SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

This study takes a fresh look at the nature of financial and real business cycles in OECD countries using annual data series and shorter quarterly economic indicators. It first analyses whether the last cycle has been different compared to previous cycles in terms of length, amplitude, asymmetry and changes of these parameters during expansions and contractions. We also study the degree of economic and financial cycle synchronization between OECD countries but also of economic and financial variables within a given country and gauge the extent to which cycle synchronization changed over time. We next describe the connection between the great moderation and the last cycle. Finally, the study discusses the synchronization between the real economy and the financial sector and provides some new evidence on the banking sector's pro-cyclicality by using aggregate and bank level. The main findings show that the amplitude of the real business cycle was becoming smaller during the great moderation, but asset price cycles were becoming more volatile. In part, this was linked to developments in the banking sector which tended to accentuate pro-cyclical behaviour. Greater synchronization of cycles may help explain the severity of the crisis.

I. Introduction

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

The recent economic and financial crisis has challenged some presumptions about the forces shaping economic cycles and the effectiveness of policy that had developed during the great moderation era. This crisis, inter alia, casts doubt over the understanding of cyclical developments and also the impact of financial markets on the cycle and the cycle on financial markets. This study examines the last cycle and highlights how the nature of business and asset price cycles have been changing over time. The length, size and asymmetry of expansionary and contractionary periods are studied using a variant of the procedure developed by Bry and Boschan (1971) to determine peaks and troughs in time series. The degree of business cycle synchronization is established on the basis of so-called concordance indices aimed at quantifying the degree of overlap of different cycles once turning points are identified.

A well-developed financial sector can play a smoothing role by allowing firms and households to smooth investment and consumption as well as by providing traction for monetary policy, but can also amplify shocks and at times be at the origin of economy-wide shocks. Indeed, capital, provisioning, liquidity and maturity mismatch in the banking sector can generate pro-cyclical behaviour in credit supply for a number of reasons including the regulatory setup, the nature of risk assessment and the prevailing incentives to take risks. In the available empirical work, there is little consensus on the degree of pro-cyclicality of the banking system. It is against this background that this study takes a new look at the pro-cyclicality of a large number of banking sector variables, including capital, liquidity and provisioning indicators by relying on dynamic panel regression analysis carried out for sector-specific and bank-level data.

The remainder of the study is structured as follows. 'The Nature of the Last Cycle' provides new evidence on the nature of real and financial cycles and compared previews cycles with the last cycle. 'The Last Cycle and the Great Moderation' discusses the link between the last cycle and the great moderation. 'The Last Cycle and the Financial Sector' analyses the extent of the pro-cyclicality of the banking sector, which might have been an important driver of the last cycle and the great moderation. 'Conclusions' finally provides some concluding remarks.

II The Nature of the Last Cycle

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

This section investigates whether the last cycle has been different than earlier cycles. For this purpose, we first set out the methodology aimed at capturing features of the cycle. We then present and discuss the results.

Determining turning points

The cycle is often determined by applying a standardized procedure to identify expansions or contractions.1 We use a variant of the procedure developed by Bry and Boschan (1971) to determine peaks and troughs in our series. We follow Avouyi-Dovi and Matheron (2005) and Everts (2007) by imposing the following rules:

  • A search is carried out in the series to pin down local minima and maxima in a window of t ± 2 for quarterly series and t ± 5 for monthly series.
  • No multiple consecutive peaks or troughs are allowed. In the occurrence of multiple peaks or troughs, the highest peak or lowest trough is selected and the rest eliminated.
  • A minimum length is imposed for peak-to-trough and trough-to-peak phases and for full peak-to-peak and trough-to-trough cycles. For monthly (quarterly) series, each phase has to be at least 5 (2) months (quarters) long and the cycle cannot be shorter than 15 (5) months (quarters).

The above algorithm is applied to the raw series but also to series filtered in two different ways to eliminate outliers and volatility stemming from high frequency (monthly) data. First, a moving average of 15 months and five quarters are applied to the monthly and quarterly series. Second, the series are filtered using a 5-point and 15-point Spencer curve for quarterly and monthly series, respectively, which is indeed a moving average with a special weighting scheme as follows: inline image with r = 2 and inline image for quarterly data (Everts, 2007) and with r = 7 and inline image for monthly data (Avouyi-Dovi and Matheron, 2005).

Cycle synchronization

A way to look at cycle synchronization is to analyse the degree of overlap of different cycles once turning points are identified. The so-called concordance index (Cxy), given below, takes the value of 1 if the two cycles overlap perfectly and is 0 if for instance series x is always in expansion at a time series y is in contraction (Avouyi-Dovi and Matheron, 2005; Harding and Pagan, 2006).

  • display math

Where St = 1 if Xt is in the phase of expansion and St = 0 if Xt is in the phase of contraction. To test the significance of the concordance index, Harding and Pagan (2006) suggest to estimate the following equation:

  • display math

Where β is the empirical correlation between Sx and Sy, and inline imageis the empirical standard deviation of S. Finally, Harding and Pagan (2006) show that if β = 0, the error term will suffer from serial correlation and therefore the equation needs to be estimated using a heteroscedasticity and autocorrelation consistent estimator.

The nature of the last cycle: length and size asymmetries

Secular time series for seven OECD countries (Canada, Germany, France, United Kingdom, United States, Japan and Sweden), for which both real GDP and real share prices are available for around 100 years, suggest that the average annual growth rate of real GDP was around 1% in the early 19th century, rising to almost 4% in the early 1970s and then moving to approximately 2% around the turn of the 21st century. The data indicate that not only has GDP growth volatility declined substantially from the 1930s onwards but also that today's low level of volatility is in line with limited volatility observed during the 19th century (Figure 1). By contrast, for real share prices, a rise in growth rates appears to be accompanied by increased volatility.

image

Figure 1. Changes in the cycle over the long-term (annual data, unweighted average of OECD countries).

Source: Calculations based on data obtained from Barro and Ursua (2008), ‘Macroeconomic Crises since 1870’, BPEA, Online Appendix.

Download figure to PowerPoint

Output (classical or level) cycles exhibit considerable length asymmetries between upswings and contractions. GDP downturns often last only a handful of quarters, while expansions typically persist for several years. Furthermore, output cycles tend to be asymmetric in size: the amplitude of the expansion is typically much larger than the contraction. Information on the last cycle was estimated separately to examine whether cyclical developments were unusual in historical context. Importantly, the last cycle was considerably more asymmetric compared to previous cycles, both in terms of length and size (Table 1).

Table 1. Output cycle asymmetries between expansions and downturns over the long run (annual data, unweighted average of OECD countries)
 All cycles excluding last cycleLast cycle onlyLast expansion relative to previous expansions
Length asymmetrySize asymmetryLength asymmetrySize asymmetryLength comparisonSize comparison
Notes:
  1. Length asymmetry, length in years of expansion/length in years of downturn. Size asymmetry, size of expansion/size of downturn. The last two columns compare the length and size of the last expansion relative to previous expansions. Source: Calculations based on data obtained from Barro and Ursua (2008), ‘Macroeconomic Crises since 1870’, BPEA, Online Appendix.

1790–2009
Level cycle2.74.28.228.82.91.6
Deviation cycle1.21.01.53.11.22.3
Growth cycle0.91.01.40.91.40.3
1946–2009
Level cycle7.338.78.228.81.61.2
Deviation cycle1.21.01.53.11.21.5
Growth cycle1.00.91.40.91.30.5

Changes in the nature of the cycle across countries may mask how different cycles are developing within a country. To give an example, in the US economy, since the beginning of the 1990s until 2006 the growth rate of real share and house prices was well above historical growth rates whereas at the same time economic growth was close to growth rates observed at the end of the 19th and mid-20th centuries (Table 2). The great moderation of volatility in economic growth appears unprecedented. Volatility in real share prices declined to levels observed in the 19th century, while volatility of real house prices increased well above levels seen in most of the 20th century.

Table 2. Changes in cycles in the United States over the long term
 1871–19141919–19391946–19711972–19901991–2006
Average growth rate
Real GDP1.631.091.612.271.82
Real share price2.695.994.841.567.70
Real oil price1.060.640.940.713.17
Real gold price0.413.80−2.2311.120.64
Real oil price0.57−0.130.3215.056.29
Standard deviation
Real GDP4.617.134.002.461.33
Real share price15.7026.4616.1716.6415.18
Real house price9.763.953.923.205.12
Real gold price2.4015.356.9137.1213.19
Real oil price27.6223.3511.5057.1824.47

Quarterly data for real output, real share prices, real house prices and real credit were used to calculate the length of the cycles and the asymmetries in the length and amplitude of the cycles for the period 1950–2008 or the longest available period for OECD countries. For most countries and variables when considering the classical cycles, the period from trough to peak and from peak to trough are more or less symmetric in length but the amplitude of the trough to peak is considerably larger for real output and real credit. The last cycle was longer, and its trough-to-peak phase was significantly longer than the peak-to-trough phase for real output, real credit and real house prices. The classical cycle for real share prices is typically shorter than the other types of cycles.2

Deviation and growth cycles tend to be shorter and more symmetric compared with the classical cycle. Measures of the deviation and growth rate cycles also show that the last trough-to-trough cycle was longer than previous cycles and exhibited more variation in the lengths of the time and amplitude from peak to trough and trough to peak.2 Note that the last observed cycle does not include the recent recession because of the insufficient number of observations to detect the local minimum for the current recession.

In comparison with real GDP, expansions of consumption are longer and shorter for investment. Also, the size asymmetry is more pronounced for consumption and less pronounced for investment, indicating less trend increase for the latter and perhaps more volatility. Share and house price cycles over the long run tend to be more symmetric. Long and large expansions of GDP are often accompanied by long and large expansions of private consumption and real house prices (Table 3).3

Table 3. Cycle asymmetries (1950 where available to 2009, level cycle, quarterly data, unweighted average of OECD countries)
 All cycles excluding last cycleLast cycle onlyLast expansion relative to previous expansions
Length asymmetrySize asymmetryLength asymmetrySize asymmetryLength comparisonSize comparison
Notes:
  1. Length asymmetry, length of expansion/length of downturn; size asymmetry, size of expansion/size of downturn.

  2. The last two columns compare the length and size of the last expansion relative to previous expansions.

Source: OECD calculations based on the OECD Economic Outlook 86 database and Datastream.
Real GDP6.314.111.532.52.53.2
Output gap1.41.02.14.21.20.8
Private consumption8.734.311.730.12.11.9
Investment2.44.05.111.61.31.0
Long-term interest rate0.91.01.61.51.50.7
Short-term interest rate1.01.01.65.32.00.7
Real short-term interest rate1.11.01.31.51.00.6
Government net lending1.21.11.77.41.21.0
Unemployment rate1.17.91.319.00.90.3
Stock market index1.22.34.110.41.91.6
Real house prices1.31.99.226.03.64.0

Overall, the nature of the cycle has changed over time, with the changes most pronounced for level cycles. For most OECD countries, output cycles have tended to become longer and more asymmetric with expansionary phases lasting longer, while the length of slowdown or contraction phases has remained approximately the same. This is widespread among different variables, with larger and longer expansions occurring for consumption, investment (including stockbuilding), as well as for share and house prices. In comparison with the average of previous expansions, the length of the latest expansion phase is about double for output, consumption and stock prices (10 years vs. 5 years), while it nearly quadrupled for house prices (almost 10 years vs. 2–3 years). The asymmetry of the size of the expansion in comparison with the contraction has also become more pronounced for level and deviation cycles but not for growth rate cycles. Another important feature for deviation and growth cycles has been the fall in the amplitude over time.

The synchronization of cycles

Synchronization, measured by the overlap of expansions and downturns of different variables with expansions and downturns of output within a country, shows marked differences across countries and sometimes there is only little synchronization. For instance, private consumption is highly synchronized with output in Canada, Japan, the United Kingdom and the United States, but not in France and Germany, while with the exception of Japan, the United Kingdom and the United States, house prices appear unsynchronized with GDP cycles.2 Rolling-window correlations show that the synchronization of GDP, real share and house prices became unprecedentedly strong during the last downturn compared with the previous 40 years (Figure 2).

image

Figure 2. 10-quarter rolling-window correlations of macroeconomic variables. Panel A – 10-quarter rolling-window correlations against the United States. Panel B – 10-quarter rolling-window correlations of macroeconomic variable pairs within a country. Note: Average/minimum/maximum is the unweighted average/lowest/highest correlation of individual OECD countries' variables against the corresponding US variable (Panel A) or of the variable pairs for each OECD country (Panel B).

Source: OECD calculations based on the OECD Economic Outlook 86 database and Datastream.

Download figure to PowerPoint

Let us now turn to cross-country synchronization. Output cycles have overlapped to a significant extent (Table 4). In particular, cycle synchronization appears strong for some country groups (for instance, among Germany, Austria and the Netherlands or the United States, the United Kingdom and Canada). Furthermore, synchronization has been higher in recent decades. The data also suggest that stock markets in OECD countries were highly synchronized over the last 40 years but that the cross-country correlation for real house prices was less pronounced and was limited to a subgroup of countries (United States, United Kingdom, Spain, France, the Netherlands, Norway and Sweden). Previous research – Duval et al. (2007), using a regression-based decomposition of output gap measures into common and idiosyncratic components – provide some evidence that synchronization across OECD countries may not have been strong, with the possible exception of euro area countries.4 Kose et al. (2008), on the other hand, found for a large sample of developed and developing countries that business cycles became more synchronized within groups of countries, and that global factors – though not group factors – declined in importance since the early 1980s for developed countries.

Table 4. Concordance of GDP cycles across countries (level cycle)
 Annual data, 1870–2009Quarterly data, 1970:q1–2008:q4
GermanyUnited StatesGermanyUnited States
Notes:
  1. Data start in 1870 except for Korea (1912), Mexico (1895) and Turkey (1923). The concordance index reported in this table takes the value of 1 if two cycles overlap perfectly and 0 if there is no overlap between the cycles.

  2. * and ** indicate statistical significance at the 10% and 5% levels. Source: Calculations based on data obtained from Barro and Ursua (2008), ‘Macroeconomic Crises since 1870’, BPEA, Online Appendix and OECD Economic Outlook 86 database.

Australia0.85**0.74**0.700.79**
Austria0.86**0.79**0.82**0.88
Belgium0.80**0.72**0.83**0.78
Canada0.78**0.86**0.80*0.97**
Denmark0.75**0.74**0.76**0.74
Finland0.81**0.67**0.78**0.79
France0.81**0.72**0.87**0.91**
Germany 0.76** 0.84**
Greece0.76**0.58**0.650.63
Iceland0.74**0.69**0.730.75
 
Ireland  0.81**0.91**
Italy0.77**0.73**0.82**0.81
Japan0.77**0.78**0.700.79
Korea0.77**0.78**0.720.82
Luxembourg  0.83**0.89*
 
Mexico0.81**0.71**0.720.83
The Netherlands0.81**0.71**0.86**0.90*
New Zealand0.66**0.67**0.710.81
Norway0.80**0.76**0.81**0.86
Portugal0.65**0.67**0.82**0.80
 
Spain0.78**0.79**0.84**0.86
Sweden0.71**0.72**0.770.83
Switzerland0.75*0.68**0.84**0.85**
Turkey0.78**0.76**0.64**0.76
United Kingdom0.79**0.86**0.82**0.90**
United States0.76** 0.84** 

However, when looking at time variations in cross-country co-movements based on rolling-window correlations, GDP and real house price growth became extremely strongly correlated with historical standards during the recent crisis. A similarly very strong synchronization of real share prices could be observed after the burst of the dot-com bubble (Figure 2). Also correlations between GDP growth, real house and share prices within countries were high in the run-up and during the crisis.5 The shocks originating from the United States in 2007 and 2008 were transmitted remarkably quickly to the rest of the world. Financial market integration (FDI) and trade openness were key elements of the rapid and strong transmission,6 magnified by intra-industry trade within subgroups of countries.

III The Last Cycle and the Great Moderation

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

This section considers the factors behind the great moderation, which while widespread did not affect all economies to the same extent. In the United States, the standard deviation of output growth and inflation declined considerably with a break occurring around the middle of the 1980s. Other OECD economies experienced similar declines in output and inflation volatility. On the other hand, not all countries enjoyed a great moderation, with output growth volatility even increasing in Iceland, while in some others (like France) output volatility was never pronounced. Four broad sets of explanations for the great moderation are advanced in the empirical literature: (i) better macroeconomic policy, (ii) good luck, (iii) structural changes in the economy, (iv) financial market deepening.

A number of findings suggest that better macroeconomic policy, particularly monetary policy, may have contributed to the great moderation. Output volatility is often correlated with the volatility of inflation, which is consistent with a story of better monetary policy (Blanchard and Simon, 2001). Monetary policy might have gained credibility because of institutional changes including greater central bank independence, the introduction of inflation targeting frameworks and a strong track record in fighting inflation, leading to a better anchoring of inflation expectations (Dalsgaard et al., 2002). Similarly, a shift in monetary policy may be related to the higher weight assigned to inflation in the monetary policymaker's objective function (Taylor, 1998). Even relatively small changes in policy rules and changes in the volatility of shocks can imply relatively large changes in the volatility of output and inflation (Canova, 2009).

A second explanation, not necessarily incompatible with better monetary policy, is good luck – in particular fewer large adverse shocks – contributing to reduced volatility.7 Some have argued that better inventory management may have contributed to the decline in aggregate volatility (Dalsgaard et al., 2002).8 However, this dynamic may only reflect smaller shocks hitting economies and other research finds that the great moderation has little to do with changes in inventory behaviour (Barrell and Gottschalk, 2004).

According to a third explanation, a large number of other changes in external conditions and the functioning of economies may have contributed to the great moderation. However, in some cases the beneficial effects were transient or ambiguous. First, the impact of globalization could either reduce or increase volatility. The rapid development of emerging economies has underpinned growth in the developed world. At the same time, cheap imported goods from China and other emerging economies have created a terms of trade gain for the advanced economies and thus a beneficial tailwind, which only turned into a headwind when rapid global growth led to sharp rise in oil and other commodity prices (Pain et al., 2006). On the other hand, greater trade and financial integration can make a country more exposed to external shocks. Second, the shift in the composition of output from manufacturing to services may have affected volatility (Black and Dowd, 2009). However, McConnell and Perez-Quiros (2000) and Stock and Watson (2003) demonstrate that the decline in volatility is common across sectors in the G7 countries. Finally, aggregation effects could have also played a role. Indeed, firm dynamics can exhibit distinct differences from aggregate developments. For example, firm-level output volatility increased, whereas aggregate volatility fell (Comin and Philippon, 2005). This could be related to developments in financial markets allowing riskier firms access to external finance (Buch et al., 2009) and the consequences of regulatory reform and technical progress leading to idiosyncratic or sector-specific shocks becoming more important and less correlated across firms and sectors (Stiroh, 2009). In addition, individual households have faced increased economic uncertainty during the great moderation in the United States, but the covariance across households has decreased, leading to smoother aggregate income developments (Dynan et al., 2006a).

Finally, financial market deepening and innovation have allowed greater consumption and investment smoothing, by allowing better risk diversification and inter-temporal smoothing (Blanchard and Simon, 2001; Catte et al., 2004; Dynan et al., 2006b; de Blas, 2009). Credit market liberalization may have helped the absorption of shocks, which may otherwise have shown up in higher inflation and growth volatility (Benk et al., 2009). However, the increasing importance of the financial sector also made the economy more vulnerable to adverse developments.

IV The Last Cycle and the Financial Sector

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

The interaction of the real economy and the financial sector

A long financial sector cycle, coupled with a greater synchronization between the real economy and the financial sector may have also played an important role in prolonging the last cycle and reducing volatility. The banking and financial sectors are indeed strongly intertwined with the real economy. Bank credit and the access to capital markets can amplify movements in the real economy (Dell'Ariccia et al., 2008). At the same time, cycles in the real economy can introduce cyclicality in bank lending.9 The demand for and the supply of bank loans and thus their cost fluctuate over the cycle because credit demand is related to production, business and residential investment (Ayuso et al., 2002), and because lending standards change over the cycle, being lax during expansions, but tight in downturns (Keys et al., 2010). Financial innovations in the US economy, mostly related to the housing sector, helped relax lending conditions longer than during previous booms, which fuelled the housing bubble. Ever rising house prices went in tandem with mortgage equity withdrawal, which financed private consumption and thus boosted the real economy.

Financial liberalization gave rise to more risk taking, especially during the great moderation period and resulted in higher leverage ratios10: banks had to increase leverage if they wanted to maintain the return on equity unchanged while having riskier clients and facing lower profits due to more intense competition (Goodhart et al., 2004). Moreover, the move from relationship banking to arm's length banking and the commodification of financial transactions and securitization increased the costs of monitoring and may have contributed to an underestimation of risks (Panetta et al., 2009). A pro-cyclical banking sector will in turn amplify the real cycle. Therefore, a policy design that reduces the banking sector's pro-cyclicality will help attenuate the real cycle.

Existing empirical work that analyses whether regulatory bank capital is pro- or counter-cyclical over the business cycle points to differences between the United States and Europe on the one hand and between the old and new EU member states on the other hand. Yet, it fails to provide a consensus on how banks react to the cycle in a given country group (Table 5). While multi-country panel studies suggest a weak counter-cyclical effect, country-specific studies for Spain, Germany, the United Kingdom and Norway come to opposite conclusions. This can be largely explained by the characteristics of the studies in terms of time span, data cleaning, country coverage, estimation method and the number of control variables.

Table 5. Literature overview on banking sector pro-cyclicality
StudyCountry coveragePeriodPro/counter-cyclicality
Regulatory capital
Ayuso et al. (2002)Spain1988–2000Pro
Stolz and Wedow (2005)Germany1995–2003Pro
Francis and Osborne (2009)United Kingdom1990:q1–2006:q4Weak Pro
Lindquist (2003)Norway1995:q1–2001:q4NO or weak Pro
Jokipii and Milne (2006)European Union1997–2004

EU-old: Pro

EU-new: Counter

Bikker and Metzenmakers (2007)29 OECD, EU, United States1992–2001

Loan growth: Pro

Macro cycle: EU-old: Counter

United States: Pro

Kim and Lee (2006)30 OECD, seven Asian1995–2004

OECD, United States: Counter

Asian countries: Pro

D'Avack and Levasseur (2007)11 Central and Eastern European countries1997–2005Pro
Loan loss provisioning
Bikker and Metzenmakers (2002)OECD, EU, United States, JPN, FRA, ITA1991–2001

Pro

ESP: NO

United Kingdom: Counter

Profitability
Beckmann (2007)16 Western European countries1979–2003Pro

Pro-cyclicality of the financial sector has increased recently

Fluctuations in bank assets relative to GDP have become more pronounced since the 1970s.11 Figure 3 shows that the deviation of the bank asset-to-GDP ratio from its trend and its percentage point changes exhibit greater volatility since the 1970s. The pronounced pro-cyclicality of the banking sector was initially triggered by the move from credit controls of the post-war period to more liberalized banking and financial sectors during the 1970s, while financial innovations played an important role later on. Goodhart et al. (2004) show that financial liberalization in OECD economies was followed by boom-bust cycles in bank lending, output and asset prices. They compare financial liberalization to a permanent productivity shock in a credit-constrained economy à la Kiyotaki and Moore (1997) in which a positive productivity shock increases the value of collateral that in turn raises the capacity to borrow, which boosts lending, investment and output until the boom turns into a bust.

image

Figure 3. Cycles in the real economy and the financial sector of OECD countries Unweighted average of OECD countries. Notes: OECD staff calculations based on data provided by Alan M. Taylor (Schularick and Taylor, 2012). The series plotted are arithmetic averages of individual series of the following countries: Australia, Canada, Switzerland, Germany, Denmark, Spain, France, United Kingdom, Italy, the Netherlands, Norway, Sweden, United States. GDP growth is the rate of growth of real GDP, deviation from trend of the bank asset/GDP ratio is the deviation of the bank asset/GDP ratio from its trend (trend is computed using the HP filter). The series are 3-year moving averages. Banking assets are defined as total domestic currency assets of banks and banking institutions.

Source: Schularick and Taylor (2012) and OECD Economic Outlook 86 database.

Download figure to PowerPoint

Examining changes in the relationship between the change of the bank asset-to-GDP ratio and the cycle over time (based on the data underlying Figure 3) shows that the banking system was counter-cyclical until the early 1970s (negative correlation between the percentage point change in the bank asset-to-GDP ratio and real GDP growth) and has become pro-cyclical only since the late 1970s with the estimated coefficient turning from negative to positive (Figure 4). It may come as a surprise that the coefficient estimate is not statistically significant for the most recent 12 year period, but this is the period when leverage exploded. The rise in pro-cyclicality correlates well with the number of banking crises for this country sample as reported by Reinhart and Rogoff (2008): No banking crisis occurred between 1945 and 1974, three countries experienced banking crises between 1974 and 1977 and 11 banking crises are identified between 1983 and 1995.

image

Figure 4. Pro-cyclicality of the banking sector: rolling window estimations. Notes: Coefficient estimates are displayed only if they are statistically significant. The estimations are performed using difference GMM. The percentage point change in the bank asset/GDP ratio is the dependent variable and GDP growth and lagged changes in percentage points of the bank asset-to-GDP ratio are the independent variables. The data points for bank asset-to-GDP ratio refer to the end of the period. The bank asset-to-GDP ratio is calculated as the unweighted average of the ratio of 13 OECD countries.

Source: Schularick and Taylor (2009) and OECD Economic Outlook 86 database.

Download figure to PowerPoint

A panel of banking sector-level and individual bank-level data is used to re-assess the pro-cyclicality of bank capital (Table 6). These estimates show how banking sector indicators have moved with the credit and output cycle. Results obtained using both aggregate and individual bank-level data sets suggest that from 1994 to 2007/2008, capital ratios, in particular the tier 1 ratio, the shareholder equity and the total equity/capital ratio have a negative association with loan growth. Country-specific estimations corroborate this result, though there are some exceptions. This pro-cyclical relationship also tends to hold, if first differences of ratios are used. The co-movements are less pronounced for the leverage ratios calculated from national balance sheets and when GDP growth is used to capture cyclical fluctuations. Bank capital ratios do not correlate with real house prices, but there is a negative correlation of leverage ratios and a weak link with the tier 2 capital ratio with real share prices.

Table 6. Panel estimation results for bank capital
 Loan growthGDP growthHouse price growthShare price growth
Notes:
  1. The results are obtained using a dynamic specification where the dependent variable is regressed on its lagged value and the measure of the cycle (real loan growth, GDP growth, real share and real house price growth). In Panel C, these results are obtained by interacting the cycle variable with country dummies. Country names are not shown if the coefficient estimates are not significant. Capital 1, 2, 3 and 4 are defined as follows. Capital 1 = tier1 capital over risk weighted assets, capital 2 = common shares over total assets, capital 3 = total equity over total assets, capital 4 = the sum of total capital and subordinated debt over total assets.

  2. * and **denote statistical significance at the 10% and 5% levels. Source: OECD calculations based on the OECD Economic Outlook 86 database, OECD Bank Profitability Database and Bankscope.

Panel A – Country-level data (Bank profitability database)
Dependent variableLevel equation
Tier 1 ratio−0.043**−0.229*0.007−0.004
Tier 2 ratio−0.0020.015−0.001−0.002
Leverage ratio (total capital/total assets)−0.019**−0.011−0.009−0.004
Leverage ratio (SNA) (total assets/total capital)0.030−0.314**−0.035−0.035**
 First difference equation
Tier 1 ratio−0.023−0.099-0.0030.002
Tier 2 ratio0.0000.040**0.005−0.002*
Leverage ratio (total capital/total assets)−0.014**−0.0280.001−0.004
Leverage ratio (SNA) (total assets/total capital)−0.029−0.343**−0.019−0.028**
Panel B – Bank-level data (Bankscope database)
Independent variableLoan growthLoan growthLoan growthLoan growth
Dependent variableCapital 1Capital 2Capital 3Capital 4
Level equation−0.002*−0.003**−0.009**−0.003
First difference equation−0.004−0.004**−0.004−0.005**
Panel C – Country-specific results
Independent variableLoan growthLoan growthLoan growthLoan growth
Dependent variableCapital 1Capital 2Capital 3Capital 4
Pro-cyclical (−)

BEL

CAN

CZE

DEU

DNK

ESP

FIN

PRT

SVK

SWE

DEU

DNK

FRA

HUN

MEX

NOR

POL

PRT

SVK

CH

CZE

DEU

DNK

FRA

HUN

MEX

NOR

POL

PRT

SVK

SWE

DEU

DNK

KOR

SWE

United States

Counter-cyclical (+)

CHE

FRA

ISL

NZLNZL

Estimations carried out with country-level and bank-level panel data sets concerning loan loss reserves, bad loan provisioning, the funding gap, various measures of bank profitability and bank liquidity all exhibit a strong pro-cyclical pattern. In addition, bank equity moves hand in hand with loan growth. This implies that deleveraging does not come about because loans drop while equity remains unchanged but because loans drop more than equity falls. Country-specific estimates broadly confirm the aggregate analysis (Table 7).

Table 7. Panel estimation results for other bank ratios
 Loan growthGDP growthHouse price growthShare price growth
Panel A – Country-level data (Bank profitability database)
Levels
Provisions−0.011**−0.113**−0.011−0.004
Funding gap−0.559**−1.562**−0.239**−0.028
Roa 10.010**0.091**0.011**0.004*
Roe 10.224**1.633**0.189*0.081**
First differences
Provisions−0.008−0.093**−0.0040.001
Funding gap−0.369**−0.880**−0.110−0.013
Roa 10.009*0.088**0.012*0.000
Roe 10.236*1.849**0.1720.041
 Level equationFirst difference equation
Panel B – Bank-level data (Bankscope database)
Provisions−0.005**−0.010**
Loan loss reserves−0.006**−0.005**
Return on assets0.001−0.001
Return on equity0.016**0.003
Liquidity 1−0.088**−0.048**
Liquidity 2−0.050**−0.022**
Funding gap−0.376**−0.137**
Bank equity growth0.356**0.295**
 Pro-cyclicality (−)Counter-cyclicality (+)
Notes:
  1. The results are obtained using a dynamic specification where the dependent variable is regressed on its lagged value and measures of the cycle (real loan growth, real GDP growth, real share price and house price growth). The return on equity (Roe) and return on assets (Roa) are based on profits before tax. Liquidity 1 = liquid assets/(deposits + short-term funding), liquidity 2 = liquid assets/(all funding), the funding gap is the ratio of deposits over loans. Country names are not shown if the coefficient estimates are not significant. These results are obtained by interacting the cycle variable with country dummies.

  2. * and ** denote statistical significance at the 10% and 5% levels. Source: OECD calculations based on the OECD Economic Outlook 86 database, OECD Bank Profitability Database and Bankscope.

Panel C – Country-specific results
ProvisionsAUT, BEL, CZE, ESP, FIN, ITA, PRT, SVK, SWE
Loan loss reservesAUT, CAN, CHE, CZE, DEU, ISL, JPN, POL, SWE, United States
Liquidity 2AUS, AUT, BEL, CAN, DNK, FIN, GBR, HUN, LUX, MEX, NLD, NZL, POL, PRT, SVK, SWE
Return on equityESP, FIN, FRA, GBR, POL, SVKNOR, NZL, United States, TUR
Bank equity growthAUS, AUT, CHE, CZE, DEU, DNK, ESP, GRC, ISL, ITA, JPN, KOR, LUX, NLD, NZL, POL, SWE, United StatesTUR

V Conclusions

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

Reassessing developments in the nature of the cycle leading up to the economic and financial crisis reveals a number of changes. While the amplitude of economic and financial cycles came down, output cycles tended to become longer and more asymmetric with expansionary phases lasting longer, while the length of slowdown or contraction phases remained approximately the same. Asset prices, on the other hand, became more volatile. There are many explanations for the great moderation era that preceded the economic and financial crisis. They focus on good luck, better policy and structural changes in the economy, some of which create vulnerabilities or were only transient factors. Though the evidence is not conclusive, the synchronization of business cycles appears to have become stronger, especially among some country groupings. The degree of synchronization of GDP, real share and house price growth during the economic and financial crisis is unprecedented both across countries and within countries.

With respect to financial markets, capital, provisioning, liquidity and maturity mismatch in the banking sector can generate pro-cyclical behaviour in credit supply for a number of reasons including the regulatory setup, the nature of risk assessment and the prevailing incentives to take risks. While existing empirical work provides little consensus on the degree of pro-cyclicality of the banking system, new estimation results provided in this study, based on aggregate and bank-level micro data sets, show a pronounced pro-cyclicality of the banking sector for most countries, even without taking into account the shadow banking system.

Acknowledgement

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References

The authors are thankful to the editor and two anonymous referees, whose comments greatly improved the study. The usual disclaimer applies.

Notes
  1. 1

    Cycles can be measured in three main ways (Harding and Pagan, 2005), and depending on data availability and frequency. These main measures are: (i) classical (or business) cycles that are fluctuations in the level of an economic variable, (ii) deviation cycles that are differences between the level and permanent component of an economic variable, and (iii) growth rate cycles that are measured by the growth rates of level variables. To obtain data for deviation cycles, filtering techniques (such as the Hodrick–Prescott filter) can be used to identify the permanent component and thus the deviation in levels from this.

  2. 2

    These results are not reported here but can be found in the working paper version of this paper (Égert and Sutherland, 2012).

  3. 3

    Using a sample covering 44 countries over five decades, Claessens et al. (2012) show that rapid credit and house price growth tend to amplify economic expansion. Jordà et al. (2011) report similar findings for a sample of 15 countries coverage more than a century: strong credit growth during economic expansion tends to be followed by deeper recessions and slower rebounds.

  4. 4

    Artis et al. (2003), Böwer and Guillemineau (2006) and Giannone and Reichlin (2006) report similar findings for the euro area. Others report stronger idiosyncratic components (Nadal-De Simone, 2002; Camacho et al., 2008) . With the formation of the euro area, the co-movement of consumption and output became stronger after the mid-1990s (Darvas and Szapáry, 2008).

  5. 5

    Previous studies argue that synchronization is strong during recessions (Canova et al., 2004) and during periods of above average growth (McAdam, 2007).

  6. 6

    Recent empirical studies show that trade and FDI (and portfolio flows) fosters co-movements among OECD economies (Böwer and Guillemineau, 2006; Artis et al., 2008; García-Herrero and Ruiz, 2008).

  7. 7

    Stock and Watson (2002) argue that the decline in volatility was too large to be explained by changes in monetary policy alone.

  8. 8

    Kahn et al. (2002) show that in the United States, inventory levels declined in the mid-1980s and Cecchetti et al. (2006) show that the contribution of inventory changes to GDP growth declined for the major economies.

  9. 9

    Asea and Blomberg (1998) document that bank lending drives and amplifies the overall real cycle in the United States and that there is also a feedback from the real cycle to bank lending.

  10. 10

    The leverage of the banking sector may have become increasingly understated as the shadow banking sector evolved, given its links to banks via contingent credit lines, guarantees and reputational risk.

  11. 11

    Real credit growth gives a biased picture about the importance of credit cycles. A growth rate of say 20% can be translated into very different figures relative to GDP at different stages of financial deepening: it would imply a 2 percentage point expansion relative to GDP for an initial credit to GDP ratio of 10% and a 16 percentage point increase to GDP for a credit stock of 80% of GDP.

References

  1. Top of page
  2. Abstract
  3. I. Introduction
  4. II The Nature of the Last Cycle
  5. III The Last Cycle and the Great Moderation
  6. IV The Last Cycle and the Financial Sector
  7. V Conclusions
  8. Acknowledgement
  9. References
  • Artis, M. J., Fidrmuc, J. and Scharler, J. (2008). The transmission of business cycles. Economics of Transition, 16(3), pp. 55982.
  • Artis, M., Massimiliano, M. and Tommaso, P. (2003). Dating the Euro area business cycle. CEPR Discussion Paper, No. 3696.
  • Asea, P. K. and Blomberg, B. (1998). Lending cycles. Journal of Econometrics, 83(1–2), pp. 89128.
  • Avouyi-Dovi, S. and Matheron, J. (2005). Interactions between business cycles, financial cycles and monetary policy: stylised facts. BIS Papers No. 22, pp. 27398.
  • Ayuso, J., Pérez, D. and Saurina, J. (2002). Are capital buffers pro-cyclical? Evidence from Spanish Data. Bank of Spain Working Paper, No. 0224.
  • Barrell, R. and Gottschalk, S. (2004). The volatility of the output gap in the G7. National Institute Economic Review, No. 188, pp. 1007.
  • Beckmann, R. (2007). Profitability of Western European banking systems: panel evidence on structural and cyclical determinants. Deutsche Bundesbank Banking and Financial Studies Discussion Paper No. 17/2007.
  • Benk, S., Gillman, M., and Kejek, M. (2009). US volatility of output and inflation, 1919-2004: a money and banking approach to a puzzle. CEPR Discussion Paper, No. 7150.
  • Bikker, J. A. and Metzenmakers, P. A. J. (2002). Bank provisioning behavior and procyclicality. De Nederlandsche Bank, Research Series Supervision, No. 50.
  • Bikker, J. A. and Metzenmakers, P. A. J. (2007). Is bank capital procyclical? A cross-country analysis. Mimeo, De Nederlandsche Bank.
  • Black, D. and Dowd, M. (2009). The changing composition of output and the great moderation. Applied Economic Letters, 16(12), pp. 126570.
  • Blanchard, O. and Simon, J. (2001). The long and large decline in U.S. output volatility. Brookings Papers on Economic Activity, 2001(1), pp. 13564.
  • de Blas, B. (2009). Can financial frictions help explain the performance of the U.S. fed? The B.E. Journal of Macroeconomics, 9(1), p. 27.
  • Böwer, U. and Guillemineau, C. (2006). Determinants of business cycle synchronization across Euro area countries. ECB Working Paper, No. 587.
  • Bry, G. and Boschan, C. (1971). Cyclical analysis of time series: selected procedures and computer programs. NBER.
  • Buch, C., Doepke, J. and Stahn, K. (2009). Great moderation at the firm level? Unconditional vs. conditional output volatility. The B.E. Journal of Economic Analysis & Policy, 9(1) (Contributions), p. 20.
  • Camacho, M., Perez-Quiros, G. and Saiz, L. (2008). Do European business cycles look like one? Journal of Economic Dynamics and Control, 32, 21121.
  • Canova, F., Ciccarelli, M. and Ortega, E. (2004). Similarities and convergence in G-7 cycles. ECB Working Paper No. 312.
  • Canova, F. (2009). What explains the great moderation in the U.S.? A structural analysis. Journal of the European Economic Association, 7(4), pp. 697721.
  • Catte, P., Girouard, N., Price, R. and André, C. (2004). Housing markets, wealth and the business cycle. OECD Economics Department Working Papers, No. 394.
  • Cecchetti, S., Flores-Lagunes, A. and Krause, S. (2006). Assessing the sources of changes in the volatility of real growth. NBER Working Paper, No. 11946.
  • Claessens, S., Kose, M. A. and Terrones, M. E. (2012). How do business and financial cycles interact. Journal of International Economics, 87(1), pp. 17890.
  • Comin, D. and Philippon, T. (2005). The rise in firm-level volatility: causes and consequences. NBER Macroeconomics Annual, 2005(20), pp. 167228.
  • Dalsgaard, T., Elmeskov, J. and Park, C. (2002). Ongoing changes in the business cycle: evidence and causes. OECD Economics Department Working Papers, No. 315.
  • Darvas, Z. and Szapáry, G. (2008). Business cycle synchronization in the enlarged EU. Open Economies Review, 19(1), pp. 119.
  • D'Avack, F. and Levasseur, S. (2007). The determinants of capital buffers in CEECs. OFCE Working Paper, No. 2007-28.
  • Dell'Ariccia, G., Detragiache, E. and Rajan, R. (2008). The real effect of banking crisis. Journal of Financial Intermediation, 17(1), pp. 89112.
  • Duval, R., Elmeskov, J. and Vogel, L. (2007). Structural policies and economic resilience to shocks. OECD Economic Department Working Papers, No. 567.
  • Dynan, K., Elmendorf, D. and Sichel, D. (2006a). Financial innovation and the great moderation: what do household data say? Mimeo.
  • Dynan, K., Elmendorf, D. and Sichel, D. (2006b). Can financial innovation help to explain the reduced volatility of economic activity? Journal of Monetary Economics, 53, pp. 12350.
  • Égert, B. and Sutherland, D. (2012). The nature of financial and real business cycles: the great moderation and banking sector pro-cyclicality. OECD Economics Department Working Papers, No. 938.
  • Everts, M. (2007). Duration of business cycles. MPRA Paper No. 1219.
  • Francis, W. and Osborne, M. (2009). On the behaviour and determinants of risk-based capital ratios: revisiting the evidence from UK banking institutions. Financial Services Authority, Occasional Paper Series, No. 31.
  • García-Herrero, A. and Ruiz, J. (2008). Do trade financial linkages foster business cycle synchronization in a small economy? Banco de Espana Documentos de Trabajo, No. 0810.
  • Giannone, D. and Reichlin, L. (2006). Trend and cycles in the Euro area: how much heterogeity and should we worry about it? ECB Working Paper, No. 596.
  • Goodhart, C., Hofmann, B. and Segoviano, M. (2004). Bank regulation and macroeconomic fluctuations. Oxford Review of Economic Policy, 20(4), pp. 591615.
  • Harding, D. and Pagan, A. (2005). A suggested framework for classifying the modes of cycle research. Journal of Applied Econometrics, 20(2), pp. 1519.
  • Harding, D. and Pagan, A. (2006). Synchronisation of cycles. Journal of Econometrics, 132(1), pp. 5979.
  • Jokipii, T. and Milne, A. (2006). The cyclical behavior of European bank capital buffers. Bank of Finland Research Discussion Papers, No. 17/2006.
  • Jordà, Ò., Schularick, M. H. P. and Taylor, A. M. (2011). When credit bites back: leverage, business cycles and crises. NBER Working Paper, No. 17621.
  • Kahn, J., McConnell, M. and Perez-Quiros, G. (2002). On the causes of the increased stability of the US economy. Economic Policy Review, Federal Reserve Bank of New York 8(1), pp. 183202.
  • Keys, B. J., Mukherjee, T., Seru, A. and Vig, V. (2010). Did securization lead to lax screening? Evidence from subprime loans. Quarterly Journal of Economics, 125(1), pp. 30762.
  • Kim, H.-W. and Lee, H.-Y. (2006). Procyclicality of buffer capital and its implications for basel II: a cross-country analysis. In H.-W. Kim, H.-S. Shin (eds), Adopting the new basel accord: impact and policy responses of Asia-Pacific developing countries. International Conference of the Korea Development Institute, ch 6.
  • Kiyotaki, N. and Moore, J. (1997). Credit cycles. Journal of Political Economy, 105(2), pp. 21148.
  • Kose, M., Otrok, C. and Prasad, E. (2008). Global business cycles: convergence or decoupling? NBER Working Paper, No. 14292.
  • Lindquist, K.-G. (2003). Banks' buffer capital: how important is the risk? Norges Bank Working Paper, No. 2003/11.
  • McAdam, P. (2007). US, Japan and the Euro area: comparing business-cycle features. International Review of Applied Economics, 21(1), pp. 13556.
  • McConnell, M. and Perez-Quiros, G. (2000). Output fluctuations in the United States: what has changed since the early 1980's? American Economic Review, 90(5), pp. 146476.
  • Nadal-De Simone, F. (2002).Common and idiosyncratic components in real output: further international evidence. IMF Working Paper, No. 229.
  • Pain, N., Koske, I. and Sollie, M. (2006). Globalisation and inflation in the OECD economies. OECD Economics Department Working Papers, No. 524.
  • Panetta, F., Angelini, P., Albertazzi, U., Columba, F., Cornacchia, W., DiCesare, A., Pilati, A., Salleo, C. and Santini, G. (2009). Financial sector pro-cyclicality: lessons from the crisis. Banca d'Italia Occasional Papers, No. 44.
  • Reinhart, M. and Rogoff, K. S. (2008). Banking crises: an equal opportunity menace. NBER Working Paper, No. 14587.
  • SchularickM. A. M. and TaylorA. M. (2012). Credit booms gone bust: monetary policy, leverage cycles and financial crises, 1870-2008. American Economic Review, 102(2), pp. 102961.
  • StirohK. J. (2009). Volatility accounting: a production perspective on increased economic stability. Journal of the European Economic Association, 7(4), pp. 67196.
  • Stock, J. and Watson, M.(2002). Has the business cycle changed and why? NBER Macroeconomics Annual, NBER.
  • StockJ. and WatsonM. (2003). Forecasting output and inflation: the role of asset prices. Journal of Economic Literature, 41(3), pp. 788829.
  • Stolz, S. and Wedow, M. (2005). Banks' regulatory capital buffer and the business cycle: evidence for German savings and cooperative banks. Deutsche Bundesbank Banking and Financial Studies Discussion Paper, No. 07/2005.
  • Taylor, J. (1998). An historical analysis of monetary policy rules. NBER Working Paper, No. 6768.