Have Business Cycles Changed? An Empirical Investigation PDF

Journal of Development Economics 109 (2014) 98–123 Contents lists available at ScienceDirect Journal of Development Economics journal homepage: www.elsevier.com/locate/devec Have business cycles changed over the last two decades? An empirical investigation☆ César Calderón a,⁎, J. Rodrigo Fuentes b a b The World Bank, USA Pontiﬁcia Universidad Católica de Chile, Chile a r t i c l e i n f o Article history: Received 25 October 2012 Received in revised form 26 February 2014 Accepted 3 March 2014 Available online 12 March 2014 JEL classiﬁcations: E32 F41 Keywords: Business cycles Peaks and troughs Emerging markets a b s t r a c t We document the properties of business cycles of 71 countries (23 industrial countries and 48 emerging market economies, or EMEs), from 1970q1 to 2012q4 using the Harding and Pagan dating algorithm. First, recessions are deeper, steeper and costlier among EMEs (especially in East Asia and Latin America). Second, recoveries are swifter and stronger among EMEs, partly due to stronger rebound effects. Third, recessions became less costly during the globalization period (1985–2007) for industrial countries and EMEs, thus reﬂecting institutional changes made during the “Great Moderation.” Fourth, the dynamic behavior of macroeconomic indicators around peaks in real GDP is more volatile in downturns associated with crisis compared to other downturns. Fifth, peaks in ﬁnancial cycles (credit and asset prices) tend to precede peaks in real output cycles. Finally, although both industrial and emerging markets have experienced deep recessions during the recent global ﬁnancial crisis, the emerging markets have recovered faster. © 2014 Elsevier B.V. All rights reserved. 1. Introduction Emerging market economies (EMEs) have experienced greater macroeconomic volatility than industrial economies. Fluctuations in output, exchange rates, and current account balances are typically more frequent, sharper, and abrupt in EMEs. This phenomenon is typically attributed to country-speciﬁc factors that amplify external shocks and have led to a higher incidence of banking, currency, and external debt crises (The World Bank, 2007).1 These country-speciﬁc characteristics include excessive dependence on a few volatile sectors, a narrow tax base, fragile ﬁnancial systems, weak institutions, and poor economic ☆ We would like to thank Gianluca Clementi, Alberto Naudón, Klaus Schmidt-Hebbel, Rodrigo Valdés and two anonymous referees for comments and suggestions, as well as participants at the WB-CEPR-CREI Conference on “The Growth and Welfare Effects of Macroeconomic Volatility,” 2007 LACEA Conference in Bogotá, 2007 Meetings of the Chilean Economic Society (SECHI), the Central Bank of Chile Seminar and the 2010 Econometric Society World Congress in Shanghai. Special thanks to David Rappoport for outstanding research assistance. The views expressed in this paper are those of the authors, and do not necessarily reﬂect those of the World Bank or its Boards of Directors. The usual disclaimer applies. ⁎ Corresponding author at: The World Bank, Ofﬁce of the Chief Economist of the Africa Region. Tel.: +1 202 458 7214. E-mail address: [email protected] (C. Calderón). 1 Recent examples of such crisis episodes are the Tequila and East Asian crises and massive depreciations of the Brazilian and Russian currencies, the subprime crisis in the U.S., and the Greek sovereign debt crisis. The recurrence of crisis episodes has increased interest in disentangling the sources of economic crises. http://dx.doi.org/10.1016/j.jdeveco.2014.03.001 0304-3878/© 2014 Elsevier B.V. All rights reserved. policies. More recently, however, the focus has gradually shifted towards the external (exogenous) environment faced by EMEs, including real shocks, e.g., shocks to commodity prices and to a country's external demand (Arora and Vamvakidis, 2005; Broda, 2004), ﬁnancial shocks, such as rising world interest rates and global risk aversion (Neumeyer and Perri, 2005; Uribe and Yue, 2006), and natural disasters (Calderón and Levy-Yeyati, 2009; Loayza et al., 2012; Raddatz, 2007). In the search for an explanation for the excess volatility of output ﬂuctuations in EMEs relative to advanced economies, Aguiar and Gopinath (2007) build a real business cycle where shocks to trend growth are the main drivers of output ﬂuctuations in EMEs while developed economies tend to experience transitory ﬂuctuations around a stable trend. The introduction of stochastic productivity trends permits the authors to replicate the stylized facts of business cycles in EMEs vis-à-vis industrial countries. Other authors show that business cycle features of EMEs can be replicated in models with ﬁnancial imperfections that amplify transitory productivity shocks (Chang and Fernández, 2013). This paper attempts to describe the main features of the business cycles of emerging market economies and industrial countries as captured by the duration, amplitude, slope, and cost of downturns and upturns in real economic activity. To accomplish this task we use a comprehensive quarterly dataset for 71 countries (23 industrial economies and 48 emerging-market economies) from 1970q1 to 2012q4. The focus of our paper is to compare the main features of real output cycles in EMEs vis-à-vis industrial countries before the outbreak of the global ﬁnancial crisis (1970–2007). Next, we examine whether those facts changed C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 with the advent of the recent global crisis by comparing the periods 1990–2006 and 2007–2012.2 The main contribution of this paper is three-fold: First, we use a common methodology for dating turning points for a large sample of countries using quarterly data.3 This analysis allows us to estimate comparable statistics of duration, depth, and speed of recessions and recoveries for both industrial and emerging-market economies. Second, we examine the evolution of the main business cycle characteristics over time. Speciﬁcally, we compare the main features of the cyclical output phases during the periods of pre-globalization and globalization (1970–84 and 1985– 2007, respectively) for both industrial and emerging-market economies. Third, we assess whether the length and size of cyclical phases of real output for industrial and emerging markets have changed during the recent global ﬁnancial crisis compared with the performance of real economic activity over the past 15 years. Here, we focus not only on the differences across country groups in the duration and depth of the recession but also on the speed and strength of the ensuing recovery. We further document the excess volatility of real output ﬂuctuations in EMEs when compared with industrial countries by ﬁrst reporting the joint distribution of output contractions by duration (short, medium, long, and protracted) and amplitude (mild, moderate, severe, and depression).4 Second, we compare the main features of recessions during crisis and their ensuing recoveries with those that do not coincide with crisis episodes.5 Third, we zoom in on the correlates of real output cycles by exploring the dynamics of macroeconomic variables around recessionary periods using event study analysis. We report the trajectory during a four-year window around peaks in real GDP associated with banking crises and compare them with periods without crisis for the following real and ﬁnancial indicators: private consumption, investment, real credit to the private sector, stock prices, and real exchange rates. Finally, this paper provides a ﬁrst assessment of the traits of recessions and recoveries during the recent global ﬁnancial crisis (GFC). We compute the duration, amplitude, and slope of recessions as well as their ensuing recoveries for industrial and emerging-market economies during the most recent cycle (2007–2012) and we compare these ﬁndings with those of output cycles over the last 15 years before the GFC (1990–2006). Next, we examine the dynamic pattern of real and ﬁnancial indicators around peaks in real economic activity during 1990–2006 and 2007–2012. In sum, we aim to answer the following questions: Are there systematic differences in the main features of business cycles (duration, amplitude, and cost) in industrial countries vis-à-vis emerging markets? Are business cycles alike within emerging markets? Are business cycles similar over time and across country groups? Are the main features of recessions and recoveries different when a crisis occurs? Do crises matter for the dynamics of macroeconomic indicators around recessionary periods? Have the main stylized facts remained invariant during the recent global ﬁnancial crisis? Are recessions longer and deeper this time around? Are the ensuring recoveries faster and stronger? The evidence presented in this paper conﬁrms that recessions in emerging-market economies are deeper, steeper, and hence, costlier than those in industrial countries, although they have the same duration. 2 The full list of countries, including their regional and income classiﬁcation as well as the sources of data, is presented in Table A1. 3 Despite large output ﬂuctuations in EMEs, empirical research on business cycles has been mainly conducted for advanced economies. Some important exceptions are Hoffmaister et al. (1998), Agénor et al. (2000), Herrera et al. (2000), Neumeyer and Perri (2005), Raddatz (2007), Aguiar and Gopinath (2007, 2008), Cerra and Saxena (2008), and Aioilﬁ et al. (2011). However, one of the limitations in most of these papers is that they either use annual data or limit themselves to a small group of countries. 4 We follow Morsink et al. (2002) in classifying the severity of recessions by duration and amplitude. The detailed discussion of this classiﬁcation is presented in Section 3.2. 5 Episodes of economic crisis, furthermore, are loosely deﬁned in this paper as the occurrence of at least one of these types of crisis: banking crisis, currency crisis, sovereign default, and restructuring of external and domestic debt. The procedure followed to identify any of these types of crisis episodes will be described in Section 3.2. Finally, we deﬁne recessions that coincide with crisis as those peak-to-trough phases that take place within a three-year window surrounding the occurrence of a crisis episode. 99 On the other hand, the ensuing recoveries in EMEs are stronger and more intense but slower and more volatile. The strong recovery among EMEs could be attributed to a larger rebound effect or to the fact that these countries have experienced a larger trend-growth rate than industrial economies during the period of analysis. We provide evidence that recessions during the globalization period (1985–2007) are less severe for Latin America and the Caribbean (LAC) compared with the previous period (1970–1984) while the main traits of recessions in East Asia and the Paciﬁc (EAP) and Eastern Europe and Central Asia (ECA) remain unchanged with the advent of globalization. Deeper and costlier recessions in EMEs are associated with a higher incidence of (ﬁnancial, currency, or debt) crises. During crisis-related downturns, troughs in consumption and investment are deeper while real credit and asset prices tend to be more volatile in EMEs (as opposed to regular recessions). Moreover, we ﬁnd that peaks in real credit and stock prices tend to precede peaks in real output during crises in EMEs, and that the domestic currency tends to depreciate in real terms, while it appreciates for industrial countries. Finally, recessions among industrial countries and EMEs, except for Latin America, are deeper and steeper during the recent global ﬁnancial crisis when compared with the pre-crisis globalization period. Interestingly, the rebound from the profound downturns among industrial economies has been slower this time around, as opposed to the fast recovery of EMEs (excluding ECA). The paper is divided into four sections. In Section 2 we brieﬂy describe some methodological issues regarding business cycle dating. Given the lack of consensus in the literature, we opt for a methodology to characterize business cycles that has the following characteristics: (a) it does not rely on arbitrary trend-cycle decompositions, (b) it provides a uniform statistical foundation to identify turning points, (c) it is robust to changes in the sample period, and (d) it is easy to replicate for a wide array of countries. Speciﬁcally, we implement the quarterly adaptation of the Bry–Boschan algorithm (BBQ) proposed by Harding and Pagan (2002). Following the traditional approach outlined by Burns and Mitchell (1946), we identify turning points in the level of output and deﬁne the different phases of the cycle (recession, recovery, and expansion) and their characteristics—including duration, amplitude, slope, and cumulative movements for each phase of the cycle. In Section 3, we ﬁrst discuss the business cycle features of our sample of 71 countries from 1970q1 to 2007q4 (before the onset of the global ﬁnancial crisis). Then, we argue that the greater depth of business cycles in EMEs relative to industrial countries is inﬂuenced by their coincidence with crisis episodes. Therefore, we compute the main features of downturns and upturns associated with crisis and examine the correlates of downturns in economic activity using event-study analysis for consumption, investment, credit, and asset prices during crisis. Next, we compare the length, depth, and intensity of business cycles during the recent global crisis (2007–2012) and the precrisis period of 1990–2006. Recessions are signiﬁcantly deeper in industrial economies and emerging markets, except for LAC. However, recoveries are faster and stronger in EMEs (other than ECA countries), and slower and weaker in industrial countries. Finally, Section 4 concludes. 2. Measuring business cycles This section outlines the methodology used to characterize business cycles for a sample of industrial countries and emerging market economies. The ﬁrst part of the section discusses the advantages and disadvantages of some of the methodological options for detecting turning points, whereas the second part describes the statistical technique used in this paper; that is, the quarterly adaptation of the Bry–Boschan (BBQ) algorithm proposed by Harding and Pagan (2002). 2.1. Methodological issues There is no single approach in the literature to characterizing the features of the business cycle. On the one hand, the seminal work by 100 C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Hamilton (1989) dates peaks and troughs by modeling the shift in the growth rate of GDP using Markov-switching (MS) methods. Harding and Pagan (2002), on the other hand, propose a non-parametric approach, which is used in this paper, to identify cyclical turning points in quarterly series; this is the so-called BBQ algorithm. These two approaches have advantages and disadvantages as discussed in Harding and Pagan (2003a,b), Hamilton (2003), Chauvet and Hamilton (2005) and Chauvet and Piger (2008). There is no consensus in the literature on the optimal method to detect turning points in a series. Chauvet and Piger (2008) argue that the MS approach outperforms the BBQ algorithm when predicting peaks and troughs in real time. Nevertheless, they also ﬁnd that the MS and BBQ approaches can accurately identify the NBER business cycle chronology of U.S. economic activity. Hence, both methodologies can provide similar results if the main purpose of the exercise is to document the historical chronology of turning points. Harding and Pagan (2003a,b) argue that the BBQ algorithm provides a simple and transparent way to detect the turning points for a time series and is not sensitive to changes in the parameterization of the data-generating process (DGP) of real GDP or to changes in the sample period of the series. Hamilton (2003), on the other hand, argues that MS and BBQ are philosophically different methods, and that the adequacy of the MS approach lies on its goal to make optimal inference on an unobserved phenomenon (a recession) based on the DGP of a series of indicators of real economic activity. In his critical review of the BBQ algorithm, Hamilton argues that the rule for identifying turning points depends on whether the country is fast- or slow-growing and on the quality of GDP data.6 However, he concurs that the BBQ algorithm is more transparent (Hamilton, 2003, p. 1693). The purpose of this paper is rather modest. We identify turning points in historical series of real GDP to characterize business cycles for a large sample of developed and developing countries.7 Predicting peaks and troughs in real time or undertaking comparisons among dating methods are beyond the scope of this paper. To accomplish the task set out, we apply the Harding and Pagan (2002) BBQ algorithm to the series of real GDP (in logs) for 71 countries. We argue that adopting a common criterion to identify cyclical phases in real economic activity is the proper way to proceed and, in that sense, the BBQ algorithm provides a transparent way to conduct cross-country comparisons of the different phases of the business cycle. Once these peaks and troughs are identiﬁed, characterization of the business cycles is undertaken by computing the duration, amplitude, slope, and cumulative variation of recessions and recoveries for our sample of countries. Recent applications of the BBQ algorithm to characterize real and ﬁnancial cycles include Claessens et al. (2009, 2011a,b, 2012) and Calderón and Servén (2013). 2.2. The Harding and Pagan algorithm8 The classical approach to analyzing business cycles, as outlined in the seminal work of Burns and Mitchell (1946), deﬁnes business cycles as the sequence of expansions and contractions in the level of either real output or employment. Speciﬁcally, this approach detects turning points in an aggregate series: typically, the log level of real GDP. Alternatively, 6 An example of the need for a country-speciﬁc rule is that we are unable to ﬁnd a recession for China using the BBQ algorithm. Concerning the quality of the data, it is well known that the quality of statistics is lower for less developed economies or such economies have large informal sectors that conceal a portion of the output produced. For instance, Jerven (2010) extensively documents the implications of poor data quality on the measurement of real GDP in Africa. 7 We are aware that business cycles are characterized by more than just turning points in real GDP. The paper concentrates on documenting cross-country differences in the duration, amplitude, and slope of recessions and recoveries in real GDP for a large group of economies rather than describing a large number of series. Nevertheless, a characterization of other variables along the cycle is provided in Section 4. 8 The description of the methodology and statistics draws heavily from Harding and Pagan (2002, 2003a) and Claessens et al. (2009, 2011a,b, 2012). empirical business cycle research has focused on identifying growth cycles by calculating deviations from long-run trends, and these trends are estimated using different techniques (deterministic trend models, the Hodrick–Prescott ﬁlter, and the band-pass ﬁlter, among others). However, the literature argues that this methodology tends to overestimate the frequency of turning points and underestimate their amplitude when compared with classical cycles (Morsink et al., 2002). Also, the dating of turning points using growth cycles rather than classical cycles is sensitive to the inclusion of new data (Claessens et al., 2009, 2011a,b, 2012). Following the classical approach, we use the Harding and Pagan (2002) extension of the Bry and Boschan (1971) algorithm to identify cyclical turning points in quarterly series (the BBQ algorithm). In fact, this algorithm requires that: (1) Complete cycles run from peak to peak and have two phases: contraction (peak to trough) and expansion (trough to peak), and peaks and troughs must alternate, and (2) The minimum duration of a complete cycle is at least ﬁve quarters and each phase of the cycle must last for at least two quarters. Local maximum and minimum values of real output (typically expressed in natural logs) can be determined by looking at the differences in our measure of real GDP. We denote yit as the (log level of) quarterly real GDP of country i in time t. Hence, Harding and Pagan deﬁne the local optima as follows: (a) A cyclical peak in the level of real output of country i occurs at time t if: 2 1−L yit N 0; ð1−LÞyit N 0 2 and ð1−LÞyi;tþ1 b 0; 1−L yi;tþ2 b 0 (b) A cyclical trough takes place in country i at time t if: 2 1−L yit b 0; ð1−LÞyit b 0 and 2 ð1−LÞyi;tþ1 N 0; 1−L yi;tþ2 N 0 and L is the lag operator, where Lkxt = xt − k. The algorithm described above ensures that yit is a local optimum relative to two quarters on either side of yit.9 This notion of local optimum and compliance with the censoring rule (minimum duration of cycle and phases) deﬁnes a complete cycle. Using the BBQ algorithm, we identify peaks and troughs in the quarterly series of real GDP for an initial sample of 75 countries over the period 1970–2007. Our sample consists of 23 industrial countries and 52 emerging-market countries. Within the latter group, we gather information for 15 Latin American countries, 11 East Asian countries, 17 countries in Eastern Europe and 9 other economies. The full sample of countries—including their regional and income classiﬁcation—is presented in the Annex, Table A1. Our quarterly real GDP data is expressed in local currency at constant prices. Detailed information on the time coverage, denomination, base year, and sources of data can also be found in Table A1. Note that the time coverage differs across countries. Most of the advanced economies and a few emerging markets have data from 1970 while the real GDP data for most emerging markets start in the second half of the 1970s or the 1980s. The real GDP ﬁgures for most transition economies (ECA countries) and lower-income emerging markets begin in the 1990s or early 2000s. The BBQ algorithm is unable to ﬁnd turning points in the real GDP data for China, Poland, the Slovak Republic, and Vietnam, leaving us with an effective sample of 71 countries with identiﬁed peak-to-trough and/or trough-to-peak phases of real output cycles. The steady and sharp growth in China's real GDP over the last 25 years prevents us from ﬁnding these 9 An even simpler sequence rule is available from the idea that a turning point in a graph at time t requires that the derivative change sign at t. Thus, treating Δyt as a measure of the derivative of yt with respect to t leads to consideration of the sequence {Δyt N 0, Δyt + 1 b 0} as signaling a peak. The problem with this method is that it conﬂicts with the requirement that a phase lasts for at least two quarters. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 101 turning points in that country's data, while the short time span of the real GDP series is the culprit for the other three countries. After computing the turning points in real output, the main features of expansions (from trough to subsequent peak) and contractions (from peak to trough) in real economic activity are characterized in terms of duration, amplitude, slope, and cumulative variation. In addition, we consider more information, from a cyclical standpoint, to characterize real output upturns. Following Claessens et al. (2011a,b, 2012) we deﬁne upturns or recoveries as the early stages of the expansion phase, when real GDP reaches the level of the previous peak coming from a trough. The main features of business cycle ﬂuctuations are deﬁned as follows: ﬁnancial crisis; that is, from 1970q1 to 2007q4.11 In relation to the business cycle features of countries during this period, we also examine whether the main traits of recessions, recoveries, and expansions changed during the recent globalization period. We compare the pre-globalization period (1970–1984) to the globalization period (1985–2007).12 The second part of the analysis examines whether cross-country differences in business-cycle features remain unchanged during the global ﬁnancial crisis (which we refer to as the GFC period); hence, we compare the behavior of business cycles during the GFC period (2007–2012) and the pre-GFC period (which we deﬁne as 1990 to 2006). (1) Duration of the cycle. This is computed as the number of quarters from peak to trough during contraction episodes and from trough to the next peak in the expansion phase. In addition, the duration of the recovery (upturn) is the number of quarters that it takes the real GDP to rebound from the trough to its previous peak. (2) The amplitude of the cycle is calculated as the maximum drop of GDP from peak (trough) to trough (peak) during episodes of contraction (expansion). For instance, the amplitude of the contraction, AC, measures the change in real GDP from a peak (y0) to the next trough (yK), that is, AC = yK − y0. The amplitude of upturns is measured as the four-quarter change in real GDP following a trough, as suggested by Sichel (1994) and Claessens et al. (2011a, 2012). (3) The slope of each phase is computed as the ratio of the amplitude of the peak-to-trough (trough-to-peak) phase of the cycle to its duration. The slope of the upturn is the amplitude from trough to the previous peak divided by its duration. (4) The cumulative variation of the cycle is estimated as the area of the triangle made up of the duration and amplitude. It reﬂects the idea of foregone output from peak to trough during contractions and the output gains during expansion episodes. For the peak-to-trough phase of the cycle, the cumulative output loss LC (i.e., an approximate measure of the overall cost of a cyclical contraction), with duration of k quarters, is deﬁned as A k c LC ¼ ∑ y j −y0 −. 2 j¼1 Table 1 presents the descriptive statistics (average, median, and standard deviation, among others) of the different phases of the cycle (recessions, recoveries, and expansions) for the full sample of countries as well as for the industrial countries and the emerging-market economies. Our discussion focuses on the various attributes of recessions and recoveries. Although we report the most important elements of expansions in real economic activity, we keep the discussion of their stylized facts to a minimum given that we are unable to distinguish temporary (cyclical) from permanent (long-term) shocks driving this trough-topeak phase of the cycle. 3. Characterizing business cycles In this section we estimate the main features of the different phases of the business cycle of 71 countries (23 industrial countries and 48 emerging market economies) from 1970q1 to 2012q4. Our analysis compares cross-country business cycle characteristics using two different types of country classiﬁcation: (a) industrial countries and emerging-market economies, and (b) the regional classiﬁcation of emerging markets. In the latter case, our effective sample includes nine countries from East Asia and the Paciﬁc (EAP), 15 countries from Eastern Europe and Central Asia (ECA), and 15 countries from Latin America and the Caribbean (LAC).10 The ﬁrst part of our analysis focuses on calculating the duration, amplitude, slope, and cumulative variation of recessions, recoveries and expansions of real economic activity before the outbreak of the global 10 Note that our sample of countries includes 9 emerging market economies from other regions: 6 countries from the Middle East and North Africa, 1 country from South Asia, and 2 countries from Sub-Saharan Africa. We also conduct an analysis by groups of countries classiﬁed according to income levels (30 high-income countries, 23 upper-middleincome countries, and 13 lower-middle-income countries) but the results do not provide more compelling evidence than that presented in this paper. Those results are not reported here but are available from the authors upon request. 3.1. Main features of real output cycles across the world before the 2008–2009 crisis Fact 1: recessions and recoveries for industrial countries and emerging market economies are not alike Recessions in emerging markets tend to last as long as those in industrial markets—with an average duration from the peak-to-trough phase of the cycle of 3.7 and 3.6 quarters respectively (see Table 1). The coefﬁcient of variation of the duration distribution is also similar across groups (0.54), implying that the standard deviation of the duration of recessions for both industrial and emerging-market economies is nearly half of the value of their average duration (approximately two quarters). However, emerging markets tend to experience deeper recessions: The median drop in real economic activity for emerging markets is larger and more abrupt than that of industrial countries. That is, real output declines by about 1.9% in recessions in industrial economies compared to a 4.8-percent drop in emerging-market economies. In other words, real economic activity declines at a rate of 0.5% per quarter among industrial countries while it falls at a rate of 1.4% per quarter among emerging markets.13 As expected, recessions are costlier among emerging markets, with a median cumulative loss of 7.5% (compared with 3.1% for industrial economies). Interestingly, we ﬁnd that the dispersion of the amplitude and slope is wider within the group of emerging-market economies than among industrial countries (the standard deviation among EMEs is more than three times that of industrial economies and the coefﬁcient of variation is slightly larger). However, that is not the case for the cumulative loss: The variation coefﬁcient (as an absolute value) of the cumulative loss 11 Related research has been conducted by Claessens et al. (2011b) and Gupta and Miniane (2009). The former analyze the cycles of 23 emerging economies for the period 1978:1-2007:4 using a similar methodology; their main concern is the correlation between recessions and credit and asset price cycles. They compare those results with results for a group of 21 OECD economies. The latter paper concentrates on contractions and recoveries for 8 Asian economies, comparing them with non-Asian emerging economies and industrial economies. Altuğ and Bildirici (2012) use a univariate Markov switching model developed by Hamilton (1989) to compare business cycles in a sample of 13 emerging-market economies and 14 developed economies. Their study analyzes the synchronization of cycles across these groups of economies and compares parametric and non-parametric methods for dating cycles. 12 There is no clear consensus on the deﬁnition and starting date of the globalization period. We consider 1985 to be the starting year given that it coincides with the onset of the Uruguay Round of trade negotiations which sped up unilateral trade liberalization programs in several developing nations (see Deardorff and Stern, 2002; Kose et al., 2012). 13 Statistical tests show that there are no signiﬁcant differences in the duration of recessions in EMEs vis-à-vis industrial countries, but that differences in terms of the amplitude and slope of recessions are statistically considerable. 102 Table 1 Basic features of real output cycles. Sample of 71 countries, 1970–2007 (quarterly information). Recessions Recoveries Expansions Duration (quarters) Amplitude (%, median) Slope (%, median) Cum. Loss (%, median) Time (%, median) Duration (quarters) Amplitude (%, median) Slope (%, median) Duration (quarters) Amplitude (%, median) Slope (%, median) All countries Average Median Standard deviation Coefﬁcient of variation 25 th percentile 75 th percentile Maximum Minimum Number of events 14.5% 12.5% 8.3% 0.57 8.1% 19.7% 0.0% 37.5% 244 3.8 3.0 2.1 0.54 2.00 5.00 2.0 13.0 244 −4.94% −2.88% 5.63% −1.14 −6.29% −1.37% −35.03% −0.13% 244 −1.32% −0.87% 1.36% −1.03 −1.69% −0.44% −7.53% −0.06% 244 −11.24% −4.95% 18.27% −1.63 −12.41% −1.90% −137.73% −0.01% 244 22.0% 19.1% 12.0% 0.54 14.7% 27.1% 4.0% 60.0% 221 4.1 3.0 3.6 0.88 2.00 5.00 1.0 21.0 221 6.06% 4.81% 4.98% 0.82 2.67% 7.51% 0.12% 41.47% 221 2.85% 1.76% 3.98% 1.39 0.63% 3.41% 0.02% 41.47% 221 18.6 13.0 17.8 0.96 5.25 26.00 2.0 102.0 228 23.48% 15.85% 24.89% 1.06 7.41% 31.16% 0.24% 192.48% 228 1.46% 1.26% 1.12% 0.77 0.82% 1.84% 0.12% 9.38% 228 Industrial countries Average Median Standard Deviation Coefﬁcient of variation 25 th percentile 75 th percentile Maximum Minimum Number of events 12.2% 12.5% 4.5% 0.36 7.9% 15.8% 3.9% 21.1% 101 3.9 3.0 2.1 0.54 2.00 5.00 2.0 13.0 101 −2.34% −1.85% 2.13% −0.91 −2.98% −0.93% −13.45% −0.13% 101 −0.61% −0.49% 0.44% −0.71 −0.86% −0.29% −2.17% −0.06% 101 −6.25% −3.12% 11.21% −1.79 −6.58% −1.31% −93.96% −0.14% 101 17.0% 15.8% 6.2% 0.36 13.8% 19.3% 6.9% 34.8% 96 3.6 3.0 2.9 0.80 2.00 5.00 1.0 18.0 96 3.34% 2.82% 2.06% 0.62 1.97% 4.34% 0.12% 11.47% 96 1.76% 1.05% 2.04% 1.16 0.38% 2.22% 0.02% 11.47% 96 22.8 18.0 20.0 0.88 7.75 32.00 2.0 101.0 98 20.10% 14.98% 20.62% 1.03 6.82% 26.96% 0.37% 130.26% 98 0.92% 0.87% 0.66% 0.72 0.60% 1.08% 0.12% 5.93% 98 Emerging markets Average Median Standard deviation Coefﬁcient of variation 25 th percentile 75 th percentile Maximum Minimum Number Of events 15.6% 13.5% 9.5% 0.61 8.0% 21.5% 0.0% 37.5% 143 3.8 3.0 2.0 0.54 2.00 5.00 2.0 13.0 143 −6.78% −4.83% 6.55% −0.97 −9.47% −2.12% −35.03% −0.14% 143 −1.82% −1.39% 1.56% −0.86 −2.26% −0.69% −7.53% −0.07% 143 −14.76% −7.51% 21.28% −1.44 −17.10% −2.32% −137.73% −0.01% 143 24.5% 21.2% 13.4% 0.55 15.4% 32.1% 4.0% 60.0% 125 4.4 3.0 4.0 0.91 2.00 5.50 1.0 21.0 125 8.16% 6.67% 5.53% 0.68 4.74% 10.59% 0.27% 41.47% 125 3.69% 2.33% 4.82% 1.30 1.17% 4.55% 0.03% 41.47% 125 15.4 9.5 15.3 1.00 4.75 21.00 2.0 102.0 130 26.03% 16.95% 27.49% 1.06 8.28% 35.57% 0.24% 192.48% 130 1.86% 1.63% 1.22% 0.66 1.25% 2.04% 0.12% 9.38% 130 Note: Recessions are deﬁned as the period (in quarters) between a peak in real GDP and its subsequent trough. Expansions, on the other hand, cover the period between a trough and the next peak in real output. Recoveries in real output represent the early stages of the expansion and take place in the period that it takes real GDP to increase from a trough to its previous peak level. Time in recession (recovery) is deﬁned as the number of quarters in which the economy is in a peak-to-trough (troughto-previous peak) phase of the cycle as a share of total time length of the series. The amplitude of the recession is computed as the percentage variation in the real GDP from its peak to its trough, while the amplitude of the recovery is computed as the 4-quarter cumulative variation in real output following a trough. The slope of the downturn is the ratio of the amplitude to the duration of the downturn (or peak-to-trough phase) whereas that of the upturn is amplitude from trough to previous peak divided by its duration. The cumulative loss combines information on the duration and amplitude to measure the overall cost of recession. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Time (%, median) C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 is smaller in EMEs compared with industrial economies (1.44 vs. 1.79, respectively). Recoveries tend to be slower among emerging markets vis-à-vis industrial economies (4.2 and 3.6 quarters, respectively) although differences in the duration of the recovery period are not statistically signiﬁcant. There is greater variability in the duration of recoveries among EMEs as compared with industrial countries (a standard deviation of 4 vs. 2.9 quarters). The strength and intensity of recoveries are different across groups: Trough-to-peak phases are sharper in both amplitude and slope for emerging markets (medians of 6.7 and 2.3%, respectively) than those of industrial countries (2.8 and 1.1%, respectively). Not only are recoveries stronger and more intense among emerging markets but they are also more volatile—i.e., the standard deviation of the amplitude and slope of recoveries in emerging-market economies more than doubles that of industrial countries. Finally, we should point out that expansions in real output tend to be longer in industrial countries as opposed to EMEs (23 vs. 15 quarters) and although the rhythm of these expansions is slower (0.9 vs. 1.6% per quarter), they resume, on average, after milder downturns. In sum, although the average duration of recessions and recoveries is roughly similar in emerging markets and industrial economies, real output contractions in the former group are deeper and more intense than those in the latter group. Moreover, country cycles show a greater degree of heterogeneity across the group of emerging economies than among industrial countries (see Table 1 and Annex Table A.2). We should note that deeper recessions in emerging markets (as opposed to industrial countries) are followed by stronger recoveries; that is, there is a more powerful cyclical rebound effect. Fact 2: the duration of recessions and recoveries is similar across geographical regions. However, recessions are deeper (in amplitude and slope) for Latin America and the Caribbean (LAC) and Eastern Europe and Central Asia (ECA), while recoveries are more dynamic in East Asia and the Paciﬁc (EAP) and ECA Fact 1 illustrates that, on average, the duration of recessions is roughly similar for emerging markets and industrial countries. Not surprisingly, differences in the duration of peak-to-trough phases of the cycle across regional groups of emerging-market countries are statistically negligible (see panel (a) of Fig. 1). However, there are large differences in the depth and cost of recessions across emerging market regions. For instance, the median amplitude of the recession is −3.8% for EAP countries, −5.5% for ECA countries, and −5.8% for LAC countries (see panel (b) of Fig. 1). Overall, recessions tend to be costlier in ECA and LAC (with cumulative losses in real output of 8.6 and 8.4%, respectively) relative to EAP (which registers a cumulative loss of 6.8%). Real economic activity recoveries, on the other hand, tend to be faster in EAP (4 quarters) as opposed to ECA and LAC (4.6 and 4.8 quarters, respectively). Their amplitude is larger in EAP (7.7%) and Eastern Europe (8.1%) compared with Latin America and the Caribbean (6.1%). Also, note that the slope or intensity of the recovery is greater in the ECA region (2.7% per quarter) (see panel (c) of Fig. 1). Fact 3: during the globalization period, recessions tend to be less frequent for industrial countries but shorter only for emerging economies when compared with the pre-globalization period. Recoveries are slower and expansions are shorter for both industrial and emerging market economies in the globalization period We next examine whether the main features of the business cycle in industrial countries and emerging-market economies have changed during the globalization period. Table 2 compares the average duration, median amplitude and slope of real cycles during the pre-globalization and globalization periods (1970–1984 and 1985–2007, respectively). Real output contractions during the globalization period tend to be less frequent among industrial countries (time in recession falls from 103 Business Cycle Features, By Region a) Duration (average, in quarters) 5 4 3 2 1 0 -1 -2 -3 -4 East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recession Recovery b) Amplitude (median, %) 10% 8% 6% 4% 2% 0% -2% -4% -6% East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recession Recovery c) Slope (median, %) 3% 2% 1% 0% -1% -2% East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recession Recovery Fig. 1. Business cycle features, by region. (a) Duration (average, in quarters). (b) Amplitude (median, %). (c) Slope (median, %). 18.6 to 9.4%) but those recessions tend to last slightly longer (4.2 quarters). In emerging markets, the frequency of a country experiencing a cyclical downturn goes up to 15% but recessions are statistically shorter (3.6 quarters, down from 4.4). Real output recoveries, on the other hand, are signiﬁcantly slower during the globalization period for both industrial countries (increasing from 3.5 to 3.9 quarters) and emerging-market economies (increasing from 3.5 to 4.6 quarters). Hence, it takes more time (between one and three more months) to reach the previous peak after the trough. In terms of expansions, the duration of trough-to-peak phases in the 104 Table 2 Basic features of real output cycles: pre-globalization vs. globalization. Sample of 71 countries, 1970–2007 (quarterly information). Recessions Recoveries Duration (quarters) Amplitude (%, median) Slope (%, median) All countries Pre-globalization (1970–1984) Globalization (1985–2007) 2-sided equality test (p-value) 16.4% 12.0% 3.9 3.8 (0.556) −2.7% −3.2% (0.557) −0.8% −1.0% (0.052) Industrial countries Pre-globalization (1970–1984) Globalization (1985–2007) 2-sided equality test (p-value) 18.6% 9.4% 3.7 4.2 (0.282) −2.4% −1.5% (0.091) Emerging markets Pre-globalization (1970–1984) Globalization (1985–2007) 2-sided equality test (p-value) 11.0% 14.7% 4.4 3.6 (0.059) Country groups by region East Asia and the Paciﬁc (EAP) Pre-globalization (1970–1984) Globalization (1985–2007) Eastern Europe and Central Asia (ECA) Pre-globalization (1970–1984) Globalization (1985–2007) Latin America and the Caribbean Pre-globalization (1970–1984) Globalization (1985–2007) Cum. Loss (%, median) Expansions Number of Events Time (%, median) Duration (quarters) Amplitude (%, median) Slope (%, median) Number of Events Duration (quarters) Amplitude (%, median) Slope (%, median) Number of Events −4.6% −5.5% (0.529) 83 160 28.5% 17.4% 3.5 4.4 (0.080) 4.3% 5.1% (0.148) 1.5% 1.8% (0.563) 71 147 20.7 12.7 (0.000) 15.6% 13.3% (0.474) 1.0% 1.4% (0.003) 84 116 −0.6% −0.5% (0.197) −3.3% −2.4% (0.276) 56 45 28.9% 10.9% 3.5 3.9 (0.491) 3.9% 2.3% (0.007) 1.4% 0.8% (0.150) 54 42 20.6 15.8 (0.243) 15.2% 7.7% (0.104) 0.9% 0.6% (0.002) 59 24 −5.4% −4.7% (0.199) −1.4% −1.4% (0.831) −8.2% −7.2% (0.521) 27 115 27.3% 21.4% 3.5 4.6 (0.313) 6.4% 7.0% (0.601) 3.1% 2.2% (0.601) 17 105 20.9 11.9 (0.007) 21.4% 15.1% (0.618) 1.5% 1.7% (0.687) 25 92 5.1% 10.9% 4.2 3.3 −4.0% −3.8% −1.2% −1.3% −6.8% −6.4% 6 20 14.3% 18.5% 1.6 4.5 14.7% 9.7% 6.7% 2.1% 5 22 48.4 17.1 71.2% 18.5% 2.1% 1.6% 5 15 11.9% 13.3% 7.0 3.1 −4.1% −7.0% −0.6% −1.8% −8.6% −8.7% 1 23 19.2% 21.4% 4.0 4.7 5.2% 1.3% 8.9% 2.8% 1 21 30.0 10.8 45.7% 14.0% 1.5% 1.8% 1 18 21.9% 18.5% 4.5 3.8 −8.2% −4.7% −1.9% −1.3% −18.9% −7.2% 15 53 50.0% 27.9% 5.6 4.8 6.5% 2.0% 6.1% 1.8% 7 44 14.2 10.3 13.8% 11.9% 1.5% 1.5% 15 43 Note: We report the average duration of real output recessions, recoveries, and expansions. The statistics for time in the cyclical phase, amplitude, slope, and cumulative loss refer to the sample median across episodes. Averages for those statistics are not reported but are available from the authors upon request. The duration of real output recessions is the number of quarters from peak to trough, while that of expansions is the number of quarters from trough to peak. Recoveries in economic activity, on the other hand, are deﬁned as the early stage of the expansion (recovery phase), consisting of the period in which output rebounds from a trough to its previous peak. The amplitude of the recession is the distance between a peak in real output and its subsequent trough while that of expansions is the distance between a trough and the next peak in real output. Moreover, the amplitude of the recovery is computed as the 4-quarter cumulative variation in real output following a trough. The slope of the recession (expansion) is the ratio of the amplitude of the peak-to-trough (trough-to-peak) phase of the cycle to its duration. The cumulative loss combines information on the duration and amplitude of the peak-to-trough phase of the cycle to measure the overall cost of the recession. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Time (%, median) C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 business cycle for industrial economies has declined during the globalization period (from 21 to 16 quarters) and it also decreases from 21 to 12 quarters for emerging markets. In sum, recoveries tend to be slower during the globalization period (when compared with the pre-globalization years) and expansions in real output have become shorter for industrial and emerging-market economies. Fact 4: recessions are slightly less deep during the globalization period, while the strength of recoveries diminishes for industrial countries In the pre-globalization era, the median pace of recessions in EMEs (as indicated by the slope of peak-to-trough phases of the cycle) is more than twice as fast as that of industrial economies (− 1.4% and −0.6%, respectively). Recoveries, on the other hand, are twice as fast; that is, 3.1 and 1.4% for EMEs and industrial countries, respectively (see Table 2). Hence, in that era, emerging markets reach the trough of their recessions and exit from them at a faster pace than industrial countries. During the globalization period, recessions are not as deep for either industrial or developing countries. The median amplitude of the peakto-trough phase for industrial countries increases from − 2.4% in 1970–84 to − 1.5% in 1985–2007, and from − 5.4 to − 4.7% over the same time frame for emerging markets. On the other hand, the pace of recessions remains statistically invariant during the globalization period for both industrial and emerging-market economies. The median slope of the recessions in industrial countries increases from −0.6 to −0.5% whereas that of emerging markets stays ﬁxed at −1.4%. For industrial countries, the strength and intensity of the real economic recovery during the globalization period is nearly halved: The amplitude declines from 3.9 to 2.3% whereas the slope falls to 0.8 from 1.4% per quarter. Emerging-market recoveries are slightly sharper, with the median amplitude moving from 6.4% in the pre-globalization period to 7% in 1985–2007. However, the initial recovery phase occurs at a slower pace; that is, the median slope of the recovery during the globalization period is 2.2% (down from 3.1%). Fact 5: the cost of recessions, as measured by either amplitude or cumulative loss of output, is greater in Latin America and Eastern Europe than in East Asia. However, the cost of recessions for LAC falls sharply during the globalization period compared with EAP countries Recessions are shorter and less profound for all regional emergingmarket groups during the globalization period (when compared with the 1970–84 period). As shown in Fig. 2, the duration of recessions decreases from 4.2 to 3.3 quarters in EAP, from 7 to 3.1 quarters in ECA, and from 4.5 to 3.8 quarters in LAC. Moreover, the amplitude of those recessions also declines sharply for LAC (with the median amplitude of the contraction shrinking from − 8.1 to − 4.7%) whereas the reduction is meager for EAP. In terms of the cumulative loss of output, we observe that the cost of recessions declines across all groups. However, the reduction is sharper among LAC countries: The median cumulative output loss drops from nearly 19% in 1970–1984 to approximately 7% in 1985–2007. Interestingly, as Fig. 2 shows, upturns have become slower in East Asia (recoveries take 4.5 quarters as opposed to just under 2 quarters during the pre-globalization period). The amplitude of the recovery has sharply declined across all EM regional groups during 1985–2007; most notably, the median amplitude of the recoveries declines from 14.7% in 1970–84 to 9.7% in 1985–2007 among EAP countries. The drop is also sharp among LAC economies, decreasing from 6.5 to 2% over the same time period. Finally, the length and amplitude of expansionary periods has also declined during the globalization period (from 48 to 17 quarters among EAP countries and from 14 to 10 among LAC economies). 105 3.2. Crisis and business cycles Section 3.1 shows that economic downturns in emerging markets are deeper, steeper, and costlier than those in industrial countries. Across regional groups of emerging markets, recessions are particularly costly in LAC during the pre-globalization period (−18.9%) and EAP in the globalization period (− 6.4%). These periods coincide with the Latin American debt crisis and the East Asian ﬁnancial crisis. Fig. 3 depicts the frequency of recessions by their duration (short, medium, long, and protracted) and their amplitude (mild, moderate, severe, and depression).14 On average, two out of ﬁve recessionary periods in industrial countries and emerging markets last between three and four quarters (regardless of their amplitude). Regardless of their duration, nearly 75% of recessions in industrial countries are mild (0–2%) or moderate (2–4%) while the proportion for emerging markets is below 40%. Finally, the mode of the distribution of recessions is different across groups: mild contractions with medium duration are more frequent among industrial countries whereas severe contractions of medium duration are more likely among emerging markets. The depth of recessions among emerging-market economies (where three out of ﬁve contractions are considered severe or depression) is related to the higher sensitivity of these countries to episodes of turmoil and crisis. Fig. 4 reports the frequency of different types of crises for industrial and emerging-market economies. Regardless of the type of crisis, we ﬁnd that these events are more likely to occur in EMEs. Moreover, these crisis episodes are preceded by overvalued currencies, excess leverage, deposit runs, or balance of payment problems. Overall, sharp output ﬂuctuations associated with crisis episodes are more likely to occur in emerging markets than in developed economies (Calderón and Servén, 2013; Claessens et al., 2009, 2011b, 2012; Tornell and Westermann, 2002). Greater output volatility and proneness to sharp recessions in EMEs (vis-à-vis industrial countries) are tightly linked to other patterns of cyclical behavior that differ from those of industrial countries: (a) greater sensitivity to ﬂuctuations in prices of primary commodities and intermediate goods; (b) the inability of the current account to act as a buffer to smooth consumption, as reﬂected by higher consumption volatility (in excess of output) and a countercyclical trade balance, and (c) procyclical macroeconomic policies resulting from highly volatile and countercyclical interest rates and procyclical government spending. Explanations of these features have been provided in a long list of works pioneered by Mendoza (1991) and Backus et al. (1992), and followed by Kose (2002), Kydland and Zarazaga (2002), Neumeyer and Perri (2005), Uribe and Yue (2006), Aguiar and Gopinath (2007, 2008), and Boz et al. (2008), among others. The empirical literature is very extensive for developed economies; for instance, see Crucini et al. (2011), Centoni et al. (2007) and the references therein. One strand of the literature suggests that productivity shocks are the main driver of business cycle ﬂuctuations. However, in regard to cyclical ﬂuctuation drivers in emerging markets, the literature examines the importance of both domestic and external factors (e.g., Kose et al., 2003).15 Historical analysis of output ﬂuctuations for selected Latin American countries shows that external factors are the main drivers of business cycles in Argentina, Brazil, Chile, and Mexico during periods of inward and outward orientations (Aioilﬁ et al., 2011). Relatedly, there is evidence that: (a) ﬁnancial and political crises tend to drive up the cost of recessions in EMEs (Cerra and Saxena, 2008), (b) the procyclicality of capital ﬂows (and of access to world capital markets) heightens the vulnerability of emerging markets to sudden stops (Calvo, 1998; Mendoza, 14 We follow Morsink et al. (2002) in classifying the severity of recessions by duration and amplitude. In terms of duration, they deﬁne recessions as short (2 quarters), medium (3–4 quarters), long (5–6 quarters), and protracted (more than 7 quarters). According to their depth (amplitude), recessions can be mild (0–2 percent), moderate (2–4 percent), severe (4–10 percent), and depressions (more than 10 percent). 15 Kose et al. (2003) ﬁnd that less developed economies are more likely to experience country-speciﬁc business cycles. 106 C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Business Cycle Features during the globalization period By Region 2.1 Pre-Globalization (1970-1984) 2.2 Globalization (1985-2007) a) Duration (average, in quarters) a) Duration (average, in quarters) 6 6 4 4 2 2 0 0 -2 -2 -4 -4 -6 -6 -8 -8 East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recessions East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recoveries Recessions b) Amplitude (median, %) Recoveries b) Amplitude (median, %) 15% 15% 10% 10% 5% 5% 0% 0% -5% -5% -10% -10% East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recessions East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recoveries Recessions c) Slope (median, %) 10% Recoveries c) Slope (median, %) 10% 8% 8% 6% 6% 4% 4% 2% 2% 0% 0% -2% -2% East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recessions Recoveries East Asia and the Pacific Eastern Europe and Central Latin America and the Asia Caribbean Recessions Recoveries Fig. 2. Business cycle features during the globalization period By Region. 2.1 Pre-globalization (1970–1984). 2.2 Globalization (1985–2007). 2006) and ampliﬁes the output effect of adverse external shocks in countries with domestic ﬁnancial frictions (Caballero, 2002). The tight association between ﬁnancial cycles and real output cycles for a wide array of countries is thoroughly documented in Claessens et al. (2009, 2012). They ﬁnd that ﬁnancial factors (credit, equity, and housing prices) play a role in explaining the duration and amplitude of the different phases of the business cycle. For instance, recessions associated with ﬁnancial disruptions—notably, credit contractions and housing price busts—tend to be deeper and costlier while recoveries in the presence of credit and housing price booms are stronger. This section documents the main features of recessions and recoveries associated with crisis episodes—more speciﬁcally, systemic banking C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 107 Severity of Recessions, 1970-2007 3.2 Emerging-market Economies 0.25 0.20 0.20 0.15 0.15 0.10 0.10 Protracted 0.05 Long Medium 0.00 Short Mild Moderate Duration 0.25 Protracted 0.05 Long Medium 0.00 Short Mild Severe Depression Amplitude Duration 3.1 Industrial Countries Moderate Severe Depression Amplitude Note: The classification of recessions by duration and amplitude follows Morsink, Helbling, and Takorick (2002). The duration of recessions is defined as short (2 quarters), medium (3-4 quarters), long (5-6 quarters), or protracted (more than 7 quarters). The amplitude of recessions is defined as mild (0-2 percent), moderate (2-4 percent), severe (4-10 percent), and depression (more than 10 percent). The figures depict the frequency of recessions in each of these categories for both industrial and emerging market economies. Reference: Morsink, James, Thomas Helbling, and Stephen Tokarick (2002). Recessions and Recoveries, Chapter 3, IMF World Economic Outlook. April Fig. 3. Severity of recessions, 1970–2007. Note: The classiﬁcation of recessions by duration and amplitude follows Morsink et al. (2002). The duration of recessions is deﬁned as short (2 quarters), medium (3–4 quarters), long (5–6 quarters), or protracted (more than 7 quarters). The amplitude of recessions is deﬁned as mild (0–2%), moderate (2–4%), severe (4–10%), and depression (more than 10%). The ﬁgures depict the frequency of recessions in each of these categories for both industrial and emerging market economies. Reference: Morsink et al. (2002). Recessions and Recoveries, Chapter 3, IMF World Economic Outlook. April. crises, currency crises, and what we denote as economic crises. Here we distinguish crisis-related recessions (banking, currency, and economic crises) from recessions that do not coincide with crisis episodes (regular recessions). Frequency of Crisis, 1970-2007 30% 27% 25% 21% 20% 17% 15% 11% 10% 10% 7% 5% 0% 0% Banking Crisis Currency Crisis Industrial Countries 0% External Debt Default Domestic Debt Default Emerging Markets Fig. 4. Frequency of Crisis, 1970–2007. The dating of the different types of crises (banking, currency, external debt default and restructuring, and domestic debt default and restructuring) are taken from Reinhart and Rogoff (2010). The data is publicly available at: http://www.carmenreinhart.com/data/browse-by-topic/topics/7/. 3.2.1. Recessions associated with crisis episodes The greater severity of output contractions in EMEs vis-à-vis industrial countries, depicted in Fig. 3, can be attributed to the fact that EMEs are not only more prone to experience sharp (and adverse) external shocks but also to the structural features of these economies (for example, non-diversiﬁed output and export structures, high-liability dollarization, fragile ﬁnancial systems), which tend to amplify rather than mitigate the effects of these shocks (Becker and Mauro, 2006; Caballero, 2001; Loayza and Raddatz, 2007; Raddatz, 2007). The combination of sharp and more frequent external shocks and the difﬁculty mitigating them results in a greater likelihood that EMEs will experience crisis episodes (Calderón and Servén, 2013; Mendoza and Terrones, 2012; Tornell and Westermann, 2002). Table 3 reports the main features of recessions associated with banking crisis, currency crisis, and economic crisis episodes as well as their ensuing recovery periods. We identify banking crisis episodes using the recent database by Laeven and Valencia (2010, 2012). The authors deﬁne systemic banking crises as those episodes with: (a) ﬁnancial distress in the banking sector signaled by bank runs, losses in the banking system, and bank liquidations, and (b) signiﬁcant intervention policy responses to ﬁnancial distress.16 Currency crisis episodes, on the other hand, are identiﬁed following the Frankel and Rose (1996) methodology based on large exchange-rate depreciations: They occur when the 16 State interventions in the banking sector are signiﬁcant, according to these authors, if at least three of the following take place: (i) the state provides extensive liquidity support (5 percent of deposits and liabilities to nonresidents), (ii) bank restructuring costs are at least 3 percent of GDP, (iii) the state undertakes signiﬁcant bank nationalizations, (iv) the state puts in place signiﬁcant guarantees, (v) the state undertakes signiﬁcant asset purchases (at least 5 percent of GDP), and (vi) the state orders deposit freezes and bank holidays (Laeven and Valencia, 2010, pp. 6–7). 108 Table 3 Basic features of real output cycles: crisis vs. regular phases of the cycle. Sample of 71 countries, 1970–2007 (quarterly information). Recessions Expansions Amplitude (%, median) Slope (%, median) Cum. Loss (%, median) Number of events Duration (quarters) Amplitude (%, median) Slope (%, median) Number of events Duration (quarters) Amplitude (%, median) Slope (%, median) Number of events 3.8 4.0 (0.627) −2.65% −6.75% (0.003) −0.79% −1.88% (0.000) −4.09% −11.26% (0.008) 206 38 3.7 6.2 (0.000) 4.70% 5.93% (0.038) 1.89% 1.05% (0.186) 187 33 19.0 19.8 (0.812) 15.37% 19.94% (0.436) 1.12% 1.46% (0.006) 179 31 3.9 4.7 (0.529) −1.85% −1.37% (0.546) −0.50% −0.39% (0.076) −3.16% −1.57% (0.546) 98 3 3.6 4.7 (0.539) 2.85% 2.40% (0.082) 1.08% 0.47% (0.082) 92 3 22.8 27.3 (0.705) 14.79% 3.67% (0.570) 0.87% 0.46% (0.082) 92 3 3.7 3.9 (0.628) −4.17% −7.06% (0.007) −1.24% −2.05% (0.022) −6.76% −12.09% (0.036) 108 35 3.8 6.4 (0.002) 6.95% 6.13% (0.228) 2.75% 1.42% (0.014) 95 30 15.0 19.0 (0.237) 15.84% 21.57% (0.357) 1.80% 1.52% (0.211) 87 28 3.7 4.2 (0.107) −2.40% −7.12% (0.000) −0.69% −1.70% (0.000) −3.65% −11.39% (0.000) 188 56 3.6 5.6 (0.000) 4.27% 7.65% (0.000) 1.75% 1.81% (0.872) 170 50 19.5 17.9 (0.595) 14.81% 23.13% (0.253) 0.98% 1.84% (0.000) 161 49 3.9 4.8 (0.417) −1.68% −4.29% (0.043) −0.48% −0.93% (0.317) −3.08% −6.35% (0.043) 97 4 3.6 3.3 (0.860) 2.82% 4.09% (0.545) 1.06% 1.05% (0.570) 92 3 23.2 16.0 (0.455) 14.77% 17.50% (0.625) 0.85% 1.21% (0.297) 91 4 3.5 4.2 (0.072) −4.03% −7.61% (0.004) −1.17% −1.83% (0.011) −5.77% −11.66% (0.002) 91 52 3.5 5.8 (0.002) 6.36% 8.37% (0.036) 2.47% 1.86% (0.628) 78 47 14.6 18.1 (0.252) 15.52% 23.53% (0.158) 1.54% 1.84% (0.158) 70 45 3.6 4.2 (0.033) −2.14% −6.13% (0.000) −0.68% −1.56% (0.000) −3.27% −9.97% (0.000) 162 82 3.4 5.3 (0.000) 4.21% 6.63% (0.000) 1.86% 1.70% (0.565) 148 72 19.4 18.5 (0.749) 14.53% 20.70% (0.143) 0.95% 1.69% (0.000) 140 70 3.8 4.7 (0.296) −1.76% −3.03% (0.251) −0.49% −0.46% (0.675) −3.10% −5.85% (0.251) 94 7 3.6 4.0 (0.753) 2.85% 2.60% (0.414) 1.09% 0.69% (0.097) 89 6 23.1 20.9 (0.780) 14.79% 11.86% (0.716) 0.86% 0.77% (0.716) 88 7 3.3 4.2 (0.012) −2.95% −6.68% (0.001) −1.08% −1.80% (0.002) −3.90% −10.81% (0.002) 68 75 3.2 5.5 (0.001) 6.59% 7.04% (0.651) 2.75% 1.81% (0.181) 59 66 13.1 18.3 (0.082) 13.60% 21.15% (0.074) 1.44% 1.75% (0.157) 52 63 Note: The statistics for amplitude, slope, and cumulative loss for recessions, recoveries, and expansions refer to the sample median across countries. Averages are presented for the duration. The duration of recessions is the number of quarters between peak and trough. Real recoveries are deﬁned as the expansion (recovery phase) that takes place during the period where output rebounds from a trough to its previous peak. The amplitude of recessions calculated as the distance between the real output at its peak and its subsequent trough. On the other hand, the amplitude of the recovery is computed as the one-year cumulative variation in real output following a trough. The slope of the recession is the ratio of the amplitude of the peak-totrough phase of the cycle to its duration. The slope of recovery is the amplitude from trough to the previous peak divided by its duration. The cumulative loss combines information on the duration and amplitude to measure the overall cost of the recession. 1/Recessions associated with crisis are deﬁned as those where the crisis (banking, currency, or economic) takes place at the same time or within the 4–6 quarter window before the start of the contractionary period. Banking crises are identiﬁed as in Laeven and Valencia (2010, 2012). 2/Currency crisis episodes are taken from Reinhart and Rogoff (2009). 3/Economic crises is deﬁned as those episodes where at least one of the following events takes place: sovereign domestic default and restructuring, sovereign external debt default and restructuring, banking crisis, and currency crisis. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 I. Financial crisis 1. All countries No Financial Crisis Financial crisis 2-sided equality test (p-value) 2. Industrial countries No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) 3. Emerging markets No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) II. Currency crisis 1. All countries No Financial crisis Financial crisis 2-sided equality test (p-value) 2. Industrial countries No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) 3. Emerging markets No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) III. Economic crisis 1. All countries No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) 2. Industrial countries No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) 3. Emerging markets No ﬁnancial crisis Financial crisis 2-sided equality test (p-value) Recoveries Duration (quarters) C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 year-on-year currency depreciation exceeds 25% and the increase in the rate of nominal depreciation is also at least 10%. Sovereign defaults, as deﬁned in Reinhart and Rogoff (2009), are events where the government 109 is unable to meet principal or interest payments on time—either on the due date or within a speciﬁed grace period. Using Reinhart and Rogoff's dating criteria we distinguish between sovereign defaults on external Business Cycle Features: Regular vs. Crisis-related Episodes, Economic Crisis By Region 5.1 Regular recessions 5.2 Recessions associated to economic crisis a) Duration (average, in quarters) a) Duration (average, in quarters) 6 6 4 4 2 2 0 0 -2 -2 -4 -4 -6 -6 East Asia and the Pacific Eastern Europe and Central Asia Recessions Latin America and the Caribbean East Asia and the Pacific Eastern Europe and Central Asia Recoveries Recessions b) Amplitude (median, %) Recoveries b) Amplitude (median, %) 10% 10% 8% 8% 6% 6% 4% 4% 2% 2% 0% 0% -2% -2% -4% -4% -6% -6% -8% Latin America and the Caribbean -8% East Asia and the Pacific Eastern Europe and Central Asia Recessions Latin America and the Caribbean East Asia and the Pacific Eastern Europe and Central Asia Recoveries Recessions c) Slope (median, %) Latin America and the Caribbean Recoveries c) Slope (median, %) 3% 3% 2% 2% 1% 1% 0% 0% -1% -1% -2% -2% -3% -3% East Asia and the Pacific Eastern Europe and Central Asia Recessions Recoveries Latin America and the Caribbean East Asia and the Pacific Eastern Europe and Central Asia Recessions Latin America and the Caribbean Recoveries Fig. 5. Business cycle features: regular vs. crisis-related episodes, economic crisis by region. 5.1 Regular recessions. 5.2 Recessions associated to economic crisis. 110 C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 debt and domestic debt.17 Finally, this paper identiﬁes episodes of economic crisis as those where at least one of the following types of crisis takes place: (a) banking crisis, (b) currency crisis, (c) sovereign external debt default, or (d) sovereign domestic debt default. In this section, we consider that recessions are associated with crisis episodes if the peak in GDP, which marks the start of the recession period, coincides with a dated crisis within a quarter interval (T − 4, T + 4, where T is the quarter where the peak in real GDP takes place). As shown in Table 3, crisis-related recessions have, on average, a duration similar to those unrelated to crisis; however, they tend to be deeper (greater amplitude), more intense (steeper slope) and costlier (larger cumulative real output losses). The median peak-to-trough change in real GDP for recessions associated with banking crises is approximately −6.8% (−2.7 in regular recessions), and their speed is nearly three times as fast as those of regular recessions (−1.9 vs. −0.7% per quarter, respectively). Overall, recessions associated with currency crises are costlier, as reﬂected by the greater cumulative output loss (11.4% relative to 3.7% in regular recessions). On the other hand, recoveries following crisis-related recessions (regardless of the type of crisis) tend to be slower than other upturns; the average duration of recoveries after banking crises is 6.2 quarters while that of regular recessions is approximately 3.7 quarters. They also tend to be stronger, with real output growing 5.9% from trough in recessions associated with banking crises as opposed to 4.7% following other recessions. Next, we focus our discussion on the differences between recessions and recoveries associated with what we generally refer to in this paper as economic crisis episodes. Recessions associated with economic crisis, on average, tend to last longer in industrial countries than in emerging markets (4.7 and 4.2 quarters, respectively). Nonetheless, in the event of a crisis, the downturn in industrial countries is smaller (3%) than the output drop in emerging markets (6.7%). The drop in real output is also steeper among emerging markets when compared with industrial countries, with declines of 1.8 and 0.5% per quarterly respectively. Overall, recessions associated with economic crisis in emerging markets are costlier than in industrial countries (with a cumulative loss of −11 vs. −5.9%, respectively). On the other side of the coin, recoveries are stronger for emerging markets when they follow an economic crisis. The amplitude of the recovery after crisis is 7% (as opposed to 6.6% when there is no crisis). However, this recovery is slower (it takes 5.5 quarters to reach the previous peak vs. 3.2 quarters when there is no crisis) and less intense (the speed of recovery drops from 2.8 to 1.8% per quarter). Focusing on the regional groups of emerging markets, the average duration of recessions in LAC (nearly 4 quarters) does not appear to be inﬂuenced by the occurrence of a crisis (Fig. 5). In the case of EAP and ECA countries, recessions are longer in duration when associated with economic crisis (they last almost 5 and 4 quarters as opposed to nearly 3 and 2.5 quarters, respectively). The depth of recessions tends to be deeper and more intense when they coincide with a crisis episode. For instance, the amplitude of peak-to-trough phases in EAP during crisis (− 6.7%) is larger than otherwise (− 2.8%). The same holds for ECA (−7.3 vs. −2.7%) and LAC (−4.1 vs. −6.5%). Recoveries after crisis episodes in Latin America and the Caribbean are slower (they take almost 5.5 quarters) than other upturns (approximately 3.5 quarters), as shown in Fig. 5. However, they are relatively stronger (8.9 vs. 7.6%). In the case of East Asia and the Paciﬁc, recoveries are fast: It takes almost 2 quarters for real output to reach its previous peak from the trough.18 Moreover, EAP recoveries following an economic crisis are larger (6.5 vs. 5.0% otherwise). Finally, Eastern Europe and Central Asia also experience faster and stronger recoveries after crisis-related recessions (see Fig. 5). 17 Episodes of sovereign default on external debt include debt rescheduling that is eliminated with less favorable terms than the original liability. Default on domestic debt includes the freezing of bank deposits and forced conversions of those deposits from foreign to local currency (Reinhart and Rogoff, 2009, pp. 11). 18 However, that is not the case when a ﬁnancial crisis takes place (it takes almost 10 quarters for real economic activity to recuperate). We should point out that recoveries following a systemic banking crisis deserve special attention in EAP: (a) It takes almost 10 quarters for these countries' real GDP to reach the pre-crisis peak, (b) the amplitude of recoveries is similar in crisis and non-crisis periods (around 7.5%), and (c) the rhythm of the recovery is slow, with the median slope declining from 3.1% per quarter to 1% per quarter. Clearly, the sharp output ﬂuctuations experienced in EAP during episodes of systemic banking crisis correspond to the 1997–98 ﬁnancial crisis experienced by the region. In fact, the peak-to-trough drop in real output during the crisis is nearly 20% in Indonesia, 16% for Thailand, 12% for Malaysia, and 9% for Hong Kong. Overall, EAP countries experience the most severe recessions when banking crises ensue, but they also experience the strongest rebound from crisis-related recessions among EMEs. 3.2.2. Dynamic behavior of real, ﬁnancial, and external indicators around recessions For this analysis, we run panel data regressions of real and ﬁnancial variables on time effects for a 17-quarter window centered on the peak of real GDP (i.e., where period T denotes the start of the recession) and distinguishing between peaks associated with crisis and peaks that are not. These regressions are conducted for the sample of industrial countries and emerging market economies over the period 1970q1–2007q4, and the coefﬁcient estimates of these regressions are depicted in Fig. 6 for recessions associated with banking crises.19 We interpret our coefﬁcient estimates as deviations from the average growth outside the fouryear window associated with the crisis episode. For the sake of simplicity, we refer to this average growth outside the window of analysis as trend growth. Fig. 6 shows the evolution of year-on-year growth in output, private consumption, real investment, domestic credit to the private sector (in per-capita terms and as a percentage of GDP), stock prices (in real terms), and the real exchange rate. By construction, the dynamic behavior of growth in real output around a recession is as expected for both industrial and emerging-market economies. After reaching a peak above trend in period T, real output drops below average in T + 1 and reaches its trough four quarters after the start of the recession (T + 4) for industrial countries and emerging-market economies. At the trough, output growth is 4 to 5 percentage points below trend for industrial countries regardless of whether the recession is associated with a banking crisis. On the other hand, the trough for emerging markets is almost 12 percentage points below trend in recessions with crisis (compared with 8 percentage points below trend for downturns without crisis). Once it reaches its trough in episodes associated with crisis, real output growth tends to converge to trend growth at a faster pace for emerging markets than for industrial countries (see Figs. 6.1a and 6.2a). The behavior patterns of private consumption and investment around peaks in real GDP is depicted in Figs. 6.1b and 6.1c for industrial countries and Figs. 6.2b and 6.2c for emerging markets, respectively. Qualitatively, we observe that the behavior pattern of private consumption and investment for both groups of countries mimics that of real output; that is, the trough in consumption and investment takes place in period T + 4. We should point out not only that the drop in consumption is smaller than that of output (relative to its trend growth) but also that the decline in investment is sharper than that of output for both samples of countries. In addition, the trough in consumption and investment in emerging markets is deeper compared with that of industrial countries (3.5 and 7 percentage-point differences, respectively).20 Consumption and investment tend to converge to trend growth at a faster pace for 19 Although the regressions are not reported here, they are available from the authors upon request. 20 Empirical evidence shows that households in EMEs have a low cost of substitution between market time and unmeasured home time (Deaton, 2005; Hicks, 2012). This creates artiﬁcially higher consumption volatility in EMEs relative to industrial countries. Moreover, the existence of a large informal sector in EMEs may create larger income volatility if recessions induce movements from formal to informal sectors. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 emerging markets than for industrial countries as a result of a stronger rebound effect. We next take a look at the dynamic behavior of credit and asset prices around peaks in real GDP. Our goal is to ascertain whether there is statistical precedence of credit and/or asset prices to real output cycles. In this context, Fig. 6 shows the evolution of real credit per capita (Figs. 6.1d and 6.2d), real stock prices (Figs. 6.1e and 6.2e) and the real 111 exchange rate (Figs. 6.1f and 6.2f) for industrial countries and emerging markets around recessions that are associated with banking crises and those that are unrelated with episodes of ﬁnancial turmoil. Growth in real credit is above trend in the run-up to the recession up to period T (T − 1) for industrial countries (emerging-market economies) and it ﬂuctuates below trend after 2 quarters. However, it still remains below average for industrial countries while it steadily declines Recession and Banking Crisis: Event Analysis 6.1 Industrial Countries 6.2 Emerging markets -0.02 -0.02 -0.04 -0.04 -0.06 -0.06 -0.08 -0.08 -0.10 -0.10 -0.12 -0.12 No Crisis No Crisis Crisis T+7 Crisis -0.06 -0.08 -0.08 -0.05 -0.05 -0.10 -0.10 -0.15 -0.15 -0.20 -0.20 -0.25 -0.25 T+5 T+6 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 No Crisis Fig. 6. Recession and banking crisis: event analysis. 6.1 Industrial countries. 6.2 Emerging markets. Crisis T+5 T+6 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 T-8 T+7 T+8 T+6 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 T-6 0.00 T-7 0.00 T-8 0.05 Crisis Crisis c) Real Investment (year-on-year growth) 0.05 T-6 c) Real Investment (year-on-year growth) No Crisis T-6 No Crisis Crisis T-7 No Crisis T+8 -0.06 T+8 -0.04 T+7 -0.04 T+7 -0.02 T-7 T-8 T+7 T+8 T+6 T+5 T+4 T+2 T+3 T+1 T T-1 T-2 T-3 T-4 0.00 T-5 0.00 T-6 0.02 T-7 b) Private Consumption (year-on-year growth) 0.02 T-8 b) Private Consumption (year-on-year growth) -0.02 T+8 T+6 T+5 T+3 T+4 T+2 T T+1 T-1 T-3 T-2 T-5 T-4 T-8 T+8 T+7 T+5 T+6 T+4 T+3 T+2 T T+1 T-1 T-2 T-3 T-4 T-5 T-7 T-6 0.00 T-8 0.00 T-6 a) Real GDP (year-on-year growth) 0.02 T-7 a) Real GDP (year-on-year growth) 0.02 112 C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Recession and Banking Crisis: Event Analysis 6.2 Emerging markets -0.10 -0.15 -0.15 -0.20 -0.20 -0.25 -0.25 -0.10 -0.10 -0.20 -0.20 -0.30 -0.30 -0.40 T+7 T+5 T+6 T+3 T+4 T+2 T T+1 T-1 T-3 T-2 T+8 T+7 T+5 T+6 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 T-6 T-7 T-8 T+8 T+7 T+6 T+5 T+4 T+3 T+2 T+1 0.00 T 0.00 T-1 0.10 T-2 0.10 T-3 0.20 T-4 0.20 T-5 Crisis e) Stock prices (year-on-year growth) 0.30 T-6 T-5 No Crisis e) Stock prices (year-on-year growth) T-7 T-4 Crisis 0.30 T-8 T-6 -0.05 -0.10 No Crisis T-7 T-8 T+8 T+7 T+5 T+6 T+4 T+3 -0.05 T+2 T 0.00 T+1 0.00 T-1 0.05 T-2 0.05 T-3 0.10 T-4 0.10 T-5 0.15 T-6 0.15 T-8 d) Real credit per capita (year-on-year growth) T-7 d) Real credit per capita (year-on-year growth) T+8 6.1 Industrial Countries -0.40 Crisis No Crisis -0.05 -0.05 -0.10 -0.10 T+8 T+7 T+6 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 T-6 T+8 T+7 T+6 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 0.00 T-4 0.00 T-5 0.05 T-6 0.05 T-7 0.10 T-8 Crisis f) Real exchange rate (year-on-year growth) 0.10 T-7 f) Real exchange rate (year-on-year growth) T-8 No Crisis -0.15 -0.15 No Crisis No Crisis Crisis Crisis Fig. 6 (continued). after 8 quarters for emerging markets. We should point out not only that deviations from trend in the growth of real credit per capita are larger in recessions associated with banking crises than otherwise, but also that these deviations are larger in emerging markets than industrial countries (Figs. 6.1d and 6.2d). Growth in stock prices (in real terms) decline below trend 2 (3) quarters before the start of the recession for the group of industrial (emerging-market) economies; that is, in period T − 2 (T − 3). Stock prices reach their trough in period T + 2 for industrial countries at more than 40 percentage points below trend in recessions associated with crisis (as opposed to 20 percentage points below in recessions without crises). Also note that stock prices grow above trend in period T + 6 for industrial countries during regular recessions and they take more time to recover during recessions that coincide with banking C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 113 Proportion of countries in a peak-to-trough cyclical phase, 1990-2012 Sample: 71 countries (quarterly information) 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Industrial Countries 2012q1 2011q1 2010q1 2009q1 2008q1 2007q1 2006q1 2005q1 2004q1 2003q1 2002q1 2001q1 2000q1 1999q1 1998q1 1997q1 1996q1 1995q1 1994q1 1993q1 1992q1 1991q1 1990q1 0.0 Emerging Market Economies Note: The gray shaded area represents the recessionary phases in the United States business cycle as identified by the BBQ algorithm. Fig. 7. Proportion of countries in a peak-to-trough cyclical phase, 1990–2012. Sample: 71 countries (quarterly information). Note: The gray shaded area represents the recessionary phases in the United States business cycle as identiﬁed by the BBQ algorithm. crises. The same qualitative behavior holds for emerging markets although stock prices recover sharply by T + 8 in recoveries associated with crisis periods (see Figs. 6.1e and 6.2e).21 Finally, ﬂuctuations from trend in real exchange rates are more volatile around recessions with crisis than recessions without crisis. Interestingly, the real exchange rate appreciates signiﬁcantly in recessions with banking crises among industrial countries—reaching its peak in period T + 4 at 5 pp above the trend growth. On the other hand, emerging markets experience a sharp depreciation of their currency in real terms, which reaches its trough in period T + 5 at approximately 15 pp below trend growth (see Figs. 6.1f and 6.2f). This ﬁnding may reﬂect the fact that crises in EMEs are usually accompanied by capital repatriation to advanced economies, which induces the real depreciation of the local currency in EMEs. 3.3. The recent global ﬁnancial crisis: is this time different? So far, we have conducted an analysis of the main features of real output cycles for industrial and emerging-market economies using quarterly data up to 2007. However, the recent global ﬁnancial crisis deserves special attention. This section compares the main features of real output ﬂuctuations during 2007–2012 vis-à-vis the 1990–2006 period. The ﬁrst year of the global ﬁnancial crisis (GFC) period was determined based on the appearance of early manifestations of the crisis in 2007, with liquidity disruptions in the interbank lending market of advanced economies. The full-blown, global credit collapse occurred in 2008q3– 2009q1, when almost 80% of the countries in the sample experienced a downturn phase in real credit (see Fig. 7). Table 4 reports the average duration, median amplitude, and slope of recessions and recoveries for industrial countries and emerging markets during the 1990–2006 and 2007–2012 periods. First, we ﬁnd that the average duration of recessions among industrial countries during the 21 As mentioned earlier, Claessens et al. (2009, 2011b, 2012) show that recessions are more closely related to credit contractions and housing price busts than stock market crashes. GFC is nearly one quarter longer than that of the pre-GFC period (4.9 vs. 4 quarters, respectively). For EMEs, the average duration of recessions is roughly similar across sub-periods. However, we should point out that the average duration increases for ECA, while it diminishes for the LAC and EAP regions. The amplitude of recessions during the GFC period is more than three times larger than the pre-GFC downturns among industrial economies (−5.1 and −1.5%, respectively). In contrast, the amplitude of the recent recession is −7.3% in EMEs, which is not quite twice the median downturn in the previous period (−4.2%). It is important to note that the severity of the recession in EMEs when compared with advanced economies is mainly inﬂuenced by EAP countries (−7.8%) and Eastern Europe (−10%). The amplitude of the recession in Latin America in 2007–2012 is roughly of a similar order of magnitude as that of 1990–2006 (−4.4 and −4.1%, respectively).22 Finally, the cost of recessions has signiﬁcantly increased for industrial countries: The cumulative output loss during recessions rises sharply from 2.3% in 1990–2006 to 9.4% in 2007– 2012. For emerging markets, recessions also become costlier during the GFC period (climbing from 7 to 9.8%) but this increase is not statistically signiﬁcant. We should note that of the regional EME groups, the cost of recessions almost triples for ECA (from 8.7 to 24.2%). Fig. 8 shows the trajectory of real GDP in 2007–2012 for industrial and emerging-market economies as well as for regional groups of EMEs. The ﬁgure clearly shows that the recession in industrial countries has been as deep as that of emerging markets but that the ensuing recovery in the former group has been not only slower but also weaker. Interestingly, the behavior of real GDP across regional EMEs provides evidence of a heterogeneous response to the crisis. While LAC appears to have dodged the crisis and resumed robust growth rapidly, the recovery in ECA is still a work in progress. A closer look at the recovery phase in 2007–2012 shows that the number of quarters needed for an industrial country to move from a trough to its previous peak increases from nearly 4 quarters in the 22 An analogous result is obtained when looking at the slope of the recession. 114 C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Table 4 Basic features of real output cycles: is this time around different? Sample of 71 countries, 1990–2012 (quarterly information). Recessions Recoveries Time (%) Duration All countries Previous cycle (1990q1–2006q4) Global ﬁnancial crisis (2007q1–2012q4) 2-sided equality test (p-value) 11.8% 20.8% 3.7 4.3 (0.029) −2.73% −5.71% (0.013) −0.96% −1.38% (0.000) Industrial countries Previous cycle (1990q1–2006q4) Global ﬁnancial crisis (2007q1–2012q4) 2-sided equality test (p-value) 7.4% 33.3% 4.0 4.9 (0.118) −1.51% −5.09% (0.007) Emerging market economies Previous cycle (1990q1–2006q4) Global ﬁnancial crisis (2007q1–2012q4) 2-sided equality test (p-value) 13.8% 16.7% 3.5 4.0 (0.213) EM Country groups by region East Asia and the Paciﬁc (EAP) Previous cycle (1990q1–2006q4) Global ﬁnancial crisis (2007q1–2012q4) Eastern Europe and Central Asia (ECA) Previous cycle (1990q1–2006q4) Global ﬁnancial crisis (2007q1–2012q4) Latin America and the Caribbean Previous cycle (1990q1–2006q4) Global ﬁnancial crisis (2007q1–2012q4) Amplitude Slope Cum. Loss Events Time (%) Duration Amplitude Slope Events −4.39% −9.59% (0.002) 138 66 19.7% 33.3% 4.3 4.9 (0.423) 5.12% 5.12% (0.201) 1.85% 1.29% (0.885) 128 38 −0.46% −1.14% (0.000) −2.25% −9.37% (0.000) 39 25 12.5% 37.5% 3.8 6.5 (0.009) 2.37% 2.61% (0.004) 0.86% 0.54% (0.303) 37 11 −4.23% −7.04% (0.185) −1.36% −1.92% (0.016) −6.98% −9.81% (0.264) 99 41 21.1% 25.0% 4.6 4.2 (0.690) 7.06% 7.44% (0.721) 2.17% 2.97% (0.348) 91 27 11.8% 8.3% 3.6 3.0 −3.07% −7.85% −1.27% −2.30% −6.44% −10.13% 23 17 14.7% 16.7% 4.2 3.0 7.7% 11.1% 2.3% 3.9% 17 8 14.3% 33.3% 3.1 5.2 −7.01% −10.03% −1.82% −1.92% −8.73% −24.20% 18 8 23.2% 45.8% 4.9 6.7 8.5% 5.0% 2.7% 0.7% 20 3 14.0% 12.5% 3.7 3.5 −4.07% −4.43% −1.14% −1.20% −6.28% −6.50% 41 10 27.8% 16.7% 4.9 4.3 6.1% 6.7% 1.7% 3.0% 37 10 Note: The statistics for time, amplitude, slope, and cumulative loss for recessions and recoveries refer to the sample median across countries. Averages are presented for the duration. The duration of recessions is the number of quarters between peak and trough. Real recoveries are deﬁned as the expansion (recovery phase) that takes place during the period where output rebounds from a trough to its previous peak. The amplitude of recessions is calculated as the distance between real output at its peak and at its subsequent trough. On the other hand, the amplitude of the recovery is computed as the one-year cumulative variation in real output following a trough. The slope of the recession is the ratio of the amplitude of the peak-to-trough phase of the cycle to its duration. The slope of recovery is the amplitude from trough to the previous peak divided by its duration. The cumulative loss combines information on the duration and amplitude to measure the overall cost of the recession. Evolution of Real GDP around the Global Financial Crisis, 2007-2012 (Index 2008q2=100) 120 115 110 105 100 95 90 2007q1 2007q3 2008q1 2008q3 2009q1 2009q3 Industrial Countries 2010q1 2010q3 2011q1 2011q3 2012q1 2012q3 Emerging Markets Note: We depict the regional median real GDP index. The gray shaded area represents the quarters where more than 40 percent of countries in our sample share a recessionary phase. Fig. 8. Evolution of real GDP around the global ﬁnancial crisis, 2007–2012. (Index 2008q2 = 100). Note: We depict the regional median real GDP index. The gray shaded area represents the quarters where more than 40% of countries in our sample share a recessionary phase. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 pre-GFC period to 6.5 quarters in the GFC period (see Table 4). We should point out that the average duration of recoveries among industrial countries in the GFC period is underestimated, as many countries in this group have yet to reach the pre-GFC peak. The median amplitude 115 of the recovery among industrial countries in 2007–2012 is slightly larger than before (2.6 vs. 2.3%) but pales in comparison with the larger drop experienced in the latter period. The same cannot be said of emerging markets: Their recoveries are stronger (nearly 7.5%) and Business Cycle Features: Is this time around different? By Region 9.1 Real cycles, 1990-2006 9.2 Real cycles during the global financial crisis, 2007-12 a) Duration (average, in quarters) a) Duration (average, in quarters) 8 8 6 6 4 4 2 2 0 0 -2 -2 -4 -4 -6 -6 East Asia and the Pacific Eastern Europe and Central Asia Recessions Latin America and the Caribbean East Asia and the Pacific Eastern Europe and Central Asia Recoveries Recessions b) Amplitude (median, %) 12% 10% 10% 8% 8% 6% 6% 4% 4% 2% 2% 0% 0% -2% -2% -4% -4% -6% -6% -8% -8% -10% -10% Recessions Recoveries b) Amplitude (median, %) 12% East Asia and the Pacific Eastern Europe and Central Asia Latin America and the Caribbean Latin America and the Caribbean East Asia and the Pacific Eastern Europe and Central Asia Recoveries Recessions c) Slope (median, %) Latin America and the Caribbean Recoveries c) Slope (median, %) 4% 4% 3% 3% 2% 2% 1% 1% 0% 0% -1% -1% -2% -2% -3% -3% East Asia and the Pacific Eastern Europe and Central Asia Recessions Recoveries Latin America and the Caribbean East Asia and the Pacific Eastern Europe and Central Asia Recessions Latin America and the Caribbean Recoveries Fig. 9. Business cycle features: is this time around different? By region. 9.1 Real cycles, 1990–2006. 9.2 Real cycles during the global ﬁnancial crisis, 2007–12. 116 C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 more intense (approximately 3% per quarter). For the group of emerging markets, Fig. 9 shows that the amplitude of recoveries in Latin America is slightly larger in the GFC period than before (6.7 vs. 6.1%), East Asia experiences a stronger recovery in the GFC period (11.1 vs. 7.7%), and Eastern Europe is going through a slower recovery when compared with its corresponding pre-GFC period (5 vs. 8.5%). Overall, the recent global ﬁnancial crisis produces longer and deeper recessions among industrial countries and emerging markets. The ensuing recoveries have been longer and slower-paced among industrial Recession and the Global Financial Crisis: Event Analysis -0.02 -0.02 -0.04 -0.04 -0.06 -0.06 -0.08 -0.08 T+6 T+5 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 T-6 T+6 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 0.00 T-3 0.00 T-4 0.02 T-5 (a) Real GDP (year-on-year growth) T-6 (a) Real GDP (year-on-year growth) 0.02 T+4 10.2 Emerging markets T+3 10.1 Industrial Countries -0.10 -0.10 Output Cycles, 1990-2006 Output Cycles, 1990-2006 Global Financial Crisis Global Financial Crisis -0.04 -0.04 -0.06 -0.06 -0.08 -0.08 -0.10 -0.10 -0.12 -0.12 Output Cycles, 1990-2006 -0.10 -0.10 -0.15 -0.15 -0.20 -0.20 -0.25 -0.25 -0.30 -0.30 T+4 T+3 T+2 T T+1 T-1 T-2 T-3 T-4 T-5 T-6 T+6 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 0.00 T-4 0.00 T-5 0.05 T-6 (c) Real Investment (year-on-year growth) 0.05 -0.05 T+6 Global Financial Crisis (c) Real Investment (year-on-year growth) -0.05 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 Output Cycles, 1990-2006 Global Financial Crisis T+5 -0.02 T+6 -0.02 T-6 T+6 T+5 T+4 T+2 T+3 T+1 T T-1 T-3 0.00 T-2 0.00 T-4 0.02 T-5 (b) Private Consumption (year-on-year growth) 0.02 T-6 (b) Private Consumption (year-on-year growth) -0.35 -0.35 Output Cycles, 1990-2006 Global Financial Crisis Output Cycles, 1990-2006 Global Financial Crisis Fig. 10. Recession and the global ﬁnancial crisis: event analysis. 10.1 Industrial countries. 10.2 Emerging markets. C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 117 Recession and the Global Financial Crisis: Event Analysis 10.1 Industrial Countries 10.2 Emerging markets -0.10 -0.15 -0.15 -0.20 -0.20 -0.25 -0.25 -0.30 -0.30 Global Financial Crisis -0.20 -0.20 -0.30 -0.30 -0.40 -0.40 -0.50 -0.50 -0.60 -0.60 -0.70 -0.70 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-5 -0.10 T-6 T+6 T+5 T+4 T+3 T+2 T T+1 T-1 T-2 0.00 T-3 0.10 0.00 T-4 0.10 T-5 (e) Stock prices (year-on-year growth) T-6 (e) Stock prices (year-on-year growth) -0.80 -0.80 Output Cycles, 1990-2006 Global Financial Crisis Output Cycles, 1990-2006 Global Financial Crisis (f) Real exchange rate (year-on-year growth) (f) Real exchange rate (year-on-year growth) 0.10 0.10 0.08 0.08 0.06 0.06 0.04 0.04 0.02 0.02 -0.04 -0.04 -0.06 -0.06 -0.08 -0.08 T+6 T+5 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-6 -0.02 T-5 T+6 T+5 T+4 T+3 T+2 T T+1 T-1 T-2 T-3 T-4 T-5 0.00 T-6 0.00 T+6 Output Cycles, 1990-2006 T+5 Global Financial Crisis T+4 Output Cycles, 1990-2006 -0.02 T+6 T+5 -0.35 -0.35 -0.10 T+4 T+3 T+2 T+1 T T-1 T-2 T-3 T-4 T-6 -0.05 -0.10 T-5 T+6 T+5 T+4 T+3 T T+2 -0.05 T+1 0.00 T-1 0.05 0.00 T-2 0.10 0.05 T-3 0.10 T-4 0.15 T-5 (d) Real credit per capita (year-on-year growth) T-6 (d) Real credit per capita (year-on-year growth) 0.15 -0.10 -0.10 Output Cycles, 1990-2006 Output Cycles, 1990-2006 Global Financial Crisis Global Financial Crisis Fig. 10 (continued). countries, and slightly shorter and fast-paced among emerging markets. Within the EME regional groups, ECA experiences the deepest recessions and this group's recovery is weak. EAP also suffers a sharp recession (more than twice as deep as those in the pre-GFC period) but recovers almost twice as fast. Finally, LAC's recession is, on average, similar to previous episodes in 1990–2006. However, the recovery is not only slightly shorter than before (by two months) but also more intense. The deeper recession experienced by ECA countries is partly attributed to their close trade and ﬁnancial connections with Western Europe as well as their own ﬁscal and external imbalances, which render the region more vulnerable than EAP and LAC. Fig. 10 presents the evolution of the real and ﬁnancial variables around the date of the crisis. As mentioned before, output contractions during the global ﬁnancial crisis have been deeper than previous ones, and this pattern of behavior holds for consumption and investment. While these two variables approach trend growth faster after six 118 Table A.1 Sample of countries, sources of data for quarterly real GDP. Code Name Region Income Start End Denomination 1/ Source 2/ ARG ARM AUS AUT BEL BGR BOL BRA BWA CAN CHE CHL CHN COL CRI CYP CZE DEN DEU DOM ECU EGY ESP EST FIN FRA GBR GRC HKG HRV HUN IDN IND IRL IRN Argentina Armenia Australia Austria Belgium Bulgaria Bolivia Brazil Botswana Canada Switzerland Chile China Colombia Costa Rica Cyprus Czech Republic Denmark Germany Dominican Republic Ecuador Egypt Spain Estonia Finland France United Kingdom Greece Hong Kong, China Croatia Hungary Indonesia India Ireland Iran AMER ECA 1IND 1IND 1IND ECA AMER AMER SSA 1IND 1IND AMER EAP AMER AMER MENA ECA 1IND 1IND AMER AMER MENA 1IND ECA 1IND 1IND 1IND 1IND EAP ECA ECA EAP SA 1IND MENA UMC LMC HIC_OECD HIC_OECD HIC_OECD UMC LMC UMC UMC HIC_OECD HIC_OECD UMC LMC UMC UMC HIC UMC HIC_OECD HIC_OECD UMC LMC LMC HIC_OECD UMC HIC_OECD HIC_OECD HIC_OECD HIC_OECD HIC UMC UMC LMC LMC HIC_OECD LMC 1970.1 1994.1 1970.1 1970.1 1970.1 1994.1 1990.1 1975.1 1994.1 1970.1 1970.1 1977.1 1979.1 1977.1 1991.1 1995.1 1993.1 1970.1 1970.1 1980.1 1970.1 2002.1 1970.1 1993.1 1970.1 1970.1 1970.1 2000.1 1973.1 1993.1 1995.1 1970.1 1990.1 1970.1 1988.1 2012.4 2011.4 2012.4 2012.4 2010.1 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2012.4 2011.4 Mill. 1993 ARS Mill. 2005 AMD Mill. 2009–10 AUD (Ch.) Mill. 2005 EUR (Ch.) Mill. 2005 EUR (Ch.) Mill. 2005 BGN Thou. 1990 BOB Mill. 2005 BRL Mill. 2006 BWP Mill. 2007 CAD (Ch.) Mill. 2005 CHF Mill. 2008 CLP (Ch.) Bill. 2000 CNY Bill. 2005 COP Mill. 1991 CRC Mill. 2005 EUR (Ch.) Mill. 2005 CZK (Ch.) Mill. 2005 DKK (Ch.) Mill. 2005 EUR (Ch.) Mill. 1991 DOP (Ch) Thou. 2007 USD Bill. 2011–12 EGP Mill. 2005 EUR (Ch.) Mill. 2005 EUR (Ch.) Mill. 2005 EUR (Ch.) Mill. 2005 EUR (Ch.) Mill. 2005 GBP (Ch.) Mill. 2005 EUR (Ch.) Mill. 2010 HKD (Ch.) Mill. 2010 HRK (Ch.) Mill. 2005 HUF (Ch.) Bill. 2000 IDR Bill. FY 2004 INR Mill. 2005 EUR (Ch.) Bill. 1997–98 IRR DS, HA, Indec (NSO) DS, HA, Central Bank of Armenia Reserve Bank of Australia OECD, European Central Bank OECD, European Central Bank OECD, European Central Bank National Statistical Institute of Bolivia (NSO) DS, HA, IGBE (NSO) HA, Statistics Botswana DS, Statistics Canada DS, Swiss State Secretariat for Economic Affairs (SECO) DS, HA, Central Bank of Chile DS, China National Bureau of Statistics (NSO) HA, DANE Colombia (NSO) Central Bank of Costa Rica OECD, European Central Bank OECD, European Central Bank OECD, European Central Bank OECD, European Central Bank HA, Central Bank of Dominican Republic DS, Central Bank of Ecuador DS, Central Bank of Egypt OECD, European Central Bank OECD, European Central Bank OECD, European Central Bank OECD, European Central Bank OECD, European Central Bank Hellenic Statistical Authority HA, Hong Kong Census and Statistics Department DS, European Central Bank European Central Bank HA, Badan Pusan Statistik (Statistics Indonesia) HA, Reserve Bank of India OECD, European Central Bank HA, Central Bank of Iran C. Calderón, J.R. Fuentes / Journal of Development Economics 109 (2014) 98–123 Annex Iceland Israel Italy Japan Kazakhstan Korea, Rep. Lithuania Luxembourg Latvia Macao, China Morocco Mexico Malta Malaysia Netherlands Norway New Zealand Panama Peru Philippines Poland Portugal Paraguay Romania Russian Federation Singapore El Salvador Serbia Slovak Republic Slovenia Sweden Thailand Turkey Taiwan, China Ukraine Uruguay United States Venezu

Have Business Cycles Changed? An Empirical Investigation PDF

Document Details

Tags

Related

Summary

Full Transcript

Upgrade to continue