather than observed -- so that standard statistical inference could be misleading -- the critical values used to assess the statistical significance of each correlation coefficient are obtained by bootstrapping the regression residuals. Strict EPL, stringent PMR and a high degree of corporatism appear to be negatively correlated with the initial impact of shocks but positively with output gap persistence. Similar but insignificant correlation signs are obtained for collective bargaining coverage and the unemployment benefit replacement rate. In a nutshell, the results from Table 2.2 suggest that strict labour and product market regulations may dampen the initial impact of a common shock while making it more persistent. 21. See data appendix for full details on data sources and methods. These indicators are time-varying and available at an annual frequency, with the exception of collective bargaining coverage which is time- invariant (country average over 1980-2000) due to lack of data at an annual frequency over the sample period. 22. This sector-based PMR indicator is used in this paper because it covers the whole sample period, unlike the OECD’s economy-wide indicator which is available only for 1998 and 2003. One drawback is that changes in the indicator for non-manufacturing sectors do not incorporate all aspects of regulatory reforms that have been undertaken by a number of OECD countries in the past decades, such as administrative reforms affecting all sectors. As a result, the resilience effects of regulatory reforms may not be fully captured by the econometric estimates presented in this paper. 23. This variable is less imperfect than union density, not least because administrative extension practices -- which remain in place in a number of continental European countries -- extend collective agreements to the non-affiliated, providing unions with greater bargaining power in practice than union membership rates would suggest. 10 [Table 2.2. Cross-country correlation coefficients between persistence/amplification coefficients and labour/product market policy indicators] 28. This is further supported by the last row of Table 2.2, which finds similar and statistically significant cross-country correlations between persistence/amplification coefficients and a synthetic indicator of labour and product market regulation (averaged for each country over 1982-2003). The rationale for constructing this indicator is that countries tend to have similar stances across policy areas, thereby making it difficult to isolate the impact of a particular policy (Table 2.3). For instance, those countries that have strict EPL also tend to have stringent PMR, and vice versa. Here, this synthetic indicator is computed as the first principal component of the previous set of policy indicators (see e.g. Nicoletti and Scarpetta, 2005). Based on the factor loadings produced by the analysis, it can be written as: Labour and product market regulationit = 0.42*(replacement rateit) + 0.45*(EPLit) + 0.48*(collective bargaining coverageit) - 0.51*(low corporatismit) + 0.37*(PMRit) This synthetic indicator has intuitive appeal. It is not very different from a simple average of the underlying policy indicators, so that it can to some extent be interpreted as a simple summary measure of the stringency of labour and product market regulation in the economy. Furthermore, it appears to explain over half of the total variance in the institutional data, which strongly suggests that the dataset can indeed be reduced into one single component without losing too much information in the process.24 [Table 2.3. Correlation coefficients between labour and product market regulation indicators] Estimating the effects of policies on business cycle patterns 29. In order to undertake more in-depth econometric analysis of the effects of labour and product market regulation on business cycle patterns and resilience, persistence and amplification coefficients φi and γi in equation [2.1] are replaced by a (linear) function of (time-varying) indicators of labour and product market regulation: iti k kk it k titit j jj it j it XXGAPGAPXXGAP εαγληϕϕ ++⎟ ⎠ ⎞ ⎜ ⎝ ⎛ −++−⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛ −+= ∑∑ −− )(1)()( ..21.. [2.2] where the Xjs and Xks are the indicators of policies and institutions in labour and product markets, namely the unemployment benefit replacement rate, EPL, PMR, collective bargaining coverage and the low corporatism dummy variable. 30. In equation [2.2] both the persistence and amplification of unobserved shocks are supposed to be functions of policies and institutions, while they were assumed to be constant and country-specific in equation [2.1]. This specification implicitly assumes that persistence and amplification coefficients depend exclusively on policy and institutional factors, and therefore does not allow for cross-country differences in these effects due to other factors.25 A positive (negative) and significant ϕj implies that the policy or institution Xj increases (reduces) the persistence of output gaps. Likewise, a positive (negative) and 24. The so-called “Kaiser rule” suggests that one should retain only factors with eigenvalues greater than one (Kaiser, 1960). Here, only the first component meets this criterion. 25. For example, the impact of a common oil-price shock depends not only on policy and institutional settings but also on the oil intensity of output, which may vary across countries. 11 significant γk implies that the policy or institution Xk amplifies (mitigates) the initial output gap effect of a shock.26 31. The tendency for policy settings in different domains to be correlated, as discussed above, tends to generate multicollinearity which prevents the estimation of equation [2.2] with the full set of policies and institutions, however.27 There is no straightforward way to address this issue, all the more so as the most recent automatic “general-to-specific” model selection procedures are currently applicable only for linear dynamic models (see e.g. Hoover and Perez, 1999; Krolzig and Hendry, 2001). Here, the multicollinearity issue is addressed in two alternative ways: • A “statistical tournament” is undertaken in order to identify the most influential policy indicators within the full set of policies and institutions. • Alternatively, equation [2.2] is estimated using the synthetic indicator of labour and product market regulation instead of the set of individual policy indicators. Identifying the labour and product market policies with most influence on the business cycle 32. In order to identify the policies and institutions with the greatest influence on resilience patterns, the following “statistical tournament” is undertaken. As a preliminary step, equation [2.2] is estimated with policy indicators entered individually in separate regressions. These results, which are presented in Table 2.4, are fairly consistent with theoretical priors and with the simple correlation exercise above. Strict EPL and stringent PMR are found to mitigate the initial impact of shocks on output gaps while at the same time increasing persistence (Table 2.4, Columns 2 and 5). Decentralised and uncoordinated wage bargaining processes appear to amplify the initial impact of shocks, providing some support for the view that wages are more responsive to changes in overall macro-economic conditions in highly corporatist systems (Table 2.4, Column 3). However, somewhat in contrast with this view, a low degree of corporatism appears to reduce output gap persistence. One tentative explanation for these findings might be that decentralised bargaining processes do not properly internalise the effects of changes in the macro- economic environment but otherwise bring in more wage flexibility. As might be expected, high rates of collective bargaining coverage are associated with stronger output gap persistence (Table 2.4, Column 4). In a number of OECD countries, high collective bargaining coverage stems from legal extension procedures, by which collective agreements become binding on parties which were originally non- signatories. One tentative interpretation is that such extension mechanisms are conducive to greater wage rigidity, thereby lengthening the adjustment process. Finally, unemployment benefit replacement rates are not found to bear any significant impact on output gap fluctuations (Table 2.4, Column 1).28 [Table 2.4. Output gap equations with labour and product market regulation indicators, 20 OECD countries, 1982-2003] 26. Also apparent from [2.2] is the fact that policies and institutions enter the estimated equation in deviations from their sample means. One implication is that ηφ provides a measure of output gap persistence in the “average” OECD country. 27. The multicollinearity issue may not be extremely severe, however. The condition number of matrix X’X, where X is a (6 x 22) matrix containing all five policy and institutional variables above and the unit vector as column vectors, is about 9. Only values in excess of 20 have been suggested as indicative of an important multicollinearity problem (Belsley et al., 1980). 28. Splitting the average unemployment benefit replacement rate into its initial replacement rate and benefit duration components yields similar results. 12 household mortgage debt -- and the strength of monetary transmission channels more broadly – would be expected to improve resilience mainly under flexible exchange rates, in practice no significant interaction was found here between these two variables. 2.4 Sensitivity analysis Using alternative output gap measures 39. The previous empirical findings are derived from OECD output gap estimates, and as such they may be sensitive to the specific methods used to produce these output gap estimates. In particular, the use of filtering methods could produce some correlation between explanatory variables (lagged output gaps) and residuals, thereby leading to biased coefficient estimates. Against this background, sensitivity analysis is carried out using three alternative measures of the output gap: the Hodrick-Prescott and Baxter-King filter estimates presented in Section 1 as well as unemployment gap estimates, with structural unemployment being derived from the panel estimation of a standard model of equilibrium unemployment, without the use of any filtering method (see Box 3 for details). Box 3. Computing unemployment gaps The unemployment gap is defined as the gap between actual and structural unemployment. Therefore, computing unemployment gap estimates requires an estimate of structural unemployment. The latter is obtained here through panel data estimation of a theoretical model of unemployment. In practice, the following reduced-form unemployment equation is estimated, consistent with a variety of theoretical models of labour market equilibrium, including standard job-search (Pissarides, 2000) and wage-setting/price-setting (e.g. Layard et al., 1991; Nickell and Layard, 1999) models: [2.3] itiit j j itjit GXU εδχβ +++= ∑ where Uit is the aggregate unemployment rate, δi is a country fixed effect,1 and Git is a cyclical variable which aims to control for the unemployment effects of aggregate demand fluctuations over the business cycle. Here, the contemporaneous GDP growth rate and six lags of it2 are used. The Xj’s are policies and institutions which theory suggests may affect structural unemployment,3 namely: the unemployment benefit replacement rate, EPL, PMR, the degree of centralisation/co-ordination of wage bargaining,4 the tax-wedge between labour cost and take-home pay5 and union density.6 Estimating variants of equation [2.3] has become mainstream in the macroeconomic literature on the determinants of structural unemployment (see e.g. Bassanini and Duval, 2006; Belot and Van Ours, 2004; Blanchard and Wolfers, 2000; Nickell et al., 2005). Estimates of equation [2.3] are presented in Table 2.7. The model is first estimated with all policies and institutions (Column 1). The unemployment benefit replacement rate, the labour tax wedge and product market regulation appear to increase structural unemployment, while a high degree of corporatism reduces it. By contrast, EPL and union density are not found to have any impact on structural unemployment, which in the case of EPL is consistent most of the theoretical and empirical literature. Both variables are thus dropped from the analysis to obtain a streamlined equation (Column 2). Based on the latter estimates, one can then compute a measure of the unemployment gap as: [2.4] iti j j itjitit UXUUUgap −+=−= ∑ )()( * δβ [Table 2.7. Structural unemployment econometric estimates, 20 OECD countries, 1982-2003] With these unemployment gap estimates in hand, the output gap equations of Section 2.3 can be re-estimated in order to check the robustness of the findings.7 15 ________ 1. The inclusion of country effects -- which are found to be jointly significant -- aims to control for omitted, country-specific determinants of structural unemployment. 2. Starting from a model with 10 lags, insignificant lags were eliminated sequentially until all remaining lags were found to be significant. 3. See e.g. Bassanini and Duval (2006), Nickell and Layard (1999), Nickell (1997, 1998), Pissarides (2000). 4. In line with Bassanini and Duval (2006), its influence is captured in Table 2.7 through a dummy for “high corporatism” -- instead of the “low corporatism” dummy used above. 5. See data appendix for details on sources and methods. 6. Union density, which is defined as the rate of union membership (see data appendix for details), aims to capture union power. While the rate of collective bargaining coverage would be arguably a better proxy -- which is why it was used in the econometric analysis above, it is not available over the whole sample for most OECD countries and therefore cannot be used here. 7. Ideally, one would rather estimate a dynamic unemployment equation in one step, identifying simultaneously the policy determinants of equilibrium unemployment and those of short-run unemployment dynamics. However, compared with the two- step estimation approach followed here, the large number of additional parameters to be estimated would imply a sizeable loss in the number of degrees of freedom and would make it more difficult for the non-linear estimation procedure to converge. 40. Table 2.8 presents the re-estimation of three key equations -- the model selected from the “statistical tournament” on labour and product market policies (Table 2.4, Column 6), the model using the synthetic indicator of labour and product market regulation (Table 2.5, Column 1) and the model incorporating the synthetic indicator of labour and product market regulation and household mortgage debt (Table 2.5, Column 5) -- using the three alternative gap estimates. The main conclusion is that the findings are reasonably robust to the method used to construct the gaps. The only two noticeable differences with respect to previous results are the following: i) Labour and product market regulation is no longer found to mitigate the initial impact of shocks when unemployment gap estimates are used; household mortgage debt no longer appears to raise gap persistence when Baxter-King filter estimates are used. [Table 2.8. Equations with alternative output gap definitions] Incorporating interactions between institutions and observed shocks 41. All econometric estimates from Section 2.3 are based on equation [2.2], which focuses on interactions between institutions and common, unobserved shocks. One potential issue with such estimates is the omission of interactions between institutions and country-specific shocks. Such omission is a potential source of estimation bias insofar as institutions that shape the propagation of common shocks would also be expected to influence the propagation of country-specific shocks. 42. In order to check whether this issue affects the estimates, the three key equations are re-estimated by adding several macroeconomic variables, or “observed shocks”, to the set of unobserved shocks (Table 2.9). Concretely, the following equation is estimated: iti k kk it k h hh it h t itit j jj it j it XXZZ GAPGAPXXGAP εαγχλ ηϕϕ ++⎟ ⎠ ⎞ ⎜ ⎝ ⎛ −+−++ −⎟⎟ ⎠ ⎞ ⎜⎜ ⎝ ⎛ −+= ∑∑ ∑ −− )(1))(( )()( .... 21.. [2.4] where the Zhs are the “observed shocks” to be interacted with policies and institutions. 16 43. In line with recent empirical literature (Bassanini and Duval, 2006; Blanchard and Wolfers, 2000; Nickell et al., 2005), four types of “observed shocks” Zh are considered for analysis (see data appendix for definitions and methodological details): i) total factor productivity (TFP) shocks; ii) terms of trade shocks; iii) labour demand shocks; and, iv) real interest rate shocks, defined as the difference between the 10-year nominal US government bond yield and annual US GDP price inflation. These are country-specific observed shocks, except for the last one which is a common observed shock in order to avoid endogeneity with respect to the output gap. As shown in Table 2.9, the main findings are robust to the use of both observed and unobserved shocks in the estimated equation. [Table 2.9. Output gap equations with both observed and unobserved shocks] Controlling for fiscal policy 44. Output response to shocks depends on both existing institutional settings in labour, product and financial markets and the monetary and fiscal policy reactions. Equation [2.2] focuses only on the former, based on the implicit assumption that monetary and fiscal policy is not exogenous but rather is shaped by the institutional framework. Assuming that monetary policy reaction to shocks is entirely driven by existing nominal and real rigidities may not be implausible.35 By contrast, the assumption that fiscal policy responds optimally to shocks is arguably a stronger one, for at least two reasons. First, as noted earlier, the response of the fiscal balance depends partly on automatic stabilisers, which vary across countries and are only partially captured by the structural policy indicators in equation [2.2]. Second, the discretionary fiscal policy reaction is likely to be shaped by a wide range of considerations in practice. For these reasons, it can not be ruled out that estimates of equation [2.2] might suffer from an omitted variable bias. 45. A limited attempt to tackle this issue is made here by re-estimating the three key equations with the share of overall tax receipts in GDP as an additional institutional variable to be interacted with shocks (Table 2.10). This variable directly captures the size of automatic stabilisers and would therefore be expected to dampen the initial impact of shocks. No attempt is made here at addressing its potential endogeneity, however.36 Therefore, the results from Table 2.10 should not be seen as an attempt to study the role of fiscal policy for resilience but rather as some robustness check for previous findings. The latter are found to be robust to such sensitivity analysis. [Table 2.10. Output gap equations with control for fiscal policy] 2.5 Assessing the overall degree of resilience of OECD countries 46. What the previous analysis holds for the analysis of the policy and institutional determinants of resilience is somewhat ambiguous. Overall, strict labour and product market regulations appear to reduce resilience to shocks by increasing output gap persistence. Strict mortgage regulations have a similar -- albeit somewhat less robust -- effect. At the same time, there seems to be an offsetting effect insofar as strict labour and product market regulations improve resilience by cushioning the initial impact of shocks in most specifications. 47. In order to determine which of these offsetting effects dominates in practice, it is possible to devise a number of resilience criteria, and then to simulate the “preferred” equation of Table 2.5 (Column 5) for different values of policy and institutional indicators to see how the latter affect the score 35. As discussed in Box 1, this assumption holds under optimal monetary policy. 36. One way at least to mitigate the endogeneity issue is to consider the country average of the fiscal policy variable over the sample period. In practice, however, using this time-invariant variable does not change the conclusions from Table 2.10. 17 • Household mortgage debt, which bears a negative relationship with mortgage market regulation and can be seen as an indicator of financial market flexibility, seems to reduce output gap persistence. This is consistent with the view that monetary policy transmission channels are stronger in less regulated financial markets. • Most of the above findings appear to be reasonably robust to the use of alternative output gap measures, to the analysis of both observed and unobserved shocks and to some control for the fiscal policy stance. 53. Insofar as strict labour and product market regulations may dampen the initial impact of shocks but make it more persistent, the implications for resilience are a priori unclear. In order to determine which of these offsetting effects dominates in practice, the preferred equation was simulated so as to assess the impact of policies and institutions on different resilience criteria. The simulations indicate that rigid labour and product markets lengthen the time it takes for output to return to potential following a shock and increase the cumulative output loss incurred over the period. However, economies with more flexible labour and product markets appear to exhibit greater output gap volatility. By contrast, whatever the resilience criterion used, strict mortgage market regulation reduces resilience by increasing output gap persistence. 54. Simulations of the preferred equation were also run in order to see how individual OECD countries are expected to score on these resilience criteria, based on their most recent policy and institutional settings. This exercise points to three main groups of countries: • English-speaking countries with flexible labour and product markets and well-developed mortgage markets, where the time needed for output to get back to potential in the aftermath of a shock and the cumulative output loss are estimated to be among the lowest across the OECD, but where output gap volatility is comparatively high. • Some small European countries with moderately stringent labour and product market regulation and well-developed mortgage markets, which according to the simulations perform relatively well on all of the resilience criteria. To a lesser extent, Germany also falls into this group. • Most other continental European countries with comparatively strict labour and product market regulation and less-developed mortgage markets, which are estimated to be less resilient on all accounts. 55. The main purpose of this paper was to explore the reasons why in recent years a number of English-speaking and Nordic countries seem to have been more resilient than most continental European countries to a range of otherwise fairly similar shocks. Overall, the evidence linking resilience to rigidities in labour and product markets is somewhat ambiguous -- both theoretically and empirically. By contrast, the analysis tentatively suggests that policy settings strengthening monetary policy transmission -- such as the low degree of mortgage market regulation prevailing within the group of resilient countries -- are unambiguously good for resilience. This result is in line with the conclusions drawn by Cotis and Coppel (2005) who, based on an analysis of recent cross-country business cycle patterns, suggest that financial market flexibility has been key to the strong resilience of certain OECD countries. However, it remains tentative insofar as it is not based on a detailed analysis of the policy and institutional determinants of monetary policy transmission. Furthermore, the univariate approach adopted here could only to a limited extent distinguish between different kinds of shocks hitting economies. It is possible that such a distinction could overturn some of the above conclusions. As well, the specifications assume symmetry in the way structural policies affect response to positive and negative shocks. In practice, however, effects may well differ depending on the direction of a shock. There is therefore ample scope for further research. 20 21 BIBLIOGRAPHY Ahrend, R., P. Catte and R. Price (2006), “Interactions between Monetary and Fiscal policy: How Monetary Conditions Affect Fiscal Consolidation”, OECD Economics Department Working Paper No. 521. Altissimo, F., M. Ehrmann and F. Smets (2006), “Inflation Persistence and Price-Setting Behaviour in the Euro Area: A Summary of the IPN Evidence”, ECB Occasional Paper No. 46. Álvarez, L., E. Dhyne, M. Hoeberichts, C. Kwapil. H. Le Bihan, P. Lünnemann, F. Martins, R. Sabbatini, H. Stahl, P. Vermeulen and J. Vilmunen (2006), “Sticky Prices in the Euro Area: A Summary of New Micro-Evidence”, Journal of the European Economic Association, Vol. 4, No. 2-3. Artis, M. (2005), “Business cycle affiliations and their determinants: where do we stand?”, European Economy, Economic Paper 227, Proceedings of the 2004 DG ECFIN Research Conference on “Business Cycles and Growth in Europe”. Bassanini, A. and R. Duval (2006), “Reassessing the Role of Policies and Institutions for Labour Market Performance: A Consolidated Analysis”, OECD Economics Department Working Paper No. 486. Belot, M. and J. van Ours (2004), “Does the Recent Success of Some OECD Countries in Lowering their Unemployment Rates Lie in the Clever Design of their Labour Market Reform?”, Oxford Economic Papers, Vol. 56, No. 4. Belsley, D., E. Kuh and R. Welsch (1980), Regression Diagnostics: Identifying Influential Data and Sources of Collinearity, Wiley Series in Probability and Statistics, Wiley. Bergman, M. (2006), “How Similar are European Business Cycles?”, in Mazzi G. and G. Savio (eds), Growth and Cycle in the Euro-zone, Palgrave, MacMillan. Blanchard, O. (1999), “European Unemployment: The Role of Shocks and Institutions”, Baffi Lecture, Banca d’Italia, Rome, January. Blanchard, O. (1998), “Revisiting European Unemployment: Unemployment, Capital Accumulation and Factor Prices”, NBER Working Paper No. 6566. Blanchard, O. and J. Galí (2005), “Real Wage Rigidities and the New Keynesian Model”, Federal Reserve Bank of Boston Working Paper No. 05-14. Blanchard, O. and L. Summers (1986), “Hysteresis and the European Unemployment Problem”, in S. Fischer (ed.), NBER Macroeconomics Annual, Vol. 1. Blanchard, O. and J. Wolfers (2000), “The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence”, The Economic Journal, Vol. 110, No. 462. Camacho, M., G. Perez-Quiros and L. Sainz (2006), “Are European Business Cycles Close Enough to be Just One?”, Journal of Economic Dynamics and Control, Vol. 30, No. 9-10. 22 TABLES AND FIGURES Table 1.1. Size of idiosyncratic relative to common fluctuations Standard deviation of idiosyncratic component relative to standard deviation of common component HP output gap, 1970-2006 BP output gap, 1970-2006 OECD output gap, 1980-2006 France 0.8 0.7 1.0 Belgium 0.8 0.8 0.8 Germany 0.9 0.8 2.4 Austria 1.1 0.9 1.3 Spain 1.1 1.0 1.3 Italy 1.1 1.0 0.7 Netherlands 1.2 0.9 1.1 Canada 1.3 1.3 1.2 United Kingdom 1.3 1.3 1.4 Japan 1.3 1.3 1.8 Denmark 1.4 1.3 1.4 Sweden 1.4 1.4 1.3 United States 1.4 1.4 1.0 Switzerland 1.4 1.4 1.1 Australia 1.5 1.4 1.2 Norway 1.6 1.5 2.1 Ireland 1.6 1.6 1.8 Portugal 2.0 1.9 2.5 Finland 2.0 1.9 2.8 Iceland 2.5 2.2 2.4 Korea 2.7 2.4 2.5 Greece 2.9 2.2 1.7 New Zealand 3.2 2.6 1.9 Average 1.6 1.4 1.6 Source: OECD Economic Outlook 80 database and calculations. 25 Table 1.2. Idiosyncratic relative to common fluctuations over time Standard deviation of idiosyncratic component relative to standard deviation of common component HP output gap BP output gap 1973-1989 1990-2006 1973-1989 1990-2006 France 0.7 0.8 0.7 0.7 Belgium 0.9 0.7 0.9 0.6 Germany 0.7 1.1 0.6 1.0 Austria 1.0 0.9 0.9 0.9 Spain 1.0 1.0 1.0 0.9 Italy 1.1 0.8 1.0 0.8 Netherlands 1.2 0.9 0.9 0.8 Canada 1.2 1.3 1.1 1.3 United Kingdom 1.4 1.0 1.3 1.0 Japan 1.1 1.4 1.0 1.5 Denmark 1.5 1.1 1.5 0.9 Sweden 1.5 1.0 1.4 1.0 United States 1.4 1.2 1.4 1.2 Switzerland 1.6 1.0 1.6 1.0 Australia 1.4 1.4 1.4 1.4 Norway 1.6 1.4 1.6 1.3 Ireland 1.5 1.8 1.6 1.5 Portugal 2.1 1.7 2.1 1.6 Finland 1.8 2.1 1.7 2.1 Iceland 1.9 2.8 2.0 2.2 Korea 2.3 3.3 2.3 3.4 Greece 3.2 2.1 2.4 1.2 New Zealand 3.4 2.1 2.7 2.0 Average 1.5 1.4 1.4 1.3 Source: OECD Economic Outlook 80 database and calculations. Table 1.3. Idiosyncratic relative to common fluctuations in the euro area Standard deviation of idiosyncratic component relative to standard deviation of common component HP output gap BP output gap 1973-1989 1990-2006 1973-1989 1990-2006 France 0.5 0.5 0.3 0.3 Belgium 0.7 0.5 0.4 0.3 Germany 0.7 0.7 0.4 0.4 Austria 0.9 0.6 0.4 0.3 Spain 0.9 0.5 0.5 0.3 Italy 0.9 0.6 0.5 0.4 Netherlands 1.2 0.5 0.6 0.2 Ireland 1.3 1.2 0.7 0.6 Portugal 1.7 0.9 1.0 0.6 Finland 1.6 1.6 1.1 1.3 Greece 2.7 1.5 1.5 0.6 Average 1.2 0.8 0.7 0.5 Source: OECD Economic Outlook 80 database and calculations. 26 Data before 1991 refer to Western Germany. BK gap data for Korea start only in 73:1. BK data stop in 05:4 due to the filtering method. LUX, MEX, TUR and Eastern European countries excluded. Source: OECD Economic Outlook 80 database and authors' calculations. Figure 1.1. Cyclical divergence accross 23 OECD economies, 1970-2006 Standard deviation of unweighted output gaps 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 1970 75 80 85 90 95 2000 05 HP GAP BK GAP OECD GAP 27 LUX, MEX, TUR and Eastern European countries excluded. Source: OECD Economic Outlook 80 database and authors' calculations. B. 1990-2006 Figure 1.4. Cyclical correlation with the euro area average and the US Correlation coefficient for quarterly HP output gaps A. 1973-1989 Correlation with euro area PRT NLD ITA IRL GRC FRA FIN ESP DEU BEL AUT USA SWE NZL NOR KOR JPN ISL GBR DNK CHE CAN AUS -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Correlation with the US Euro area countries Other OECD countries Correlation with euro area PRT NLD ITA IRL GRC FRA FIN ESP DEU BEL AUT USA SWE NZL NORKOR JPN ISL GBR DNK CHE CAN AUS -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Correlation with the US Euro area countries Other OECD countries 30 Data before 1991 refer to Western Germany. BK gap data for Korea start only in 73:1. LUX, MEX, TUR and Eastern European countries excluded. Source: OECD Economic Outlook 80 database and authors' calculations. Cycles are defined as lasting five quarters at least. Turning points require that the upturn or downturn lasts at least over two subsequent quarters. Figure 1.5. Business cycle turning points accross 23 OECD economies Average number of turning points over current and previous three quarters HP output gap measure BK output gap measure 6 5 4 3 2 1 0 1 2 3 4 5 6 1970 75 80 85 90 95 2000 05 Peaks Troughs 6 5 4 3 2 1 0 1 2 3 4 5 6 1970 75 80 85 90 95 2000 05 Peaks Troughs 31 Data before 1991 refer to Western Germany. LUX eand SI excluded. Source: OECD Economic Outlook 80 database and authors' calculations. BK output gap measure Cycles are defined as lasting five quarters at least. Turning points require that the upturn or downturn lasts at least over two subsequent quarters. Figure 1.6. Business cycle turning points accross euro members Average number of turning points over current and previous three quarters HP output gap measure 3 2 1 0 1 2 3 1970 75 80 85 90 95 2000 05 Peaks Troughs 3 2 1 0 1 2 3 1970 75 80 85 90 95 2000 05 Peaks Troughs 32 Persistence of shocks:1 coefficient Implied half-life of output gaps (in years) Amplification of shocks:1 coefficient Estimate for the US: USA 0.44 1.67 0.41 Estimates for other OECD countries and test for statistical differences in coefficients with respect to the US: AUS 0.34 1.3 0.24 * * AUT 0.47 1.8 -0.85 *** BEL 0.45 1.7 -0.25 * CAN 0.27 1.1 0.34 ** ** CHE 0.41 1.6 -0.16 * DEU1 0.42 1.6 -0.15 DNK 0.31 1.2 -0.49 * * ** ESP 0.54 2.3 -0.41 ** ** ** FIN1 0.49 2.0 0.24 FRA 0.50 2.0 -0.53 *** GBR 0.41 1.6 -0.30 * IRL 0.49 1.9 0.37 ITA 0.47 1.9 -0.37 ** JPN 0.50 2.0 -0.57 *** NLD 0.50 2.0 -0.55 ** NOR 0.57 2.5 -0.97 *** *** *** NZL 0.38 1.4 -0.61 *** PRT 0.56 2.4 -0.38 *** *** ** SWE1 0.40 1.5 0.18 Time dummies yes Observations 434 R-squared 0.85 Non-linear least squares. * (**, ***): estimated coefficient differs significantly from corresponding coefficient obtained for the US at the 10% (5%, 1%) level. 1. (weighted) average over periods 1982-1990 and 1993-2003. Source : Authors' estimates. Table 2.1. Output gap equations with country dummies (20 OECD countries, 1982-2003) iηϕ iγ 35 Persistence of shocks: coefficient Amplification of shocks: coefficient Benefit replacement rate 0.12 -0.39* EPL for regular contracts 0.62*** -0.43** PMR 0.58*** -0.46** Collective bargaining coverage 0.29 -0.23 Low corporatism -0.52*** 0.54*** Labour and product market regulation (synthetic indicator) 0.5** -0.51** Source : Authors' estimates on the basis of country-specific persistence and amplification coefficients estimated in Table 2.1 and data sources described in the data appendix. coefficients and labour/product market policy indicators Table 2.2. Cross-country correlation coefficients between persistence/amplification (based on simpleregressions of country-specific coefficients on the average value of each policy indicator over the period 1982-2003, using bootstrapped critical values to assess statistical significance) iγiϕ 36 Table 2.3. Correlation coefficients between labour and product market regulation indicators Benefit replacement rate EPL PMR Collective bargaining coverage1 Low corporatism Benefit replacement rate 1 EPL 0.29 1 PMR 0.15 0.37 1 Collective bargaining coverage1 0.52 0.44 0.40 1 Low corporatism -0.53 -0.57 -0.38 -0.48 1 1. time-invariant indicator (country average over the period 1980-2000) Source : Authors' estimates on the basis of data sources described in the data appendix. Correlation coefficients, 1982-2003 37 Persistence of shocks: coefficient ξ i Amplification of shocks: coefficient ξi Household mortgage debt -0.38* 0.03 Intermediation of financial system 0.36* 0.11 Flexible exchange rate regime -0.43** 0.32 Source : Authors' estimates on the basis of country-specific persistence and amplification coefficients estimated in Table 2.1 and data sources described in the data appendix. coefficients and monetary/financial variables Table 2.6. Correlation coefficients between country-specific persistence/amplification based on simpleregressions of country-specific coefficients on the average value of each indicator over the period 1982-2003, using bootstrapped critical values to assess statistical significance) 40 1 2 Full model with all policies and institutions (dependent variable: unemployment rate) Final model selected after dropping insignificant variables (dependent variable: unemployment rate) Policies and institutions: Benefit replacement rate 0.064 0.072 [3.11]*** [4.14]*** Labour tax wedge 0.142 0.139 [5.20]*** [5.12]*** High corporatism -1.203 -1.590 [2.89]*** [4.11]*** PMR 0.711 0.661 [6.31]*** [8.97]*** Union density -0.031 [1.36] EPL 0.259 [0.85] Cyclical controls: GDP growth (t) -0.164 -0.150 [4.14]*** [3.74]*** GDP growth (t-1) -0.305 -0.300 [7.33]*** [7.19]*** GDP growth (t-2) -0.246 -0.237 [6.02]*** [5.68]*** GDP growth (t-3) -0.256 -0.252 [6.26]*** [6.10]*** GDP growth (t-4) -0.139 -0.131 [3.15]*** [2.81]*** GDP growth (t-5) -0.134 -0.132 [4.04]*** [3.84]*** GDP growth (t-6) -0.129 -0.123 [3.73]*** [3.53]*** Country fixed effects yes yes Time dummies no no Observations 434 434 R-squared 0.98 0.98 Ordinary least squares (Within estimates). Absolute value of t statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. Source : Authors' estimates based on Bassanini and Duval (2006). Table 2.7. Structural unemployment econometric estimates (20 OECD countries, 1982-2003) 41 1 2 3 Model selected from statistical tournament Model with synthetic indicators of labour and product market regulation alone Model with synthetic indicators of labour and product market regulation and monetary factors Persistence coefficients: 0.995 0.983 0.975 [22.75]*** [22.08]*** [21.09]*** 0.430 0.408 0.394 [10.43]*** [9.52]*** [8.88]*** Effect of institutions on persistence: Labour and product market regulation2 0.094 0.090 [4.14]*** [3.86]*** Household mortgage debt2 -0.658 [2.36]** EPL 0.145 [4.84]*** Effect of institutions on amplification of shocks: Labour and product market regulation2 -0.127 -0.126 [2.76]*** [2.80]*** Household mortgage debt3 PMR -0.452 [5.90]*** Country fixed effects yes yes yes Time dummies yes yes yes Observations 431 431 409 R-squared 0.81 0.80 0.80 Non-linear least squares. Absolute value of t statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. 1. See Section 1.1 in the main text for the definition of the Hodrick-Prescott-filtered output gap. 2. synthetic indicators calculated as the first component of a factor analysis performed on the following set of policies and institutions: unemployment benefit replacement rate, EPL, corporatism regime, collective bargaining coverage and PMR. 3. time-invariant indicator (country average over the period 1990-2002) Source : Authors' estimates. Table 2.8. Equations with alternative output gap definitions: Hodrick-Prescott filter gap1 estimates (20 OECD countries, 1982-2003) η φ jϕ kγ 42 1 2 3 Model selected from statistical tournament Model with synthetic indicators of labour and product market regulation alone Model with synthetic indicators of labour and product market regulation and monetary factors Persistence coefficients: 1.203 1.229 1.221 [35.54]*** [36.87]*** [34.85]*** 0.277 0.278 0.271 [10.38]*** [10.64]*** [9.76]*** Effect of institutions on persistence: Labour and product market regulation1 0.028 0.029 [2.07]** [2.05]** Household mortgage debt2 -0.322 [1.80]* EPL 0.066 [3.72]*** Effect of institutions on amplification of shocks: Labour and product market regulation1 -0.070 -0.063 [2.96]*** [2.60]*** Household mortgage debt2 PMR -0.074 [2.06]** Country fixed effects yes yes yes Time dummies yes yes yes Observations 394 394 376 R-squared 0.92 0.92 0.92 Non-linear least squares. Absolute value of t statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. 1. synthetic indicators calculated as the first component of a factor analysis performed on the following set of policies and institutions: unemployment benefit replacement rate, EPL, corporatism regime, collective bargaining coverage and PMR. 2. time-invariant indicator (country average over the period 1990-2002) Source : Authors' estimates. Table 2.9. Output gap equations with both observed and unobserved shocks (20 OECD countries, 1982-2003) η φ jϕ kγ 45 1 2 3 Model selected from statistical tournament Model with synthetic indicators of labour and product market regulation alone Model with synthetic indicators of labour and product market regulation and monetary factors Persistence coefficients: 1.052 1.021 1.017 [22.61]*** [21.47]*** [20.67]*** 0.400 0.380 0.379 [9.42]*** [8.49]*** [8.19]*** Effect of institutions on persistence: Labour and product market regulation1 0.100 0.080 [3.49]*** [2.63]*** Household mortgage debt2 -0.508 [1.80]* EPL 0.128 [3.53]*** Tax receipts as a share of GDP 0.044 -0.406 -0.064 [0.13] [1.01] [0.14] Effect of institutions on amplification of shocks: Labour and product market regulation1 -0.173 -0.184 [2.49]** [2.61]*** Household mortgage debt2 PMR -0.408 [3.90]*** Tax receipts as a share of GDP -0.906 0.040 0.504 [0.87] [0.03] [0.41] Country fixed effects yes yes yes Time dummies yes yes yes Observations 395 395 374 R-squared 0.82 0.83 0.83 Non-linear least squares. Absolute value of t statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. 1. synthetic indicators calculated as the first component of a factor analysis performed on the following set of policies and institutions: unemployment benefit replacement rate, EPL, corporatism regime, collective bargaining coverage and PMR. 2. time-invariant indicator (country average over the period 1990-2002) Source : Authors' estimates. Table 2.10. Output gap equations with control for fiscal policy (20 OECD countries, 1982-2003) η φ jϕ kγ 46 Source: Authors' estimates. See main text for details. Figure 2.1. Cross-country comparison of the value of explanatory variables -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 U SA G BR C AN N ZL AU S JP N C H E D EU D N K IT A ES P SW E N O R IR L BE L FI N AU T N LD FR A PR T Synthetic indicator of labour and product market regulation, 2003 (number of standard deviations around the 2003 OECD average) -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 IT A FR A AU T BE L FI N ES P IR L JP N SW E C AN PR T N O R AU S D EU N ZL U SA G BR D N K N LD C H E Household mortgage debt (number of standard deviations around the 2002 OECD average) 47 Source: Authors' estimates. See main text for details. Figure 2.2 (continued). Impulse-response functions (assuming a 1 percentage point common negative shock to the output gap, based on Table 2.4, column 5) Austria -1.6 -1.2 -0.8 -0.4 0.0 0.4 0 1 2 3 4 5 6 7 8 9 10 output gap years Finland -1.6 -1.2 -0.8 -0.4 0.0 0.4 0 1 2 3 4 5 6 7 8 9 1 output gap years 0 Netherlands -1.6 -1.2 -0.8 -0.4 0.0 0.4 0 1 2 3 4 5 6 7 8 9 10 output gap years Switzerland -1.6 -1.2 -0.8 -0.4 0.0 0.4 0 1 2 3 4 5 6 7 8 9 1 output gap years 0 50 Source: Authors' estimates. See main text for details. Figure 2.3. Simulated degrees of resilience according to three alternative criteria (based on Table 2.4, Column 5, using 2003 values of policy and institutional indicators) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 N LD C H E D N K PR T N O R D EU SW E IR L FI N ES P BE L FR A AU T N ZL AU S G BR JP N C AN U SA IT A Output gap volatility (number of squared standard deviations of the common shock, assuming there are no idiosyncratic shocks) 0.0 1.0 2.0 3.0 4.0 5.0 6.0 U SA C H E G BR N ZL C AN AU S D N K N LD JP N D EU N O R SW E IR L ES P PR T FI N BE L AU T FR A IT A Time T needed for output to get back to potential (in years, following a 1 percentage point negative common shock to output gaps) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 C H E N LD D N K G BR U SA N ZL D EU AU S PR T N O R C AN SW E JP N IR L ES P FI N BE L AU T FR A IT A Cumulative output loss between 0 and T (as a percentage of output, following a 1 percentage point negative common shock to output gaps) 51 DATA APPENDIX: DATA SOURCES AND METHODOLOGY OECD measure of the output gap: Definition: OECD measure of the gap between actual and potential output as a percentage of potential output, in %. The methodology followed by the OECD in order to estimate output gaps is based on a Cobb-Douglas production function with Harrod-neutral technological progress and is described in detail in Giorno et al. (1995). The approach may best be qualified as hybrid in the sense that it relies on both structural economic relationships to estimate NAIRUs (see Richardson et al., 2000) -- and univariate filters -- mostly HP filters, to estimate trend participation rates, trend hours worked and trend total factor productivity. Source: OECD, Economic Outlook 78, December 2005. Aggregate unemployment rate: Definition: unemployed workers as share of the labour force, in %. Source: OECD, Database on Labour Force Statistics; OECD, Annual Labour Force Statistics. Data adjustments: while the primary source is the OECD Database on Labour Force Statistics, Annual Labour Force Statistics -- which are usually available over longer time periods -- were also used in some cases to extrapolate unemployment rates backwards (under the assumption of similar percentage changes in unemployment rates in both sources). Unemployment benefit replacement rate: Definition: average unemployment benefit replacement rate across two income situations (100% and 67% of average production worker (APW) earnings), three family situations (single, with dependent spouse, with spouse in work) and three different unemployment durations (1st year, 2nd and 3rd years, and 4th and 5th years of unemployment). Source: OECD, Benefits and Wages. Data adjustments: original data are available only for odd years. Data for even years are obtained by linear interpolation. Employment Protection Legislation (EPL): Definition: OECD summary indicator of the stringency of Employment Protection Legislation. Source: OECD, Employment Outlook 2004. 52 Total factor productivity shock: Definition: deviation of the logarithm of Total Factor Productivity (TFP) from its trend calculated by means of a Hodrick-Prescott (HP) filter (smoothing parameter λ = 100). The calculation proceeds in three steps. First, growth in the Solow residual in the business sector is calculated as: ααα /)]log()1()log()log([)log( KNYTFP ∆−+∆−∆=∆ , where Y refers to real business sector GDP, N to total employment, K to the gross capital stock and α to labour income as a share of business sector income. Second, an index log(TFP) of the logarithm of TFP is obtained by cumulating the annual values of ∆log(TFP). Finally, the TFP shock variable is computed as the difference between log(TFP) and its HP filtered trend. Source: Bassanini and Duval (2006), Annex 2. Terms of trade shock: Definition: logarithm of the relative price of imports weighted by the share of imports in GDP, i.e. terms of trade shock = (M/Y) * log(PM / PY), where M and Y denote total imports and GDP in nominal terms, respectively, and (PM / PY) is the ratio of the deflator of total imports to the GDP deflator. Source: Bassanini and Duval (2006), Annex 2. Real interest shock: Definition: difference between the 10-year nominal US government bond yield (in %) and the annual change in the US GDP deflator (in %). Source: Bassanini and Duval (2006), Annex 2. Labour demand shocks: Definition: logarithm of the labour share in business sector GDP purged from the short-run influence of factor prices. The methodology follows Blanchard O. and J. Wolfers (2000). First, measures of real wages and employment in efficiency units are computed as Wefficiency units = (W / PY) / TFP and Nefficiency units = N * TFP, respectively. The simplest possible measure of labour demand shocks would be the negative of the sum of the logarithm of the ratio of labour input in efficiency units to real output in the business sector, on the one hand, and the logarithm of real wages in efficiency units, on the other hand: – [log(Nefficiency units / Y) + log(Wefficiency units)] = – log[(N * TFP) / Y] – log[(W / PY) / TFP] = –log [(W * N) / (PY * Y)] = – (labour’s share of business sector income). However, this simple measure of labour demand shocks is accurate only to the extent that the production function is Cobb-Douglas and factor proportions adjust instantaneously to changes in factor prices. Insofar as the latter assumption is unlikely to be verified in the short-run, changes in the labour share reflect both genuine labour demand shocks and the lagged adjustment of factor proportions to changes in factor prices. Therefore, it is necessary to purge the labour share from the short-run influence of factor prices. For simplicity and comparative purposes, this is done here by following the same methodology as Blanchard (1998). Concretely, a wage measure which takes into account the gradual adjustment of factor proportions is computed as: log(Wadjusted) = λ * log(Wadjusted) + (1–λ) * log(Wefficiency units), where the value of parameter λ is set equal to 0.8 in line with estimates on annual data provided by Blanchard. The labour demand shock is then constructed as – [log(Nefficiency units / Y) + log(Weadjusted)]. The negative sign implies that an increase in 55 this variable should be interpreted as an adverse labour demand shock. Finally, this variable is set equal to zero in 1970 (or in the first year of data availability for those countries where long time series are unavailable). Source: Bassanini and Duval (2006), Annex 2. Cyclically-adjusted primary fiscal surplus: Definition: Primary fiscal surplus adjusted for cyclical factors. Source: OECD, Economic Outlook 80. 56