Policy Research Working Paper 10354 Potential Growth A Global Database Sinem Kilic Celik M. Ayhan Kose F. Ohnsorge F. Ulrich Ruch Development Economics Prospects Group March 2023 Policy Research Working Paper 10354 Abstract Potential growth—the rate of expansion an economy can over time. In 2011–21, potential growth was below its sustain at full capacity and employment—is a critical driver 2000–10 average in nearly all advanced economies and of development progress. It is also a major input in the roughly 60 percent of emerging market and developing formulation of fiscal and monetary policies over the busi- economies. Second, adverse events, such as the global finan- ness cycle. This paper introduces the most comprehensive cial crisis and the COVID-19 pandemic, contributed to database to date, covering the nine most commonly used the decline. At the country-level also, national recessions measures of potential growth for up to 173 countries over lowered potential growth even five years after their onset. 1981–2021. Based on this database, the paper presents three Third, the persistent impact of recessions on potential findings. First, all measures of global potential growth show growth operated through weaker growth of investment, a steady and widespread decline over the past decade, with employment, and productivity. all the fundamental drivers of growth losing momentum This paper is a product of the Prospects Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at skiliccelik@imf.org, akose@worldbank.org, fohnsorge@worldbank.org, and fruch@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Potential Growth: A Global Database Sinem Kilic Celik, M. Ayhan Kose, F. Ohnsorge, and F. Ulrich Ruch 1 Keywords: production function; filters; growth expectations; developing economies. JEL Classification: E30, E32, E37; O20 1 Kilic Celik (IMF; skiliccelik@imf.org), Kose (Prospects Group, World Bank; Brookings Institution; CEPR; CAMA; akose@worldbank.org); Ohnsorge (Prospects Group, World Bank; CEPR; CAMA; fohnsorge@worldbank.org); Ruch (Prospects Group, World Bank; fruch@worldbank.org). Kaltrina Temaj provided outstanding research assistance. We thank Amat Adarov, Carlos Arteta, Martin Bailey, Eduardo Borensztein, Ajai Chopra, Kevin Clinton, Brahima Coulibaly, Antonio Fatas, Erik Feyen, Jakob de Haan, Graham Hacche, Elena Ianchovichina, Ergys Islamaj, Gerard Kambou, Jean Pierre Lacombe, Valerie Mercer Blackman, Joseph Mawejje, Ugo Panizza, Zia Qureshi, David Robinson, Naotaka Sugawara, Jonathan Temple, Christopher Towe, and Garima Vasishtha, as well as participants of many seminars for excellent feedback and comments. The findings, interpretations and conclusions expressed in this paper are entirely those of the authors and should not be attributed to the institutions they are affiliated with. 1 I. Introduction The global economy headed into the COVID-19 pandemic and the Russian invasion of Ukraine after a decade of slowing growth. The pandemic-induced global recession of 2020 further deepened this slowdown and Russia’s invasion of Ukraine in February 2022 has already left additional scars. These adverse shocks have reduced not just actual global output growth but have also dampened potential growth—the rate of increase of potential output, defined as the level of output an economy can sustain at full capacity utilization and full employment. Potential growth is a critical determinant of a wide range of macroeconomic and development outcomes. Sound fiscal and monetary policy decision about stimulus or austerity cannot be taken without being grounded in a firm understanding of potential growth. Potential growth is of fundamental importance to short- and long-run macroeconomic analysis and policy but it is not directly observable. In an extensive literature, three main methods of estimating potential output growth have been employed, each of which has its advantages and disadvantages. Thus, measures of potential growth based on production function estimates make it possible to study the contributions of the fundamental drivers of growth—namely, the growth of the factors of production and technical progress—but involve assumptions that may be viewed as restrictive. A second method uses economic analysts’ long-term (five-year-ahead) output growth forecasts, which may be assumed to incorporate their judgments. The third method obtains measures of potential growth from statistical filters of actual growth data; it may be best at ensuring consistency between estimates of potential growth and output gaps, on the one hand, and indicators of domestic demand pressures, on the other. This study introduces the most comprehensive international database yet for the nine most commonly used measures of potential growth, based on these three methods, for the largest available sample of countries over the period 1981-2021. In addition, this study addresses the following questions. First, how has potential growth evolved in recent decades? Second, how have recessions and other adverse developments affected potential growth? Finally, through which channels have such developments affected potential growth? The study makes several contributions to the literature. First, it introduces the first comprehensive database of the nine most commonly used measures of potential growth for the largest available country sample—of up to 173 economies (37 advanced economies and 136 emerging market and developing economies [EMDEs])—over 1981-2021. These measures comprise one based on the production function approach; five based on the application of univariate filters (Hodrick-Prescott, Baxter-King, Christiano-Fitzgerald, Butterworth, and Unobserved Components filters); one based on a multivariate Kalman filter; and two based on long-term growth forecasts. Previous studies have limited themselves to a single method of measuring potential growth, such as the production function approach (OECD 2014), or multivariate filters (ADB 2016; IMF 2015). This study builds on earlier work published before the pandemic that utilized several measures of potential growth (Kilic Celik, Kose, and Ohnsorge 2020; World Bank 2018). Second, the study documents a broad-based and long-standing slowdown in potential 2 growth. All measures of potential growth show a decline in global potential growth in the decade before the pandemic and that it was internationally widespread. Earlier studies documented the decline for only a subset of measures (for example, Chalaux and Guillemette 2019; Kilic Celik, Kose, and Ohnsorge 2020). Third, this study is the first to systematically compare the long-term damage to potential growth of short-term economic disruptions—such as recessions, banking crises, and epidemics—in a large set of countries. Thus far, only a few studies have estimated the effects of recessions on potential output growth, and they were confined to an OECD sample and the production function approach (Furceri and Mourougane 2012; Mourougane 2017). This study broadens the earlier research by estimating the effects of recessions, banking crises, and epidemics in a large sample of advanced economies and EMDEs and for a wide range of potential growth measures. Fourth, this study estimates empirically, using a set of local projection models, the multiple channels through which short-term economic disruptions have dampened potential growth. Specifically, it estimates the effects of disruptions on the growth of the labor supply, the growth of investment, and the growth of total factor productivity (TFP) in a consistent framework. Previous studies have typically examined overall effects on growth or effects through individual channels. The theoretical literature has analyzed, typically using DSGE models, several mechanisms through which short-term output disruptions (associated with recessions and other adverse events) may have longer-term effects. Weak aggregate demand during such disruptions may reduce the expected profitability of, and thus discourage, productivity- increasing research and development (Fatás 2000). It may similarly discourage investment in productivity-raising new technologies that would otherwise have improved productivity (Anzoategui et al. 2019). Investors who expect weak aggregate demand to persist will be reluctant to invest; reduced investment will tend to lower asset prices which, through wealth effects, will further depress consumption (Caballero and Simsek 2017). If aggregate demand weakness is accompanied by a financial crisis, financial market frictions can restrict firms’ access to credit and start-up capital, further reducing investment and productivity growth. 2 Damage to potential output from short-term disruptions can also result from productivity losses due to resource misallocation (Dieppe, Kilic Celik, and Okou 2021; Furceri et al. 2021); these may be partially offset by productivity gains stemming from the exit of low- productivity firms (Bloom et al. 2020). Finally, high unemployment that accompanies weak aggregate demand tends to lead to human capital losses and reduced job search activity among the long-term unemployed (Blanchard and Summers 1987; Lockwood 1991). Empirical estimates have documented that some of these mechanisms were indeed at work during past recessions. An analysis of data for a large sample of countries during 1960- 2018 found that financial crises, especially when accompanied by a rapid buildup of debt, 2 For details of these empirical findings involving financial markets, see Claessens and Kose (2017), Queralto (2013), and Wilms, Swank, and de Haan (2018). 3 were associated with persistent productivity losses (Dieppe, Kilic Celik, and Okou 2021). Among a large sample of firms in six EMDEs in Europe, firms in sectors that faced the largest adverse demand shocks during the 2009 global recession reduced capacity most (Nguyen and Qian 2014). In a sample of 61 countries during 1954-2010, banking crises were followed by lower labor productivity growth, consistent with a loss of human capital during these crises (Oulton and Sebastia-Barriel 2016). Other studies found that the return of actual output growth or levels to pre-recession trends was non-linear and depended on the persistence, depth, and source of the recession and on whether it was accompanied by financial crises. 3 None of these studies, however, systematically examines the various channels through which short-term disruptions reduce potential growth. The study reports the following key findings. Trend decline in potential growth. An internationally widespread decline in average annual potential growth occurred between 2000-10 and 2011-21. This is shown by all estimates of potential growth, globally and for the main country groups—advanced economies and EMDEs. Global potential growth, as estimated using the production function approach, fell to 2.6 percent a year during 2011-21 from 3.5 percent a year during 2000-10; advanced-economy potential growth fell to 1.4 percent a year during 2011-21, 0.8 percentage point below its 2000-10 average; and EMDE potential growth fell to 5.0 percent a year during 2011-21 from 6.0 percent a year during 2000-10. The weakening of potential growth was highly synchronized across countries: during 2011-21, potential growth was below its 2000-10 average in 96 percent of advanced economies and 57 percent of EMDEs. This widespread decline reflected a multitude of factors. All the fundamental drivers of growth faded in 2011-21: TFP growth slowed, investment weakened, and labor force growth declined. Persistent impact of recessions on potential growth. Recessions, even five years later, were associated, on average, with a decline of about 1.4 percentage points in potential growth. While the magnitude of the estimated decline in potential growth five years after a recession depended on the measure (with a range of 0.2-1.4 percentage points), it was always statistically significantly negative. The effect was somewhat stronger in EMDEs than in advanced economies: in EMDEs, potential growth was still, on average, 1.6 percentage points lower five years after the recession, whereas, in advanced economies, it was 1.3 percentage points lower. Larger impact of recessions than other adverse events on potential growth. The longer- term effect of recessions on potential growth tended to be somewhat more severe than the effects of other adverse events. Banking crises were associated with initially larger falls in potential growth (peaking at 1.8 percentage point after two years) as a result of a collapse in investment. However, this tended to unwind quickly such that the fall in potential growth after five years was only 1.2 percentage point. Epidemics were associated with more modest, but still statistically significant, short- and medium-term declines in potential growth. These were more severe in EMDEs than in advanced economies, which may have been better able to limit the economic damage with fiscal and monetary 3 For a discussion of the impact of financial crises on growth, see Ball (2014); Claessens, Kose, and Terrones (2009, 2012); Furceri and Mourougane (2012); and Haltmeier (2012). 4 stimulus. Adverse effects through multiple channels. Recessions affected potential growth through multiple channels. Four to five years after an average recession, the annual growth of investment, employment, and productivity remained significantly lower than in “normal” years (by 3 percentage points, 0.7 percentage point and 0.7 percentage point, respectively). This contrasts with banking crises, which tended to be associated mostly with lasting losses of productivity growth, and epidemics, which were mainly associated with lasting employment losses, possibly reflecting economic shifts caused by behavioral responses to epidemics. Different features of potential growth estimates. The comprehensive database also allows a comparisons across potential growth measures. Forecast-based estimates tend to be systematically higher than other estimates, and estimates based on univariate filtering techniques systematically lower. Estimates based on filtering techniques tend to be the most volatile and to track actual growth most closely, as expected. Estimates based on the production function approach tend to be the most stable and the least correlated with actual growth as they capture slow-moving drivers of potential growth. The study proceeds as follows. Section II presents the database. Section III describes movements in potential growth around the world in recent decades. Section IV estimates the effects on potential growth of recessions. Section V documents the channels through which these operates. Section VI concludes. II. Database Three main methods of estimating potential growth estimates have been used in the literature, and several different measures can be derived using variants of them. The comprehensive database developed here allows a comparison of the behaviors of such measures. The database includes nine measures of potential growth for up to 173 countries over periods as long as 1981-2021. The baseline measure of annual potential growth, estimated using the production function approach, is available for up to 30 advanced economies and 64 EMDEs for 1998-2021 (table 1, annex A). Six univariate and multivariate filter-based estimates of potential growth, which require quarterly data, are available for up to 37 advanced economies and 52 EMDEs for 1980Q1-2022Q1 (annexes B and C). IMF World Economic Outlook-based estimates of potential growth are available for up to 37 advanced economies and 136 EMDEs for 1990-2022 (annex D). Consensus forecast-based estimates of potential growth are available for up to 34 advanced economies and 44 EMDEs for 1990-2022. The database also includes projections for a subset of measures. For the production function approach, projections are available for 2022-32. These projections and the methodology on which they are based are presented and analyzed in Kilic Celik, Kose, and Ohnsorge (2023). For the filter-based estimates, forecasts are available up to 2024Q4. This study discusses aggregates for the global economy and for particular country groups. 5 These aggregates are real GDP-weighted averages (at 2010-19 prices and market exchange rates) for a balanced sample of 30 advanced economies and 53 EMDEs for 2000-21, unless specified otherwise. The 53 EMDEs comprise 6 economies in East Asia and the Pacific (EAP), 9 economies in Europe and Central Asia (ECA), 16 economies in Latin America and the Caribbean (LAC), 5 economies in the Middle East and North Africa (MNA), 3 economies in South Asia (SAR) and 14 economies in Sub-Saharan Africa (SSA). Data for about half of EMDEs (mainly in ECA and SSA) are not available before 1998. Hence, to ensure broad country coverage, the sample period is restricted to 2000-21 when discussing international averages. However, when discussing the robustness of trends among different measures, the sample is restricted to those countries for which data are available for all measures. II.1 Basic concepts Three main methods of estimating potential growth have been employed in the literature, sometimes with different objectives. Some have been used to analyze short-term movements in potential growth, while others have focused on long-term developments (Basu and Fernald 2009). Estimates of movements in potential growth in the short term may be computed using time-series filtering techniques, including univariate or multivariate filters, while estimates of potential output growth over longer periods are usually based on structural models that include a production function or on long-term growth forecasts. In the short term, when factors of production cannot be reallocated in response to shocks, potential growth may be viewed as the growth of output that can be sustained without putting pressure on given productive capacity and inflation (Okun 1962). Potential output growth can be buffeted in the short term by temporary disruptions and boosts to supply that may dissipate over the longer term. For example, a shift in the composition of demand may render part of the existing capital stock obsolete, effectively reducing potential output and its growth in the short-term. However, over the longer term, firms would be expected to adjust to the new structure of demand, returning potential output growth toward its previous path. The short-term measure is particularly relevant for demand management and monetary policy, since temporary supply constraints or upward demand shocks tend to reduce the effective slack in the economy, with implications for macroeconomic policy and the monetary policy rate. Central banks, in particular, need to focus on movements in potential growth in the short term as they gauge deviations of actual from potential output levels over the horizon of monetary policy transmission, around one to two years. In the production function framework, potential output growth is a function of growth in the factors of production—the capital stock and the labor force, along with current technological progress (Solow 1962). Potential output growth in the long term thus depends on these fundamental drivers, an implicit assumption being that the factors of production are allocated to their most productive uses, regardless of temporary supply shocks. Finance and economy ministries often focus on potential growth over longer periods, aware that boosting it will promote fiscal sustainability over longer time horizons. 6 II.2 Measures of potential output growth The literature has largely focused on three methods of estimating potential growth: a production function method, time-series filters, and analysts’ growth forecasts. Production function method. The production function approach represents potential output as a function of the fully utilized capital stock, fully employed labor force, and technology as measured by TFP. For analytical convenience, the production function is often assumed to have a particular form, known as Cobb-Douglas. Potential TFP growth is estimated as the predicted value of a parsimonious panel regression of five-year averages of trend TFP growth on lagged per capita income relative to the advanced-economy average (to proxy convergence-related productivity catchup), education and demographic indicators, and trend investment (annex A). Potential labor supply is estimated as the population-weighted aggregate of predicted values of age- and gender-specific labor force participation rates from regressions on policy outcomes and cohort characteristics, business cycles, and country effects. The potential capital stock is assumed to match the actual capital stock. Time-series filtering methods. These methods employ univariate or multivariate filters. Univariate filters involve estimates of trend output using only GDP data series (annex B). Multivariate filters use the empirical relationship between GDP and other variables (such as inflation, unemployment rates, commodity prices or financial variables) to help distinguish short-run deviations of output from trends (annex C). The database in this study employs the following five univariate filters: the Hodrick-Prescott filter, the Baxter- King filter, the Christiano-Fitzgerald filter, the Butterworth filter, and a filter based on an unobserved components model. An additional multivariate filter uses financial variables and commodity prices, a Phillips curve relationship, a Taylor rule, and Okun’s law. Growth forecasts. This method is applied using two sets of long-term (five-years-ahead) growth forecasts, from Consensus Economics and the IMF’s World Economic Outlook database (annex D). These forecasts are based partly on models used by the analysts and partly on the analysts’ judgement. Judgment can play an important role during periods of major structural change, which models may not be well-equipped to capture. Each approach comes with advantages and disadvantages (table 2). Even in data-poor environments, univariate filters are straightforward to implement. Multivariate filters utilize additional information that can ensure that the measure of potential output is better aligned with its determinants, as suggested by economic theory. In particular, the multivariate filter-based estimates can ensure that estimated output gaps in the short term are consistent with indicators of domestic demand pressures (such as inflation, unemployment, current account balances, and capacity utilization). All statistical filters, however, have drawbacks: in particular, they suffer from well-known “end-point” problems that tend to lead to large revisions as new data become available. The approach employed here includes forecasts of real GDP growth to minimize this problem. Since they capture high-frequency movements, measures of potential growth based on filtering techniques correlate strongly with actual output growth and with each other. 7 The production function approach has the advantage of taking into account the fundamental drivers of output on the supply side—factor inputs and technology—that dominate in the long run. While estimates of potential growth based on this approach are often consistent with long-term growth averages, they correlate less closely with actual growth in the short term. Potential growth measured by the production function approach is also only weakly correlated with potential growth estimates obtained from filtering techniques. The production function approach has a number of drawbacks, however. It assumes a particular functional form of the relationship between factor inputs, technology, and output. Its application relies on imperfect measures of, or proxies for, the growth of potential TFP, labor supply, and the capital stock. And it is unable to capture cyclical shocks to capacity and supply that may cause short-term fluctuations in potential output. Finally, the approach provides measures of potential output growth, but derivation of potential output levels would require additional steps to identify an “anchor level” in which the output gap is closed. Long-term growth forecasts generally incorporate analysts’ judgment and, thus, capture factors that cannot be econometrically modelled. As a result, similar to estimates based on the production function approach, these forecasts are only weakly correlated with filter- based estimates of potential growth. However, in practice, forecasts can be sticky and, at times, difficult to interpret. II.3 Comparison of different potential growth measures The estimated potential growth rates resulting from the application of these methods differ in their levels and evolutions over time. This section briefly explores these differences. First, differences among potential growth estimates were wider for advanced economies than EMDEs (figures 1.A and B). During 2000-21, potential growth estimated from forecasts was the highest among the nine measures in more than half the country-year pairs (figure 1.C). The lowest estimates were generally produced by the univariate filters. At the country level, the same pattern was found: forecast-based measures of potential growth tended to be the highest and measures from univariate filters the lowest, especially over the past decade. Second, multivariate filter-based estimates of potential growth had narrower confidence bands than those based on univariate filters (figure 1.D). This likely reflects the use of additional demand pressure indicators in the multivariate filter that help identify the output gap more accurately. Confidence intervals cannot be computed for estimates based on the production function approach or analysts’ forecasts. Third, global, advanced-economy, and EMDE potential growth estimates based on univariate and multivariate filters typically have the highest variances, while those based on the production function approach have the lowest (figure 1.E). At the country level, univariate filter estimates have the largest variance (in about 75 percent of cases). Fourth, univariate filter-based estimates have the least persistence, especially in advanced 8 economies, while estimates from forecasts and the production function approach have the most persistence across all groups of countries (figure 1.F). 4 These findings are intuitively appealing, as filter-based estimates are designed to capture time-series variation, whereas the others rely on more persistent drivers of potential growth. Fifth, estimates from different multivariate and univariate filters tend to be highly correlated, with a median within-country correlation coefficient above 85 percent (figure 2.A). However, they correlate only moderately with estimates from the production function approach and analysts’ forecasts. Similarly, production function-based and forecast-based estimates correlate only moderately with each other, whereas estimates from the two sources of growth forecasts are highly correlated with each other. Finally, as expected, estimates of potential growth based on filters derived from the unobserved components model most closely track actual growth, with an average correlation coefficient of 0.95 across the country sample, followed by estimates based on the multivariate filter and other univariate filters (figure 2.B). As expected, given its construction from slow-moving variables, the production function approach deviates more from actual growth (with a correlation of 0.45 with actual growth). The correlation is even lower for forecast-based measures of potential growth, which tend to change only when forecasters modify their views about long-term growth drivers. III. Evolution of potential growth This section first reviews the evolution of potential growth over the past two decades. It then focuses on potential growth during the last two global recessions, of 2009 and 2020. While both sub-sections rely mostly on the production function-based measures of potential growth, the findings are consistent with those from the other measures of potential growth. III.1 Potential growth over time Global potential growth, as estimated using the production function approach, fell to 2.6 percent a year over 2011-21 from 3.5 percent a year during 2000-10 (figure 3.A). 5 The weakening of potential growth was internationally widespread. Thus during 2011-21, potential growth was below its 2000-10 average in 96 percent of advanced economies and 57 percent of EMDEs. Economies with potential growth below its 2000-10 average accounted for about 80 percent of global GDP in 2022 (figure 3.B). Per capita potential growth estimates also show a trend decline over time, to 2.0 percent a year in 2011-21 from 2.7 percent a year during 2000-10 (figure 3.C). These estimates suggest a trend slowdown in global potential growth around the cyclical shocks that depressed actual 4 The coefficient on lagged potential growth from a regression with one autoregressive term is taken to capture the degree of persistence here. 5 Data for half the EMDEs (mainly in ECA and SSA) are not available before 1998. Hence, to ensure broad country coverage, the sample period is restricted to 2000-2021 for discussing country groups. However, when robustness of trends among different measures is discussed, the sample is restricted to those countries for which data are available for all measures. 9 growth below its elevated average in the early 2000s. The finding of a decline in potential growth is robust with respect to the measure used, although the magnitude of the slowdown differs across the measures. To ensure comparability, a smaller sample of 30 advanced economies and 25 EMDEs is used for which all nine measures are available. By all these measures, global potential growth slowed by 0.9-1 percentage point a year from its average in 2000-10, to 2.5-2.9 percent a year in 2011-21 (figure 3.D). In advanced economies, the potential growth slowdown set in before the global financial crisis. After a sharp decline during 2008-10—the period of the global financial crisis and the start of the euro area sovereign debt crisis—potential growth stabilized in 2011-21 as investment growth recovered. However, at 1.4 percent a year over 2011-21, potential growth in advanced economies was 0.8 percentage point below its 2000-10 average (figure 4.A). As in the broader set of advanced economies, potential growth in the G7 economies (Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States) was 1.5 percent a year on average in 2011-21, 0.5 percentage points below its 2000-10 average. EMDEs, by contrast, enjoyed a short-lived pre-global recession surge in potential growth in the 2000s that subsequently faded. In the wake of the global financial crisis and associated global recession, a surge in public investment underpinned EMDE potential growth, offsetting softening growth of both TFP and labor supply. As EMDE policy stimulus was unwound and as investment growth plummeted in commodity-exporting EMDEs amid the oil price slide in 2014-2016, EMDE potential growth slowed sharply in 2015-19. A sharp investment growth slowdown during the 2010-19 also depressed potential growth in China whereas the slowdown was milder in other EMDEs where investment growth remained more robust and demographics were more favorable. Overall, at 5.0 percent a year, EMDE potential growth during 2011-21 fell 1.0 percentage point a year short of its average during 2000-10 (figure 4.B). Across EMDE regions, potential growth fell furthest in those regions that had benefited from rapid per capita income convergence in the early 2000s or included many commodity- exporting EMDEs (figures 4.C.and D). The slowdown in potential growth in 2011-21 relative to its 2000-10 average was sharpest in MNA, where investment growth plunged amid the oil price drop of 2014-16 and conflict and policy uncertainty persisted in parts of the region. In EAP, potential growth in 2011-21 was 1.4 percentage points a year lower than in 2000- 10. This decline mostly reflected a slowdown in potential growth in China, partly as a result of policy efforts aimed at rebalancing growth away from investment towards more sustainable growth engines; adding to this was slower growth of both TFP and the working-age population. In ECA and LCA, potential growth in 2011-21 was 0.5-0.6 percentage point a year lower than in 2000-10. The ECA region’s previous two decades of rapid integration into European Union production networks, beginning in the 1990s, gradually diminished its potential for further catchup productivity growth. The region also hosts several energy- 10 exporting countries (including Russia) which suffered recessions or slowdowns in the wake of the 2014-16 slump in oil prices. In LAC, potential growth suffered from weakened productivity growth, partly as a result of adverse terms-of-trade shocks and bouts of policy uncertainty, as well as less favorable demographics. Potential growth in SSA also declined somewhat (by 0.2 percentage points a year in 2011- 21 relative to 2000-10). A sharp slowdown in TFP growth was only partially offset by favorable demographics and rapid capital accumulation, which accelerated as resource discoveries were developed into operating mines and oil fields and governments undertook large-scale public infrastructure investments. In 2011-21, potential growth in SAR remained broadly unchanged from 2000-10. Growth of the labor force benefited from a demographic dividend. The share of the population of working age rose by more than one-tenth between 2000 and 2021, reaching 67 percent in 2021. Capital and TFP also maintained their growth momentum in 2011-21. Growth in investment remained broadly robust over this period—growing faster than in the EMDE average—and the investment-GDP ratio rose by 5 percentage points of GDP between 2000 and 2021, to more than 28 percent of GDP in 2021. III.2 Potential growth during global recessions The 2000-21 period spans two global recessions—the 2009 recession that was triggered by the global financial crisis and the 2020 recession that was caused by the COVID-19 pandemic. These recessions disrupted fixed capital investment and caused widespread employment and output losses. In the case of the 2020 recession, disruptions of education systems caused by pandemic-induced reductions in social interaction also slowed down human capital accumulation. By the production function-based measure of potential growth, global potential growth slowed by 1.2 and 1.3 percentage point from two years before the global recessions of 2009 and 2020, respectively, to the recession year itself (figure 5.A). The slowdowns in potential growth in EMDEs differed more between the two recessions (1.3 percentage points in 2007-09 and 1.7 percentage points in 2018-20) than the slowdowns in advanced economies (1.2 percentage points in 2007-09 and 1.1 percentage points in 2018-20; figures 5.B and C). The considerably smaller slowdown in EMDEs in the 2009 global recession largely reflected the investment-driven support for potential growth in China during the global financial crisis. In EMDEs excluding China, potential growth declined by 1.2 and 2.0 percentage points in the 2009 and 2020 recessions, respectively (figure 5.D). In advanced economies, the slowdown in potential growth in the two global recessions reflected steep declines in investment and TFP growth, whereas in EMDEs it reflected mostly a decline in TFP growth (figures 6.A-D). In both country groups, slowing labor force growth also contributed. The steeper slowdown in potential growth in EMDEs in 2020 than in 2009 reflected the deeper collapse in investment but also the pandemic- induced fall in potential labor force participation. Although both global recessions resulted in a slowdown in potential growth, they differed in the behavior of potential growth in the subsequent recoveries. The global financial crisis 11 was followed by a decade of investment weakness and reduced productivity growth, leading to a failure of potential growth to return to pre-recession rates. In contrast, the 2020 global recession was followed by the swiftest first-year output rebound of any global recession over the past eight decades (World Bank 2021). This was accompanied by strong growth in investment, especially in advanced economies, and a productivity rebound, which together lifted potential growth to pre-recession rates globally, in advanced economies, and in EMDEs. However, the impact of this initial rebound in potential growth is likely to be temporary because of the persistent headwinds faced by the fundamental drivers of potential growth (Kilic Celik, Kose, and Ohnsorge 2023). These estimated movements in potential growth around global recessions were similar for almost all measures of potential growth, except those based on forecasts. Potential growth declined in the two recession years globally, in advanced economies, in EMDEs, and in EMDEs excluding China. 6 On average across the eight measures that showed declines in the two recessions, global potential growth slowed by about 1.3 percentage points from two years before the recession to the year of the recession. 7 The slowdown was larger in EMDEs (1.5 percentage points) than in advanced economies (1.2 percentage points). The recession year in both episodes generally saw the trough in potential growth for all measures. The estimated decline in potential growth was smallest for production function- based measures and largest for measures obtained using univariate filters. IV. The long-term effects of short-term shocks on potential growth The COVID-19-induced output collapse of 2020 renewed concerns about the impact of recessions on the level and growth of potential output. A number of studies have documented the lasting effects of country-specific recessions and financial crises on the level or growth of actual or potential output (Cerra and Saxena 2008; Furceri and Mourougane 2012; Mourougane 2017). However, these studies have mostly focused on OECD countries using only production function-based estimates of potential growth. This section broadens the scope of the earlier literature in three dimensions. First, it examines the effect of country-specific recessions on potential growth in a much larger sample of countries, including both advanced economies and EMDEs. Second, it employs all the measures of potential growth described above to obtain a better understanding of the linkages between recessions and potential growth. Third, in addition to recessions, it considers other adverse events, such as banking crises and epidemics, and compares their effects on potential growth. IV.1 Methodology A (country-specific) recession is defined as a period from a peak in output preceding a 6 For the COVID-19-induced global recession of 2020, this is broadly consistent with the findings of persistently lower potential output levels by Bodnár et al. (2020) for the euro area and Fernald and Li (2021) for the United States. 7 Measures based on consensus forecasts for long-term growth are not covered here because they have a much smaller country sample. 12 business cycle trough to the trough, with a trough defined as a year in which output growth is both negative and at least one standard deviation below its long-term (1995- 2020) average (as in Huidrom, Kose, and Ohnsorge 2016). This definition yields up to 124 recessions in 37 advanced economies and up to 351 recessions in 101 EMDEs during 1980- 2020. Almost half of such recessions at the country level occurred during global recession years (1975, 1982, 1991, 2009, 2020; figure 7.A). Recessions at the country level, on average, lasted 1.5 years and were associated with a contraction in actual output of 4.0 percent, on average (figure 7.B). In advanced economies, recessions were, on average, somewhat less severe than in EMDEs (with drops of 3.5 percent and 4.3 percent, respectively; figures 7.C and D). The duration of recessions was similar, at 1.5 years, in the two country groups. A local projection method (LPM) is employed to estimate the evolution of potential growth following recessions (annex E). The model estimates the cumulative effect of recessions on potential growth, following Jordà (2005) and Teulings and Zubanov (2014). In impulse responses, the model estimates the effect of short-term shocks (the recession, banking crisis, or epidemic event) over a horizon h on potential growth while controlling for other determinants: yi,t+h – yi,t = αh + βhshocki,t + γh ∆yi,t-1 + fixed effectsi + εi,t , where yi,t is potential growth. The model controls for country-fixed effects to capture time- invariant cross-country differences. The variable shocki,t is a dummy variable for a recession event (or banking crisis or epidemic), the main variable of interest. Lagged potential growth yi,t-1 controls for the history of potential growth. IV.2 Results Long-term effect of recessions. Even five years after recessions, potential growth as measured by the production function approach is estimated, on average, to have been 1.4 percentage points lower than if a recession had not occurred (figure 8.A). Coefficient estimates for the recession dummy are statistically significantly negative for the first five years after a recession. The effect was somewhat stronger and more persistent for EMDEs, with 1.6 percentage points lower potential growth five years after a recession compared to 1.3 percentage points for advanced economies (figures 8.B.C). These results are broadly robust to the choice of potential growth measure and the definition of recessions. Four to five years after recessions, potential growth as measured by most methods other than the production function approach is estimated to have been 0.2-1.3 percentage points lower than if a recession had not occurred (annex E). 8 Recessions could alternatively be defined as years of negative output growth, regardless 8 The only exceptions are, for advanced economies, forecast-based estimates from the IMF World Economic Outlook database and, for EMDEs, multivariate filters and Hodrick-Prescott-filtered estimates. One possible reason for the unresponsiveness of some forecast-based measures might be that forecasters’ perception of long-term growth is stickier for advanced economies than for EMDEs. 13 of the depth of the output decline. This alternative definition of events would yield 541 recessions events (151 events in 37 advanced economies and 390 events in 101 EMDEs), around 14 percent more than the baseline sample of 475 events. 9 Potential growth slowed statistically significantly following recessions defined in this way also. Long-term effect of other adverse events. The effects of banking crises and epidemics on potential growth are also examined and compared with those of recessions (annex E). The banking crises examined are those identified in Laeven and Valencia (2020). This yields a sample of 25 banking crises in 32 advanced economies and 41 banking crises in 91 EMDEs during the period 1990-2021. During the year of an average banking crisis globally, actual output rose by 0.7 percent—well below the average annual global output growth during the sample period of 1990-2021 (3.5 percent) and even further below average annual EMDEs output growth over this period (4.1 percent). The average crisis lasted less than 1 year. The five recent epidemics examined are: SARS (2002-03), swine flu (2009), MERS (2012), Ebola (2014-15), and Zika (2015-16). They affected 96 countries—32 advanced economies and 64 EMDEs. On average, they were accompanied by close-to-zero output growth, compared to the average growth of 4.0 percent in these countries during the sample period outside these episodes. Like recessions, both banking crises and epidemics have reduced potential growth, but the time profiles of their effects differed from those of recessions. Banking crises tended to have stronger short-term impacts than recessions but somewhat smaller long-term effects on potential growth. 10 Overall, 81 percent of banking crises were associated with recessions within three years (figure 8.D). Using estimates based on the production function approach, potential growth slowed more steeply in the first 1-2 years after banking crises than after recessions, but the initial decline in potential growth after banking crises was subsequently partly reversed, whereas the slowing effect of recessions strengthened over time (figures 8.A and 9.A). The long-term effects of banking crises on other potential growth measures are estimated to have been even weaker than the effect on measures based on the production function approach (annex E). 11 The effect of banking crises was stronger but shorter-lived in EMDEs than in advanced economies; five years after a banking crisis, the effect was no longer statistically significant in EMDEs but still significant in advanced economies (figures 9.B and C). The fading effect of banking crises on potential growth may in part reflect the lack of a lasting impact on the growth of employment and investment, especially in EMDEs, as the disruptions of banking crises 9 By this alternative definition, the average recession is associated with an actual output contraction of 3.7 percent and lasts 1.6 years. 10 Results for currency crises and debt crises suggest limited and short-lived impacts that are statistically significant only in the year of the event (currency crises) or up to two years after the event (debt crises). 11 The exercise is repeated for banking crises that were followed by recessions within a three-year window. There were 20 such cases events in the sample used here. The results indicate statistically significant impacts of recessions combined with banking crises, with somewhat larger short-term effects but similar long-term effects to banking crises, but the difference between the responses of potential growth to banking crises with and without recessions is not statistically significant. 14 were often followed by economic rebounds. The strong initial impact of banking crises on potential growth, as well as their declining and highly heterogeneous longer-term effects, are in line with estimates of actual output losses reported in the literature. Candelon, Carare, and Miao (2016) document significant growth slowdowns in the first year following banking crises which become more muted in subsequent years. Similarly, Dwyer, Devereux, and Baie (2013) document wide heterogeneity in growth impacts five years after banking crises. 12 In a comprehensive review of the literature, Claessens and Kose (2018) also find that the duration of a recession depends on the features of the financial stress that accompanies it. In particular, house price busts, especially when combined with credit crunches, can prolong recessions, whereas a rapid recovery in housing and asset markets can accelerate the broader economic recovery from financial stress. Epidemics, too, had somewhat more modest, but still statistically significant, negative long-term effects on potential growth than did recessions—larger in EMDEs than in advanced economies (figures 2.8.A and 2.9.D). Based on the production function measure, potential growth five years after an epidemics was 0.9 percentage point lower than it would otherwise have been (compared with declines of 1.2 and 1.4 percentage points after banking crises and recessions, respectively). One reason for the more muted effect of epidemics than of recessions is their more muted effect on productivity over the medium term. Experience since 2020, when the COVID-19 pandemic erupted, has shown how rapidly productivity can rebound when pandemic restrictions are lifted and disruptions are resolved. V. How do short-term shocks affect potential growth? The previous section established that recessions have been associated with significantly slower potential growth for several subsequent years. This section assesses three possible channels through which this process unfolded: employment, investment, and TFP growth. The literature provides ample evidence that all three channels suggested by the production function approach are likely to have been important in weakening potential growth following recessions and other adverse events. V.1 Effects of recessions Employment and labor supply. In a recession, unemployment generally rises significantly and remains elevated for a prolonged period. For example, in the sample of recessions examined here, unemployment remained 1.8 percentage points higher, on average, three years after the recession than would have been the case otherwise (annex E). Such a lasting effect is in line with other findings in the literature. In the United States, for example, a 1 percentage point increase in state-level unemployment during the 2007-09 12 Even if the effect of banking crises on output growth has been short-lived, their effect on output levels has been persistent. Cerra and Saxena (2008) showed this for actual output levels five to ten years after financial crises; Ollivaud and Turner (2014) showed this for potential output levels three to seven years after the global financial crisis. 15 recession was associated with 0.3 percentage point lower employment rates in 2015 (Yagan 2019). Following recessions, lingering uncertainty about future sales prospects may discourage firms from hiring (Baker, Bloom, and Davis 2016; Bloom 2009, 2014). Financial constraints may force the more indebted firms into greater job cuts in the event of demand drops (Giroud and Mueller 2017). Long spells of unemployment may discourage workers and erode the skills of the long-term unemployed (Ball 2009; Blanchard 1991; Blanchard and Summers 1987). Thus, the decrease in employment over a prolonged period after a recession tends to have adverse consequences for labor supply and potential output. Investment and capital accumulation. Gross fixed investment typically falls more sharply in response to economic downturns than other components of GDP (Kydland and Prescott 1982). A recession can cause investors to reassess long-term growth prospects. A downgrade in growth forecasts could erode prospects of long-term returns on investment or risks around expected returns and, thus, discourage investment. Access to finance for investment may also become more restricted and discourage investment, especially for younger, more innovative, and riskier firms (Fort et al. 2013). 13 Reduced capital accumulation in a recession will directly reduce potential growth. Total factor productivity. A collapse in investment growth not only directly reduces potential growth but also indirectly by slowing the adoption of productivity-enhancing embodied technologies and the reallocation of resources towards more productive uses (Dieppe, Kilic Celik, and Okou 2021; Syverson 2011). Workers losing their jobs during recessions may enter permanently lower-skilled career paths (Huckfeldt 2022). Skills mismatches between job market entrants and job requirements are larger during recessions than expansions and tend to be long-lasting, suggesting persistent productivity losses from such mismatches (Liu, Salvanes, and Sørensen 2016). Recessions are also likely to be associated with reduced spending on research and development, with negative consequences for the growth of TFP. All three channels were at work during the recessions considered in this study (annex E). Five years after the average recession, TFP growth is estimated to have been 0.7 percentage point lower than it would have been without a recession and, in EMDEs, 0.9 percentage point lower (figures 10.A and 11.A). Investment growth declined steeply in the first year of the average recession and remained significantly lower five years later— 3 percentage points below what it would have been without a recession, both globally and in EMDEs (figures 10.B and 11.B). The effect was somewhat shorter-lived for employment. Four years after the average recession, employment growth was about 0.7 percentage point lower than what it would have been otherwise. However, for EMDEs, this effect was no longer statistically significant by the fifth year (figures 10.C and 11.C). The absence of a longer-lasting employment response in EMDEs is, in part likely to reflect the large, flexible informal economies that help these countries absorb shocks to labor markets. 13 Similar lasting impacts of investment weakness have been shown for banking crises (Wilms, Swank, and de Haan 2018). 16 V.2 Effects of banking crises and epidemics The effects of banking crises on the growth of TFP, investment, and employment tended to be short-lived (figures 10.D-F and 11.A-F). Five years after the average banking crisis, neither investment growth nor employment growth were statistically significantly lower than otherwise; only TFP growth was still significantly lower. Epidemics were associated, even five years later, with statistically significantly lower TFP growth, investment growth, and—in contrast to recessions and banking crises—potential labor supply growth. The effect of epidemics on investment growth after five years was somewhat stronger, and the effect on TFP growth weaker, than the effects of recessions (figures 10.D-F). Banking crises had larger long-term adverse effects on TFP growth, investment growth, and employment growth in advanced economies than EMDEs, possibly reflecting the larger role of finance in, and greater financial development of, advanced economies. Conversely, epidemics had larger long-term adverse effects on these variables in EMDEs than in advanced economies, in part perhaps because EMDE governments and central banks had less policy room to dampen the economic effects of epidemic disruptions (figures 11.A-F). VI. Conclusions Potential growth, the growth an economy can generate at full employment and full capacity, is critical for a sustained increase in living standards. It also anchors the calibration of macroeconomic policies. This study introduced the most comprehensive international database of potential growth, including the nine most widely used measures of potential growth for up to 173 countries over 1981-2021. At the global level, all measures point to a steady decline in potential growth in the past decade. This decline was internationally widespread, with potential growth in 2011-21 falling below its 2000- 10 average in 70 percent of countries. The decline in potential growth between 2000-10 and 2011-21 was almost as large in advanced economies (0.8 percentage point per year) as in EMDEs (1.0 percentage point per year). The study also presented an application of the new database by studying the effects of recessions and other adverse events on potential growth. Recessions, on average, have been followed, even five years later, by a drop of 1.4 percentage points in potential growth. The magnitude of this estimated decline varies somewhat among the possible measures of potential growth, but it is virtually always statistically significant. This lasting effect of recessions operates through the many channels: Four to five years after recessions, investment growth, productivity growth, and employment growth all remained statistically significantly lower. In addition, this study compared the effects of recessions with those of other adverse events, such as banking crises and epidemics. The long-term effect of recessions was somewhat deeper than that of banking crises and more broad- based than that of epidemics. Understanding the behavior of potential growth is of fundamental importance to short- and long-run macroeconomic analyses and policy formulation. The new database will facilitate future research on a number of topics related to potential growth. 17 Role of human capital accumulation in driving potential growth. To improve estimates of potential growth based on the production function approach, broader measures of human capital could be constructed, using information beyond the education enrollment and completion metrics and life expectancy data used in this study. The COVID-19 pandemic demonstrated the critical importance of a broader measures of human capital that takes into account such factors as morbidity and the quality of schooling (Angrist et al. 2021; World Bank 2018). The World Bank’s Human Capital Index offers one such measure but is thus far available only for very few countries and years (World Bank 2021). In addition, there is some evidence that increased human capital is more growth-enhancing in the presence of better institutions (Ali, Egbetokun, and Memon 2018). Future specifications could take into account such interaction effects. 14 Effects of climate change-related weather events on potential growth. There is growing evidence that climate change-related weather events are causing increasingly frequent and severe damage to output and that they have consequences for potential growth. Some of these are associated with increased migration (Missirian and Schlenker 2017); shorter working hours in industries with widespread outdoor labor due to excessive heat (ILO 2019); falls in total factor productivity (Economides and Xepapadeas 2018); and increased economic volatility (Panton 2020). Overall, climate change has been shown to be associated with significant output losses (Cantelmo, Melina, and Papageorgiou 2019; Colacito, Hoffman, and Phan 2018; Kahn et al. 2019). Conversely, increased investment designed either to increase resilience to adverse climate events or to mitigate climate change could provide a boost to potential growth (IMF 2019). Some of these diverging forces are explored in Kilic Celik et al. (2023). In any event, it will be essential to analyze the implications of climate change for potential growth. Role of natural resources in the measurement of potential growth. Particularly for countries that rely heavily on natural resources, production function-based estimates of potential growth could be improved by taking into account natural resources as a factor of production whose depletion can reduce potential growth. In addition, research could take into account the adverse implications of natural resources for other factors of production and productivity. For example, natural resources affect the growth benefits of foreign direct investment (Hayat 2018) and of aggregate investment (Gylfason and Zoega 2006). They can also have reduce productivity through rent-seeking behavior (Torvik 2002) and sectoral shifts (Stokke 2008). Implications of emerging trends in drivers of growth. Measures of TFP based on the production function approach could be refined to capture new developments. For example, the energy transition could generate large sectoral shifts, with consequences for TFP growth, and major investments (IMF 2021). The broadening use of digital technologies, the shift from trade in goods to trade in equipment services (“servitization”), and shifts in global value chains could change the patterns of cross-country technology transfers and hence affect productivity growth and foreign direct investment flows. Servitization and digitalization have been associated with productivity gains in the affected firms and 14 Loayza and Pennings (2022) have developed tools to model long-term growth. These include applications such as how public investment affects growth, the determinants of TFP, and the evolution of growth in resource-rich economies. 18 industries (Cette, Nevous, and Py 2022; Gal et al. 2019). Conversely, concerns have been raised that friendshoring or nearshoring of global value chains may be associated with productivity losses (Moran and Oldenski 2016; Quian, Liu, and Steenbergen 2022). Better measures of output gaps. Output gap estimates are important inputs into macroeconomic policy decisions, especially monetary ones. Hence, multivariate filter-based potential growth estimates could be tailored to capture more closely the relationship between domestic inflation and domestic monetary policy by controlling for additional external factors. These include global output gaps, global commodity price cycles, and global financial cycles. Especially for EMDEs, estimates could also be extended backwards in time and systematically tested, and adjusted, for major structural breaks. 19 FIGURE 1 Estimates of potential growth A. Advanced-economy average annual B. EMDE average annual potential growth potential growth (range across methodologies) (range across methodologies) Percent Percent 3 8 6 2 4 1 2 0 0 2000-10 2011-21 2000-10 2011-21 C. Methodologies generating highest and D. Uncertainty in global potential growth lowest estimates of potential growth Percent of country-year pairs Percent MVF MVF CI PF MVF UVF Forecasts 8 UCM UCM CI 100 80 6 60 4 40 2 20 0 0 2000-10 2011-21 2000-10 2011-21 -2 2000 2003 2006 2009 2012 2015 2018 2021 Highest Lowest E. Standard deviation of potential growth F. Persistence in potential growth estimates, estimates, 2000-19 2000-19 Percent PF MVF UVF Forecasts Percent 2.5 1.2 PF MVF UVF Forecasts 1.0 2.0 0.8 1.5 0.6 1.0 0.4 0.5 0.2 0.0 0.0 World Advanced economies EMDEs World Advanced economies EMDEs Source: World Bank. Note: “PF” stands for production function approach, “MVF” for multivariate filter, “UVF” for univariate filter, and “Forecasts” for five-year-ahead growth forecasts from the IMF World Economic Outlook. “EMDE” = emerging market and developing economies. Aggregates refer to weighted averages ( constant real GDP weights at average 2010-19 prices and exchange rates). A.B. Blue bars denote production function-based estimates. Orange whiskers indicate the range of eight estimates. C. Graph shows the share of country year pairs during each period in which each methodology generates the highest or the lowest estimate of potential growth. Only country-year pairs are considered for which at least two methodologies are available. “UVF” stands for any of four univariate filters (Christiano-Fitzgerald filter, Baxter-King, Hodrick-Prescott, or Butterworth). Unbalanced sample of 30 advanced economies and 25 EMDEs for 1998-21. 20 D. “UCM CI” and “MVF CI” are 95 percent confidence bands of each methodology. Unbalanced sample of 30 advanced economies and 25 EMDEs for 2000-21. E. Standard deviation of potential growth estimates over 2000-2019. “UVF” is the maximum among the univariate filters. Unbalanced sample of 30 advanced economies and 40 EMDEs. F. Coefficient estimates on lagged potential growth from an AR1 regression of global, advanced-economy, and EMDE potential growth for 2000-2019. “UVF” is the minimum among the univariate filters. Unbalanced sample of 30 advanced economies and 25 EMDEs for 2000-21. 21 FIGURE 2 Comparison of potential growth estimates A. Correlation of potential growth, 2000-21 B. Correlation of potential growth estimates with actual growth, 2000-20 For. For. Correlation coefficient PF MVF HP BK CF BW (WEO) (CF) UCM 1.0 PF 0.8 MVF 0.6 HP 0.4 BK CF 0.2 BW 0.0 For. (WEO) -0.2 For. (CF) PF MVF UVF For. UCM UCM Source: World Bank staff estimates. Notes: “PF” stands for production function approach; “HP” for Hodrick-Prescott filter; “BK” for Baxter- King filter; “MVF” for multivariate filter; “CF” for Christiano-Fitzgerald filter; “For. (WEO)” or “For.” for five-year-ahead growth forecasts from the IMF World Economic Outlook database; “For. (CF)” for five- year-ahead growth forecasts from the Consensus Economics; “UCM” for Unobserved Components Model. A. Figure shows the within-country correlation during 2000-20 between different measures of potential growth. Red represents greater than 80 percent, orange represents 60-80 percent, yellow represents 40-60 percent, and light blue represents 20-40 percent. Unbalanced sample of 37 advanced economies and 63 EMDEs for 2000-21. B. Blue bars show the median of within-country correlation during 2000-20 between different measures of potential growth and actual growth. Orange whiskers represent the 25th and 75th percentiles of within- country correlation during the same period. Unbalanced sample of 37 advanced economies and 95 EMDEs for 2000-20. 22 FIGURE 3 Evolution of potential growth A. Potential growth B. Share of economies and GDP with potential growth below 2000-10 average, 2011-21 Percent Percent Potential growth 100 8 Actual growth 6 80 2000-2021 potential growth 4 60 2 40 0 20 2000-10 2011-21 2000-10 2011-21 2000-10 2011-21 0 World Advanced EMDEs economies World AEs EMDEs C. Per capita potential growth D. Global potential growth Percent Percent 2000-10 2011-21 6 Potential growth 5 Actual growth 4 4 2000-2021 potential growth 2 3 0 2 2000-10 2011-21 2000-10 2011-21 2000-10 2011-21 1 0 World AEs EMDEs PF MVF UVF For. UCM Sources: World Bank, UN population statistics. Note: AEs = advanced economies; EMDEs = emerging market and developing economies. A.B.C. Based on potential growth derived using production function approach. GDP-weighted average. Sample includes 30 advanced and 53 emerging market and developing economies. B. Number of economies and their share of global or group GDP with potential growth in each period below its 2000-10 average. Horizontal line indicates 50 percent. Unbalanced sample of 30 advanced economies and 53 EMDEs for 2000-21. D. Based on common sample of 30 advanced economies and 25 EMDEs for 2000-21 to ensure consistency in samples across methodologies. Orange whiskers indicate range implied by GDP-weighted average of country-specific standard deviations of potential growth estimates for each approach. 23 FIGURE 4 Drivers of potential growth A. Contributions to potential growth B. Contributions to potential growth Percent Percent 10 TFP Capital Labor Potential growth 10 TFP Capital Labor Potential growth 8 8 6 6 4 4 2 2 0 0 2000-21 2000-10 2011-21 2000-21 2000-10 2011-21 2000-21 2000-10 2011-21 2000-21 2000-10 2011-21 2000-21 2000-10 2011-21 World AEs EMDEs EMDEs EMDEs excl.China C. Potential growth in EMDE regions D. Potential growth in EMDE regions Percent Potential growth Percent Potential growth 10 10 Actual growth Actual growth 2000-2019 potential growth 8 2000-2019 potential growth 8 6 6 4 4 2 2 0 0 2000-10 2011-21 2000-10 2011-21 2000-10 2011-21 2000-10 2011-21 2000-10 2011-21 2000-10 2011-21 MNA SAR SSA EAP ECA LAC E. Share of economies with potential growth F. Share of economies with potential growth below 2000-10 average, 2011-21 below 2000-10 average, 2011-21 Percent Percent 100 100 80 80 60 60 40 40 20 20 0 0 EAP ECA LAC MNA SAR SSA Source: World Bank staff estimates. Note: GDP-weighted averages of production function-based potential growth estimates. TFP growth stands for total factor productivity growth. AEs = advanced economies; EMDEs = emerging market and developing economies. A.-D. Sample of 30 advanced economies and 53 EMDEs. E.F. Number of economies and their share of GDP in a region among 53 EMDEs with potential growth in each period below its 2000-10 average. Horizontal line indicates 50 percent. Regional samples include the largest available coverage for each region. EAP stands for East Asia and Pacific (6 countries), ECA stands for Europe and Central Asia (9 countries), LAC stands for Latin America and the Caribbean (16 countries), MNA stands for Middle East and North Africa (5 countries), SAR stands for South Asia (3 countries), and SSA stands for Sub-Saharan Africa (14 countries). In all MNA countries, potential growth was higher in 2000-10 than in 2011-21 (and higher than the full-period average) because of a commodities boom in the first decade of the 2000s that was followed by a commodity price plunge, political tensions, and conflict in the second decade of the 2000s. 24 FIGURE 5 Potential growth around the global recessions of 2009 and 2020 A. World: Potential growth B. Advanced economies: Potential growth Percent Percent 5 4 Average Range Average Range 4 3 3 2 2 1 1 0 0 -1 -1 -2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 2009 global recession 2020 global recession 2009 global recession 2020 global recession C. EMDEs: Potential growth D. EMDEs excluding China: Potential growth Percent Percent 10 Average Range 8 Average Range 8 6 6 4 4 2 2 0 0 -2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 2009 global recession 2020 global recession 2009 global recession 2020 global recession Sources: World Bank; World Economic Outlook. Note: EMDEs = emerging market and developing economies. “Average” is an unweighted average of seven potential growth measures (excluding expectations). “Range” reflects the maximum and minimum. Figures show potential growth around global recessions in t=2009 and t=2020. Unbalanced sample of 30 advanced economies and 25 EMDEs for 2007-21. 25 FIGURE 6 Drivers of potential growth around the global recessions of 2009 and 2020 A. World: Contributions to potential growth B. Advanced economies: Contributions to potential growth Percentage points Percentage points 2.0 Capital TFP Labor 1.2 Capital TFP Labor 1.0 1.5 0.8 1.0 0.6 0.4 0.5 0.2 0.0 0.0 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 2009 global recession 2020 global recession 2009 global recession 2020 global recession C. EMDEs: Contributions to potential growth D. EMDEs excluding China: Contributions to potential growth Percentage points Percentage points 4.0 Capital TFP Labor 2.5 Capital TFP Labor 3.5 2.0 3.0 2.5 1.5 2.0 1.5 1.0 1.0 0.5 0.5 0.0 0.0 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 t-2 t-1 t=0 t+1 t+2 2009 global recession 2020 global recession 2009 global recession 2020 global recession Sources: World Bank; World Economic Outlook. Note: EMDEs = emerging market and developing economies. Figures show the contributions of capital, total factor productivity (TFP), and labor to potential growth around t=2009 and t=2020. Unbalanced sample of 30 advanced economies and 25 EMDEs for 2007-21. 26 FIGURE 7 Characteristics of recessions A. Share of countries with recessions B. Actual growth during recessions: World Percent Percent 100 Advanced economies 8 EMDEs 6 80 4 60 2 0 40 -2 20 -4 -6 0 -8 1981-89 1990-99 2000-07 2008-09 2010-19 2020 t-3 t-2 t-1 t=0 t+1 t+2 t+3 t+4 t+5 C. Actual growth during recessions: Advanced D. Actual growth during recessions: EMDEs economies Percent Percent 8 8 6 6 4 4 2 2 0 0 -2 -2 -4 -4 -6 -6 -8 -8 t-3 t-2 t-1 t=0 t+1 t+2 t+3 t+4 t+5 t-3 t-2 t-1 t=0 t+1 t+2 t+3 t+4 t+5 Source: World Bank. Note: Recessions are defined as the period from the peak preceding a business cycle trough to the trough, with a trough defined as a year in which output growth is both negative and at least one standard deviation below its long-term average. Sample includes 91 recession events in 33 advanced economies and 190 recession events in 77 EMDEs during 1981-2020. B. Unweighted averages of actual growth during recessions as defined in annex 2E denotes the peak year preceding the recession. 27 FIGURE 8 Effects of recessions on potential growth A. World: Response of potential output B. Advanced economies: Response of potential growth after recessions output growth after recessions Percentage points Percentage points 0.0 0.0 -0.5 -0.5 -1.0 -1.0 -1.5 -1.5 -2.0 -2.0 -2.5 -2.5 Year 1 Year 3 Year 5 Year 1 Year 3 Year 5 C. EMDEs: Response of potential output D. Share of adverse events associated with growth after recessions recessions Percentage points Percent 0.0 With recessions Without recessions -0.5 100 80 -1.0 60 -1.5 40 -2.0 20 -2.5 0 Year 1 Year 3 Year 5 Banking crises Epidemics Source: World Bank. Note: Recessions are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years in which output growth is both negative and one standard deviation below the long-term average. Banking crises are identified as in Laeven and Valencia (2020). Epidemics include SARS (2003), swine flu (2009), MERS (2012), Ebola (2014), and Zika (2016). A.-C. Blue bars are coefficient estimates from local projections model. Orange whiskers indicate 90 percent confidence interval. Methodological details are in annex E. Sample includes unbalanced panel of 28 advanced economies 50 EMDEs for 1998-2020. D. Share of events associated with recessions is the share of events that coincide with a recession in a 3- year window, out of the total number of events. Sample includes unbalanced panel of 33 advanced economies and 98 EMDEs for 1981-2020. 28 FIGURE 9 Effects of banking crises and epidemics on potential growth A. Response of potential output growth after B. Response of potential output growth in banking crises advanced economies 5 years later Percentage points Percentage points 0.0 0 -0.5 -1 -1.0 -1.5 -2 -2.0 -2.5 -3 Year 1 Year 3 Year 5 Recessions Banking crises Epidemics C. Response of potential output growth in D. Response of potential output growth after EMDEs 5 years later Epidemics Percentage points Percentage points 2 0.0 1 -0.5 0 -1.0 -1 -1.5 -2 -2.0 -3 -2.5 Recessions Banking crises Epidemics Year 1 Year 3 Year 5 Source: World Bank. Note: Blue bars are coefficient estimates from local projections model. Orange whiskers indicate 90 percent confidence interval. Methodological details can be found in annex E. Recessions are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years in which output growth is both negative and one standard deviation below the long-term average. Banking crises are identified as in Laeven and Valencia (2012, 2018, 2020). Epidemics include SARS (2003), swine flu (2009), MERS (2012), Ebola (2014), and Zika (2016). Sample includes unbalanced panel of 32 advanced economies 97 EMDEs for 1981-2020. 29 FIGURE 10 Effects of adverse events on growth of employment, TFP, and investment A. Response of potential TFP growth after B. Response of investment growth after recessions recessions Percentage points Percentage points 0.0 0 -5 -0.5 -10 -15 -1.0 -20 Year 1 Year 3 Year 5 Year 1 Year 3 Year 5 C. Response of employment growth after C. Response of employment growth after 5 recessions years later Percentage points Percentage points 0.5 1.0 0.0 0.5 -0.5 0.0 -1.0 -0.5 -1.5 -2.0 -1.0 -2.5 -1.5 Year 1 Year 3 Year 5 Banking crises Epidemics E. Response of potential TFP growth 5 years F. Response of investment growth 5 years later later Percentage points Percentage points 0.0 2.0 0.0 -2.0 -0.5 -4.0 -6.0 -1.0 -8.0 Banking crises Epidemics Banking crises Epidemics Source: World Bank. Note: Blue bars are coefficient estimates from local projections model. Orange whiskers indicate 90 percent confidence interval. Recessions are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years in which output growth is both negative and one standard deviation below the long-term average. Banking crises are identified as in Laeven and Valencia (2020). Epidemics include SARS (2003), swine flu (2009), MERS (2012), Ebola (2014), and Zika (2016). Sample includes unbalanced panel of 32 advanced economies 97 EMDEs for 1981-2020. 30 FIGURE 11 Effects of adverse events on growth of employment, TFP, and investment in advanced economies and EMDEs A. EMDEs: Response of potential TFP growth B. EMDEs: Response of investment growth 5 5 years later years later Percentage points Percentage points 0.5 9 6 0.0 3 0 -0.5 -3 -1.0 -6 -9 -1.5 -12 Recessions Banking crises Epidemics Recessions Banking crises Epidemics C. EMDEs: Response of employment growth 5 D. Advanced economies: Response of potential years later TFP growth 5 years later Percentage points Percentage points 2 0.2 0.0 1 -0.2 0 -0.4 -1 -0.6 -0.8 -2 Recessions Banking crises Epidemics -1.0 Recessions Banking crises Epidemics E. Advanced economies: Response of F. Advanced economies: Response of investment growth 5 years later employment growth 5 years later Percentage points Percentage points 6 2 4 2 0 0 -2 -4 -2 -6 -8 -4 -10 Recessions Banking crises Epidemics Recessions Banking crises Epidemics Source: World Bank. Note: Blue bars are coefficient estimates from local projections model. Orange whiskers indicate 90 percent confidence interval. Recessions are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years in which output growth is both negative and one standard deviation below the long-term average. Banking crises are identified as in Laeven and Valencia (2012, 2018, 2020). Epidemics include SARS (2003), swine flu (2009), MERS (2012), Ebola (2014), and Zika (2016). Sample includes unbalanced panel of 32 advanced economies 97 EMDEs for 1981-2020. 31 ANNEX A Production function approach The production function approach assumes that potential output can be captured by a Cobb-Douglas production function with constant returns to scale (Solow 1957): 15 Yt = AtKtaLt(1-a) , where Yt is potential output, At is potential total factor productivity (TFP), Kt is the potential capital stock, and Lt is potential employment. To extend the sample beyond 2019—the latest available data from Penn World Tables—TFP was recalculated as the Solow residual of output, employment (extended using data from Haver Analytics) and capital (extended using investment data from Haver Analytics and the perpetual inventory method; table 3). Labor and capital shares are the within-country averages of those reported in Penn World Tables. Human capital is not separately accounted for in the production function approach but affects TFP growth and labor supply growth, as described below. Two of the three components of potential output—potential TFP and potential employment—are proxied by the fitted values from panel regression estimates. The third component, the contribution of capital to potential growth, is assumed to be the same as the contribution of capital to actual growth, as shown in the Penn World Tables (and extended using data from Haver Analytics). This approach yields an unbalanced panel dataset for 30 advanced economies and 64 EMDEs for 1998-2021 (table 4). The same approach, using appropriate assumptions, can be used to project potential growth into the future. These assumptions and the approach for projections for 2022-32 are detailed in Kilic Celik, Kose, and Ohnsorge 2023. Capital stock data from Penn World Tables 10.0 is used until the latest available year in the dataset (2019 for most countries in the sample). For 2020-21, investment data are compiled from national statistical agencies and Haver Analytics, while the capital stock is estimated from investment data by the perpetual inventory method using historical average depreciation rates. 16 Potential TFP growth is defined as the fitted value of a panel fixed effects regression for 33 advanced economies and 92 EMDEs for 1983-2020 of Hodrick Prescott-filtered trend of actual TFP growth (the Solow residual) on determinants of productivity. These include GDP per capita relative to advanced economies, education (secondary school completion rate), the working-age share of the population, and the five-year moving average real investment growth (as in Abiad, Leigh, and Mody 2007; Bijsterbosch and Kolasa 2010; Feyrer 2007; Turner et al. 2016). 17 To allow for nonlinearities in the productivity dividends 15 The potential growth estimates may be biased if the assumption of constant returns to scale is not valid (Dribe et al. 2017). For a detailed discussion of drawbacks of growth accounting, see Dieppe and Kilic Celik (2021). That said, the approach is widely used for its conceptual simplicity and ease of interpretation. 16 Implicitly, this approach does not account for the possibility that inefficient investment is written off during downturns. Hence, it may overstate the capital stock during downturns. 17 The results are robust to using GDP per capita instead of GDP per capita in percent of advanced-economy GDP per capita. GDP per capita relative to a frontier (advanced economies) is used here to proxy the catch-up effect highlighted in the literature on stochastic frontier analysis (Growiec et al. 2015). 32 from education, schooling is interacted with a dummy for schooling in the bottom two- thirds across the sample. A dummy is included for commodity exporters during the period 2003-07. This dummy is intended to capture the impact of the exceptionally large commodity price boom that temporarily lifted commodity exporters’ growth during this period. Potential TFP is thus: Δtfpi,t = α0 + α1 GDP per capitai,t + α2 wapi,t , + α3 educationi,t + α4 educationi,t * Dedu , + α5 Dcebi,t + α6 Δinvi,t + εi,t , where Δtfpi,t is the logarithmic first difference of trend TFP, GDP per capitai,t is GDP per capita in percent of advanced-economy per capita GDP, wapi,t is the working-age share of the population, educationi,t is the percent share of the population who completed secondary school, Δinvi,t is the five-year moving average of real investment growth, Dedu is a dummy variable taking the value of 1 if the secondary completion rate is in the bottom two-thirds of the distribution, and Dcebi,t is a dummy variable for the period 2003-07 taking the value 1 if the country is a commodity exporter. 18 The data were compiled using a wide range of sources: UN Population Statistics (for population growth, the working-age share of the population); Barro and Lee (2013) (for secondary school completion); the World Development Indicators (for secondary school completion and GDP per capita relative to the advanced economies); and Haver Analytics (for investment). The regression results are broadly in line with the previous literature (table 5). TFP growth slows as per capita incomes converge toward advanced-economy levels (Barro and Sala-i-Martin 1997). A better-educated population and accelerated investment growth are associated with higher TFP growth. However, the impact of education diminishes as education levels rise toward advanced-economy levels (Benhabib and Spiegel 1994, 2005; Coe, Helpman, and Hoffmaister 1997; Kato 2016). As a result, the coefficient on secondary school completion rates is only significant for countries with completion rates below the top third. The results are broadly robust to a number of alternative specifications (tables 5 and 6). Two different methodologies are used to estimate trend TFP growth (a linear-quadratic trend and 3, 5, and 7-year moving averages) instead of the HP-filtered trend. The 3- and 7-year rolling averages of investment growth are used. In most specifications, the coefficient estimates remain significant and retain their signs; however, the working-age 18 This approach is similar to Abiad, Leigh, and Mody (2007) and Bijsterbosch and Kolasa (2010). Abiad, Leigh and Mody (2007) estimate five-year non-overlapping averages of TFP growth as a function of per capita GDP, schooling, population growth, trade openness and a nonlinear function of current account deficits and FDI for a sample of 22 European countries for 1975-2004. Bijsterbosch and Kolasa (2010) estimate five-year non-overlapping averages of labor productivity growth as a function of relative productivity levels (which here is proxied with relative per capita GDP), the share of high-skilled workers in employment, and investment in percent of value added for sectoral data for eight European countries for 1996-2005. 33 population share became insignificant in some specifications. The inclusion of R&D spending, which is available only for a much smaller sample, and urbanization also do not materially change the results. Potential labor supply is defined as the product of the working-age population and the fitted value of age- and gender-specific regressions of labor force participation rates (lfpra,g,t) in percent on their structural determinants (Xa,g,t) and controlling for cohort effects, fixed effects, and the state of the business cycle—defined as the deviation of the logarithm of real GDP from the Hodrick-Prescott-filtered trend. The vector Xa,g,t includes gender-specific education outcomes (secondary and tertiary completion rates in percent of the population over the age of 25 and enrollment rates in percent of population of the age group that officially corresponds to the level of education, age-specific fertility rates (births per woman), and life expectancy (in years). These are interacted with a dummy variable Demde which takes the value of 1 for EMDEs. The vector Ca,g,t includes all the control variables: 19 lfpra,g,t = αa,g + βa,g Xa,g,t + γa,g Xa,g,t * Demde + δa,g Ca,g,t +εa,g,t . Data on the working-age population comes from the UN Population Statistics Database. Data for age- and gender-specific labor force participation rates are available from Key Indicators of the Labor Market (KILM) of the ILO Population Statistics Database for 1990-2019, which is spliced by Labour Force Statistics of the OECD for 1960-2020 for 33 advanced economies and 16 EMDEs. This produces data for age- and gender-specific labor force participation rates for 1960-2020 for up to 38 advanced economies and 142 EMDEs. 20 Completion rates of secondary and tertiary education are from Barro and Lee (2013) and the World Bank’s World Development Indicators; age-specific fertility rate and life expectancy are from the UN’s World Population Projections database; gender-specific secondary and tertiary school enrollment rates are from the World Development Indicators. The regression sample includes up to 35 advanced economies and 133 EMDEs for 1987-2020. 21 The regression results are broadly in line with findings in the previous literature (table 7). First, among teenage and younger women, fertility rates are associated with higher labor force participation as mothers are more likely to discontinue their education and 19 This approach combines those by Fallick and Pingle (2007) and Goldin (1994). For the United States, Fallick and Pingle (2007) estimate labor force participation by age group and gender as a function of cohort and age fixed effects as well as business cycle fluctuations. Goldin (1994) models aggregate labor force participation rates as a function of country-level variables such as female schooling. The regression used here incorporates both cohort effects and country-level variables modelling human capital and other factors driving labor force participation. 20 This is an unbalanced sample because some of the exogenous variables are not available for the full period for all countries. However, the regression results are robust to restricting the sample to the balanced panel with fully available data. 21 Since UN data for life expectancy is only available for five-year periods, historical life expectancy data from the World Developing Indicators database is used. For projection years or missing data, UN World Population Statistics are spliced with data from World Development Indicators database. 34 participate in the labor force, especially in advanced economies (Azevedo, Lopez-Calva, and Perova 2012; Fletcher and Wolfe 2009; Herrera, Sahn, and Villa 2016). This effect is more muted in EMDEs, potentially reflecting an earlier average age of marriage, which tends to be associated with lower female labor force participation (United Nations 2012). Second, for relevant age groups, educational attainment is associated with higher participation rates, except for young men and women aged 20-24. The positive correlation between completion rates and labor force participation may partly reflect higher compensation for more educated workers. For young men, higher tertiary educational attainment is associated with lower labor force participation. This might reflect the lack of demand for employment in sectors where these educated workers would expect to be employed, discouraging them from labor force participation (Klasen and Pieters 2013). However, for men aged 50-64 and all workers aged 65 years and older, education becomes an insignificant determinant of labor force participation (as in Fallick and Pingle 2007). Tertiary enrollment rates in all relevant age groups are associated with lower labor force participation rates, as students devote time to completing their degree (Kinoshita and Guo 2015; Linacre 2007; and Tansel 2002). Third, life expectancy is one of the main determinants of participation for workers aged 50 and above (Fallick and Pingle 2007). For the younger ones among them, between the ages of 50-64, higher life expectancy is associated with higher labor force participation, possibly reflecting the need to accumulate savings for a longer retirement period or the positive association between better health among older workers and higher incomes (Haider and Loughran 2001). Among those aged 65 years or older, higher life expectancy is associated with higher labor force participation in advanced economies, but does not significantly change participation in EMDEs. Life expectancy may be a weak proxy for a healthy old age in EMDEs with less-developed health care systems or where differences in life expectancy might mostly reflect differences in infant mortality (Eggleston and Fuchs 2012). Fourth, labor force participation is procyclical—albeit less so in EMDEs than in advanced economies—in most age groups until the age of 50. Labor force participation rises when real GDP is above its HP-filtered trend and declines when real GDP is below its HP- filtered trend. 22 As the age increases, the sensitivity to cyclicality decreases and participation eventually becomes countercyclical (Balakrishnan et al. 2015; Duval, Eris, and Furceri 2011). This may reflect greater ability of more experienced workers to remain employed or return to employment after spells of unemployment during recessions (Elsby, Hobijn, and Şahin 2015; Shimer 2013). However, participation becomes pro-cyclical again (although not statistically significant) for workers aged 65 and above as they become eligible to retire and may be readier to drop out of the labor force in a weaker economy. This result is broadly robust to defining the business cycle as deviations of real GDP from the 10-year moving average or from a linear-quadratic trend (tables 8, 9). 22 In several instances, there were no statistically significant differences between advanced economies and EMDEs in the cyclicality of their labor force participation. Hence, the interactions were omitted from the regressions. 35 ANNEX B Univariate filters Univariate statistical filters decompose a series yt into trend, cyclical, and noise components. The trend component is used as a proxy for potential output. Although they are all essentially weighted moving averages of the series yt , they differ in their weights. Five univariate filters are applied to estimate potential output: filters based on Hodrick and Prescott (1997), three band-pass filters (Baxter and King 1999; Butterworth 1930 and Gomez 2001; Christiano and Fitzgerald 2003), and a filter based on an Unobserved Components Model. The measures are estimated for 37 advanced economies and 52 EMDEs for 1980Q1-2022Q1 (table 10). Forecasts from the Global Economic Prospects report provide data to 2024. A smaller sample is used in comparisons with other approaches, to ensure consistency of samples (tables 11 and 12). Hodrick-Prescott filter The Hodrick-Prescott (HP) filter minimizes deviations of a series yt from its trend τt , assuming a degree of smoothness λ of the trend. The HP filter chooses the trend τt that minimizes: T T −1 2 ∑ ( yt −τ t ) + λ ∑ [ (τ t +1 −τ t ) −(τ t −τ t −1) ] 2 t =1 t =2 , where T is the sample size. A larger λ indicates a smoother trend. For λ=0, the trend is equal to the actual series and for λ>+∞ the trend is a linear time trend with a constant growth rate. Typically, the value of λ is set at 1600 for quarterly data. The trend is estimated based on past values as well as projected values of the series yt. Band-pass filters The three band-pass filters aim to isolate fluctuations in a time series which lie in a specific band of frequencies. They eliminate slow-moving components (trend) and very high frequency components (noise) and define the intermediate components as the business cycle. Specifically, the three band-pass filters differ in their approximations of the optimal linear filter (also known as the “ideal” band-pass filter) to deal with finite time series. The Baxter and King (BK) filter is a moving average of the data with symmetric weights on lags and leads. Therefore, it loses observations in the beginning and towards the end of the sample. It is particularly well-suited when the raw series follows a near-independent and identically distributed process (Christiano and Fitzgerald 2003). Specifically, the BK filter is given by: ˆt = b( L) yt c , where b(L) is the lag polynomial given by: k b( L ) = ∑b L k j j j= −k , 36 with bkj = bk‒j . Note that k observations will be lost in both ends of the sample. The higher k, the closer the filter is to the ideal filter but also the higher are the number of lost observations. The default business cycle frequencies used here (required for estimation) are between 1.5 to 8 years. The Christiano and Fitzgerald (CF) filter is a one-sided moving average of the data with weights that minimize the distance between the approximated and the “ideal” filter. Since the filter is one-sided, it does not lose observations towards the end of the sample. It is most suitable for random-walk series. The optimal cycle at time t is given by: p ˆt = c ∑b j=− f p, f j yt − j , where are the optimal weights of the CF filter that solve: Min p, f E  ˆt − ct ) 2 y  (c  bj , and ct is the filtered series under the “ideal” (infinite sample) band-pass filter. By default, the CF filter business cycle frequencies are set between 1.5 to 8 years. The Butterworth (BW) filter—widely used in electrical engineering for signal extraction— isolates only low-frequency fluctuations, not high-frequency ones. Pollock (2000) proposes the use of this filter for macroeconomic time series filtering as an alternative to the traditional linear filters such as the Hodrick-Prescott filter. The low-pass BW filter is characterized by two parameters λ and n and can be specified as: , where L is a lag operator, λ is the smoothness parameter and n is the degree of the filter. Unobserved Components Model Most univariate filters can be nested into the Unobserved Components Model. 23 In contrast to other univariate filters, the Unobserved Components Model does not impose specific parameter assumptions about the degree of smoothing, lead and lag windows, or business cycle frequencies. Instead, it relies on assumptions about the underlying process followed by output gaps and potential growth, and is estimated using the Kalman filter (Harvey 1990): LYt = LȲt + YGAPt , (1) 23 For example, if the trend and cyclical components are uncorrelated white noise, the unobserved components model coincides with the Hodrick-Prescott filter if the noise-to-signal ratio matches the Hodrick- Prescott filter’s smoothing parameter (Hamilton 2018). 37 LȲt = LȲt-1 + Gt + εȲt , (2) Gt = (1 - τ)Gss + τ Gt-1 + εGt , (3) YGAPt = β1YGAPt-1 + β2YGAPt – 2 + γtYGAP , (4) where LY is the log of seasonally adjusted quarterly real GDP, LȲ the log of potential output, YGAP the output gap, Gt potential output growth, Gss the steady state level that growth is assumed to converge to over the long term, and εY and εG are independently and identically distributed disturbances. Note that the shock εY shifts the level of potential output whereas εG is a shock to potential output growth. Equation (3) assumes that potential growth converges (at a speed of convergence τ) to its steady level Gss after a shock. The output gap follows a commonly used second-order autoregressive process (equation 4). The Kalman filter algorithm yields (posterior) time-varying variance- covariance matrices for the smoothed estimates of the unobserved state variables, potential growth and the output gap. The standard deviation of potential growth is used to calculate the 95 percent confidence band around estimated potential growth. 38 ANNEX C Multivariate filters The unobserved components model can be expanded to include additional indicators of domestic demand pressures to help identify the output gap (Benes et al. 2010). The most commonly used indicators are inflation and the unemployment rate. Specifically, the univariate model (1-4) is further augmented with a Phillips Curve relationship between inflation and output gaps (equation 5), an Okun’s Law relationship between unemployment rates and output gaps (equations 6-9), a relationship between capacity utilization and output gaps (equations 10-13), and a set of equations describing the Taylor rule (equations 14-17). Given the large variation in available data across economies, switches are employed to add selected equations to each country model based on the country’s specific dataset. If house prices or the unemployment rate data is not available for a specific country, the relevant equations would not be included. At minimum, all countries have output, inflation, and commodity price data. 24 Model components The Phillips Curve relates inflation to the output gap, controlling for the impact of supply side shocks such as import prices on domestic inflation. πt = ρ π t–1 + (1 – ρ)π t+1 + α1YGAPt + λ1πmt + επ , (5) where πt is annualized quarter-on-quarter inflation at time t, πmt is import price inflation at time t, and YGAPt is the output gap at time t. Expectations are assumed to be an average of adaptive and rational expectations, weighted by ρ. Inflation expectations are linked to fixed horizon forecasts of inflation from Consensus Economics where available. 25 Okun’s Law relates the unemployment gap UGAPt (defined as the difference between the actual unemployment rate Ut and the equilibrium, or natural, unemployment rate Ūt in equation 6) to the output gap (in equation 7) as: UGAPt = Ut – Ūt , (6) UGAPt = γUGAPt–1 – α2YGAPt + εtUGAP . (7) Following Blagrave et al. (2015), the equilibrium unemployment rate process is specified in deviation from steady state. Equation (8) specifies the process for Ut . It implies that following a shock, the non-accelerating inflation rate of unemployment (NAIRU) Ūt converges back to its steady state value Uss according to the parameter τ1 and has a trend component GU which has an autoregressive process (9): 24 Three economies—Lesotho, Namibia, and Tanzania—have only output, inflation, and commodity price data. 25 Fixed-horizon forecasts transform the fixed-event forecasts (for example, for 2022 and 2023) provided by Consensus Economics to be one year-ahead forecasts (in other words, at a fixed horizon in the future). See Bordo and Siklos (2017) and Siklos (2013) for details. 39 Ūt - Uss = τ1(Ūt-1- Uss) + GUt + εUt , (8) GUt = τu GUt-1 + εGt , (9) Since capacity utilization Ct is highly pro-cyclical, it can help identify the cyclical component of output even when other indicators (such as, say, a stable unemployment gap during jobless recoveries or stable inflation in highly open economies) do not signal cyclical upturns. Equations (10)-(13) describe the relation between capacity utilization and output gaps and the exogenous process for capacity utilization, where is the steady state of capacity utilization rate, CGAPt is the capacity utilization gap, defined as the difference between actual and non-inflationary capacity utilization Ct, and GCt is the growth of capacity utilization: CGAPt q CGAPt −1 + α 3YGAPt   = +εCGAPt Ct = CGAPt +    Ct τ2 (Ct −1 − Css ) + GCt + εCt Ct − Css = GCt = τc GCt −1 + εGt      A Taylor rule describes monetary policy in economies where short-term policy interest rates are used as an instrument of monetary policy: it = τi it −1 + (1 − τi )(rt* + π* * t + γ π ( πt + 4 − πt ) + γ YGAPYGAPt ) + ε it , (14) where it is the nominal policy interest rate that responds to forecast inflation from its target (πt*) and the output gap. The ex ante real interest rate is defined using the Fisher equation as: rt= it – π4t +1, (15) where π4t +1 is the year-on-year change in consumer prices. The neutral real interest rate is modelled as in Laubach and Williams (2003): rt* cGt + Z t , = (16) Z t Z t –1 + ε Zt                                , = (17) An output gap process closes the model. Inflation and unemployment might fail to capture all domestic demand pressures, such as credit or asset price growth or commodity price cycles. 26 This may lead to an underestimation of the output gap and an overestimation of potential output, especially at the peak of the cycle. Instead of assuming that the output gap process is exogenous, as in the traditional multivariate Kalman filter, three additional indicators are included in the output gap equation: house price, credit, and commodity 26 See Borio (2013, 2014) and Summers (2014) for advanced economies, Jesus et al. (2015) for Latin America and the Caribbean, Kemp (2015) for South Africa, and Enrique et al. (2016) for East Asia and the Pacific. The cyclical component of copper prices helps explain mining sector output gaps in Chile (Blagrave and Santoro 2016). 40 price growth: , (18) where crt , hprt , and comprt are cyclical components of year-on-year private sector credit growth deflated by consumer price inflation, quarterly seasonally-adjusted house prices, and export-weighted real average commodity prices, respectively, and rt-rt* is the deviation of the real policy rate from its equilibrium level. Estimation The model uses the Kalman filter algorithm and Bayesian techniques on quarterly data covering 1980Q1-2022Q2 for up to 36 advanced economies and 54 EMDEs. A key parameter determining the shape of potential output is the variance of the output gap relative to potential growth innovations. The variance of the innovations εYGAPt and εGt are set such that their ratio equals the typically used smoothness parameter of the Hodrick-Prescott filter. The prior for the elasticity of output gap with respect to commodity price β3 (the central bank’s response to deviations of inflation from target) and the coefficient on potential growth in the neutral real interest rate follows a normal distribution in the case of commodity prices to allow for a potentially negative impact of commodity price increases in commodity importers. The prior distributions for all standard deviations are inverse gamma distributions. All other estimated priors follow a beta distribution. The standard deviations of εCGAPt and εUGAPt are set as the OLS standard errors of equations (5) and (9) based on Hodrick-Prescott-filtered data. Steady state values of growth, unemployment, and capacity utilization are calibrated to the sample means of their corresponding HP-filtered series. Estimates of potential growth from the Multivariate Filter Model and the Unobserved Components Model used in this paper are based on LȲt and include both level and growth shocks to potential growth. As in the case of the Unobserved Components Model, the Kalman filter algorithm yields (posterior) time-varying variance-covariance matrices for the filtered estimates of all unobserved state variables, including potential growth. From this matrix, the standard deviation of potential growth is used to calculate the 95 percent confidence band around estimated potential growth. Data Based on the univariate and multivariate filters, output gaps and potential growth are estimated for up to 37 advanced economies and 52 EMDEs for as long a period as 1980Q1- 2024Q4 (table 10). A smaller sample is used in comparisons with other approaches, to ensure constant samples (tables 11 and 12). GDP, inflation, unemployment rates, private sector credit growth, and capacity utilization rates are from Haver Analytics. House price growth is from Bank for International Settlements, commodity prices are from the World 41 Bank’s Pink Sheet, and export weights are from the UN Comtrade database. Country- specific output gaps are aggregated using real GDP weights at 2010-19 exchange rates and prices. ANNEX D Long-term growth expectations Expectations of output growth over long horizons capture forecasters’ assessment of long- term sustainable growth since they are stripped of unpredictable short-term shocks. Two sources of expectations are used: the International Monetary Fund’s World Economic Outlook (WEO) database, published twice a year, and Consensus Economics, published on a quarterly basis. Since the longest available forecast horizon is 5-years for IMF’s WEO, 5-year-ahead forecasts are selected for both sources for consistency across these two measures. The IMF’s WEO provides five-year-ahead forecasts for up to 173 countries (37 advanced economies, 136 EMDEs) for 1990-2021. Consensus forecasts are available for up to 78 countries (34 advanced economies and 44 EMDEs) for 1990-2022 and the database includes the April vintages. 42 ANNEX E Local projection estimation A local projection estimation is used to explore the evolution of potential growth, employment growth, potential TFP growth, and investment growth following recessions, banking crises, and epidemics. The model estimates the cumulative impact of recessions, following Jordà (2005) and Teulings and Zubanov (2014). 27 In impulse responses, the model estimates the effect of short-term shocks (the recession, banking crisis, or epidemic event) over a horizon h on potential growth while controlling for other determinants: yi,t+h – yi,t = αh + βh shocki,t + γh ∆yi,t–1 + fixedeffectsi + εi,t , where yi,t is potential growth. The model controls for country-fixed effects to capture time- invariant cross-country differences. 28 The variable shocki,t is a dummy variable for a recession event (or banking crisis or epidemic), the main variable of interest. Lagged potential growth yi,t–1 controls for the history of potential growth. For channels, the same specification is used, where yi,t is employment growth, potential TFP growth, or investment growth. This model also controls for country-fixed effects to capture time-invariant cross-country differences. Lagged potential growth yi,t–1 controls for the history of employment growth, potential TFP growth, or investment growth. Banking crises are defined as in Laeven and Valencia (2018) and the ones corresponding to the potential growth measures are listed in table 13. Epidemics include SARS (2003), swine flu (2009), MERS (2012), Ebola (2014), and Zika (2016) and affected countries are listed in table 14. Results for the impact of recessions, banking crises, and epidemics on alternative measures of potential growth are shown in tables 15-18. Results for the impact of recessions, banking crises, and epidemics on employment, total factor productivity, and investment growth are shown in tables 19-20. 27 Plagborg-Møller and Wolf (2021) show that vector autoregression (VAR) and LPM estimations yield the same impulse response functions but Li, Plagborg-Møller and Wolf (2022) show that LPM estimators have larger variance (but lower bias), especially for the medium- and long-term horizons, than VAR estimators. 28 A dummy for time effects is not necessary because the time variable t refers to the time since the start of the event and pertains to different years for different countries. 43 TABLE 1 Methodology time and country coverage Methodology Time coverage* Advanced economies EMDEs Production 1998-2032 30 (AUS, AUT, BEL, CAN, 64 (ALB, ARG, ARM, BDI, BEN, BGD, BGR, function CHE, CYP, DEU, DNK, BOL, BRA, BRB, CAF, CHL, CHN, CMR, COL, approach ESP, EST, FIN, FRA, CRI, DOM, ECU, EGY, GAB, GTM, HND, HUN, GBR, GRC, HKG, HRV, IDN, IND, IRN, IRQ, JAM, JOR, KAZ, KEN, IRL, ISR, ITA, JPN, KOR, KGZ, LAO, LSO, MAR, MDA, MEX, MNG, LTU, LVA, NLD, NOR, MOZ, MRT, MUS, MYS, NAM, NER, NIC, PAK, PRT, SVK, SVN, SWE, PER, PHL, POL, PRY, QAT, ROU, RWA, SDN, USA) SEN, SRB, TGO, THA, TJK, TUN, TUR, URY, VNM, ZAF) Multivariate 1981-2024 37 (AUS, AUT, BEL, CAN, 52 (ALB, ARG, AZE, BGR, BHR, BLZ, BOL, filter CHE, CYP, CZE, DEU, BRA, BWA, CHL, CHN, CMR, COL, CRI, DOM, DNK, ESP, EST, FIN, ECU, EGY, GEO, GTM, HND, HUN, IDN, IND, FRA, GBR, GRC, HKG, IRN, JOR, KAZ, KEN, KWT, LSO, MAR, MEX, HRV, IRL, ISL, ISR, ITA, MKD, MNG, MYS, NAM, NGA, NIC, PAN, PER, JPN, KOR, LTU, LUX, PHL, POL, PRY, ROU, SAU, SLV, THA, TUN, LVA, MLT, NLD, NOR, TUR, TZA, URY, VNM, ZAF) NZL, PRT, SGP, SVK, SVN, SWE, TWN, USA) Univariate 1980Q1-2024Q4 37 (AUS, AUT, BEL, CAN, 52 (ALB, ARG, AZE, BGR, BHR, BLZ, BOL, filters CHE, CYP, CZE, DEU, BRA, BWA, CHL, CHN, CMR, COL, CRI, DOM, DNK, ESP, EST, FIN, ECU, EGY, GEO, GTM, HND, HUN, IDN, IND, FRA, GBR, GRC, HKG, IRN, JOR, KAZ, KEN, KWT, LSO, MAR, MEX, HRV, IRL, ISL, ISR, ITA, MKD, MNG, MYS, NAM, NGA, NIC, PAN, PER, JPN, KOR, LTU, LUX, PHL, POL, PRY, ROU, SAU, SLV, THA, TUN, LVA, MLT, NLD, NOR, TUR, TZA, URY, VNM, ZAF) NZL, PRT, SGP, SVK, SVN, SWE, TWN, USA) WEO five- 1990-2022 37 (AUS, AUT, BEL, CAN, 136 (AFG, AGO, ALB, ARE, ARG, ARM, ATG, year ahead CHE, CYP, CZE, DEU, AZE, BDI, BEN, BFA, BGD, BGR, BHR, BHS, expectations DNK, ESP, EST, FIN, BIH, BLZ, BOL, BRA, BRB, BRN, BTN, BWA, FRA, GBR, GRC, HKG, CAF, CHL, CHN, CMR, COD, COG, COL, HRV, IRL, ISL, ISR, ITA, COM, CPV, CRI, DJI, DMA, DOM, DZA, ECU, JPN, KOR, LTU, LUX, EGY, ERI, ETH, FSM, GAB, GEO, GHA, GIN, LVA, MLT, NLD, NOR, GMB, GNB, GNQ, GRD, GTM, GUY, HND, HTI, NZL, PRT, SGP, SVK, HUN, IDN, IND, IRN, IRQ, JAM, JOR, KAZ, SVN, SWE, TWN, USA) KEN, KGZ, KHM, KIR, KNA, KWT, LAO, LBN, LBR, LBY, LCA, LSO, MAR, MDA, MDG, MDV, MEX, MKD, MLI, MMR, MNG, MOZ, MRT, MUS, MWI, MYS, NAM, NER, NGA, NIC, NPL, OMN, PAK, PAN, PER, PHL, PNG, POL, PRY, QAT, ROU, RWA, SAU, SDN, SEN, SLB, SLV, SOM, SRB, SSD, STP, SUR, SWZ, SYC, SYR, TCD, TGO, THA, TJK, TLS, TON, TUN, TUR, TZA, UGA, URY, UZB, VCT, VNM, VUT, WSM, YEM, ZAF, ZMB) Source: World Bank. Note: Country codes are available at https://www.iban.com/country-codes. 44 TABLE 2 Methods to estimate potential growth Methodology Advantages Disadvantages Production Produces estimates that help explain the Relies on proxies for potential productivity function approach movement of potential output in terms of and labor supply growth and capital its inputs. accumulation that could be subject to measurement errors. Relies on assumption Low correlation with actual output growth. of specific functional form. Time-series filters Univariate filters are straightforward to “End-point” problems can lead to large implement, even in data-poor revisions as new data become available. 37 environments. Multivariate filters produce output gaps Strong correlation with actual output that are consistent with indicators of growth, which could reflect short-term domestic demand pressures (inflation, shocks to potential growth or, alternatively, unemployment, current account deficits, are associated with cyclical movements. capacity utilization). Long-term growth In principle, incorporate judgment and, In practice, tend to be sticky and, at times, expectations thus, capture factors that cannot be in ways that are challenging to interpret. modelled during periods of high volatility. Source: World Bank. 45 TABLE 3 Variable list Variable Units Source Sample GDP in U.S. dollars Millions of U.S. dollars, at IMF World Economic Outlook 194 countries, market exchange rates database 1980-2021 Real GDP in local Millions of local currency Haver Analytics 93 countries, currency 1980Q2-2021Q4 GDP per capita U.S. dollars at market IMF World Economic Outlook 182 countries, exchange rates database; UN population statistics 1980-2021 Population, by age and Number UN population statistics and 184 countries, gender projections 1950-2035 Labor force, by age and Number ILO, Key Indicators of the Labour 180 countries, gender Market (KILM) database; OECD 1960-2020 Labour Force Statistics Investment growth Percent Haver Analytics 187 countries, 1961-2021 Secondary education Percent of population that Barro and Lee (2013); World 179 countries, completion rate completed secondary Development Indicators 1960-2020 education in percent of population in relevant age group Tertiary education Percent of population that Barro and Lee (2013); World 174 countries, completion rate completed tertiary education Development Indicators 1960-2020 in percent of population in relevant age group Secondary education Percent of population of the World Development Indicators 193 countries, enrolment rate age group corresponding to 1970-2020 the level of education Tertiary education Percent of population of the World Development Indicators 192 countries, enrolment rate age group corresponding to 1970-2020 the level of education Life expectancy Years UN population statistics; UN 181 countries, population projections 1985-2035 Fertility rate Number of births per 1,000 UN population statistics; UN 175 countries, women population projections 1960-2095 Employment Number Penn World Table 181 countries, 1950-2019 Urban population Share of total population World Development Indicators 194 countries, 1960-2020 R&D spending In percent of GDP World Development Indicators 144 countries, 1996-2019 Consumer price inflation Percent Haver Analytics 93 countries, 1980Q1-2021Q4 Inflation expectations Percent Consensus Economics 74 countries, 1980Q1-2021Q4 Unemployment rate Percent of labor force Haver Analytics 66 countries, 1980Q1-2021Q4 Capacity utilization rate Percent of capacity Haver Analytics 31 countries, 1980Q1-2021Q4 Import price inflation Percent Haver Analytics 74 countries, 1980Q1-2021Q4 Private credit growth Percentage points of GDP Haver Analytics 57 countries, 1980Q1-2021Q4 Average commodity Index World Bank; Federal Reserve Bank 93 countries, export price of St. Louis; UN Comtrade 1980Q1-2021Q4 Monetary policy rates Percent Haver Analytics 80 countries, 1980Q1-2021Q4 House prices Bank for International Settlements 55 countries, 1980Q1-2021Q4 46 TABLE 3 Variable list (continued) Variable Units Source Sample WEO real GDP growth Percent IMF World Economic Outlook 175 countries, forecasts database 1990-2021 Consensus real GDP Percent Consensus Economics 78 countries, growth forecasts 1990-2022 Source: World Bank. 47 TABLE 4 Coverage for production function-based estimates of potential growth Economy Sample Economy Sample Economy Sample period period period Australia 1998-2032 Europe and Central Asia Middle East and North Africa Austria 1998-2032 Albania 1998-2032 Egypt, Arab Rep. 1998-2032 Belgium 1998-2032 Armenia 1998-2032 Iraq 2001-2019 Canada 1998-2032 Bulgaria 2000-2032 Iran, Islamic Rep. 1998-2032 Cyprus 1998-2032 Hungary 1998-2032 Jordan 1998-2032 Croatia 1998-2032 Kazakhstan 1998-2032 Morocco 1998-2032 Denmark 1998-2032 Kyrgyz Republic 2000-2032 Qatar 1998-2016 Estonia 1998-2032 Moldova 2013-2032 Tunisia 1998-2032 Finland 1998-2032 Poland 1998-2032 France 1998-2032 Romania 1998-2032 South Asia Germany 1998-2032 Serbia 1998-2032 Bangladesh 1998-2032 Greece 1998-2032 Tajikistan 1998-2032 India 1998-2032 Hong Kong SAR, 1998-2032 Turkey 1994-2030 Pakistan 1998-2032 China Iceland 1998-2032 Israel 1998-2032 Latin America and Caribbean Sub-Saharan Africa Italy 1998-2032 Argentina 1998-2032 Benin 1998-2032 Japan 1998-2032 Barbados 1998-2016 Burundi 1998-2032 Korea 1998-2032 Bolivia 1998-2032 Cameroon 1998-2032 Latvia 1998-2032 Brazil 1998-2032 Central African Republic 1998-2019 Lithuania 2000-2032 Chile 1998-2032 Gabon 1998-2032 Netherlands 1998-2032 Colombia 1998-2032 Kenya 1998-2032 Norway 1998-2032 Costa Rica 1998-2032 Lesotho 1998-2032 Portugal 1998-2032 Dominican Republic 1998-2032 Mauritania 2000-2032 Slovak Republic 1998-2032 Ecuador 1998-2032 Mauritius 1998-2032 Slovenia 1998-2032 Guatemala 1998-2032 Mozambique 1998-2032 Spain 1998-2032 Honduras 1998-2032 Namibia 1998-2032 Sweden 1998-2032 Jamaica 1998-2032 Niger 1998-2032 Switzerland 1998-2032 Mexico 1998-2032 Rwanda 2000-2016 United Kingdom 1998-2032 Nicaragua 1998-2032 Senegal 1998-2032 United States 1998-2032 Paraguay 1998-2032 South Africa 1998-2032 Peru 1998-2032 Sudan 1998-2019 East Asia and Pacific Uruguay 1998-2032 Togo 1998-2032 China 1998-2032 Indonesia 1998-2032 Malaysia 1998-2032 Mongolia 1998-2032 Philippines 1998-2032 Thailand 1998-2032 Vietnam 2013-2021 Source: World Bank. Note: Methodology and assumptions underlying projections for 2022-32 are detailed in Kilic Celik, Kose and Ohnsorge (2023). 48 TABLE 5 Regression results for total factor productivity Dependent variable: Baseline 3-year moving 5-year moving 7-year moving Linear-quadratic HP-trend average average average trend TFP growth GDP per capita rel. to -0.06*** -0.07*** -0.07*** -0.06*** -0.06*** advanced economies (0.000) (0.001) (0.002) (0.002) (0.001) Working-age population 4.16* 3.05 4.70 6.86** 3.13 (0.100) (0.326) (0.143) (0.044) (0.321) Secondary completion rate 0.003 0.003 0.010 0.009 -0.029*** (0.701) (0.807) (0.375) (0.397) (0.002) Secondary completion rate 0.009* 0.012* 0.009 0.004 0.004 (bottom two-thirds) (0.061) (0.068) (0.142) (0.466) (0.464) Investment growth 0.088*** 0.178*** 0.185*** 0.169*** 0.118*** (five-year moving average) (0.000) (0.000) (0.000) (0.000) (0.000) Commodity exporters 0.592*** 1.094*** 0.778** 0.664** 1.001*** credit boom dummy (0.000) (0.002) (0.035) (0.040) (0.000) Number of observations 706 694 692 687 706 Number of countries 125 125 125 125 125 Within R-square 0.26 0.27 0.29 0.29 0.25 Source: World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. Estimations are based on standard errors clustered around countries. The methodology is defined in annex.3. Sample includes unbalanced panel of 33 advanced economies 92 EMDEs for 1983-2020. p-statistics are shown in parentheses. 49 TABLE 6 Regression results for total factor productivity Dependent variable: TFP growth HP-trend HP-trend HP-trend HP-trend GDP per capita relative to advanced economies -0.06*** -0.06*** -0.06*** -0.05*** (0.000) (0.000) (0.000) (0.000) Working-age population 5.96** 4.70 6.54** 6.13** (0.024) (0.115) (0.038) (0.047) Secondary completion rate -0.002 -0.001 0.013 0.000 (0.770) (0.847) (0.139) (0.968) Secondary completion rate 0.007 0.011** 0.012** 0.006 (bottom two-thirds) (0.125) (0.028) (0.013) (0.255) Investment growth 0.009 (three-year moving average) (0.672) Investment growth 0.084*** 0.111*** (five-year moving average) (0.000) (0.000) Investment growth 0.007 (seven-year moving average) (0.763) Commodity exporters credit boom dummy 0.953*** 0.924*** 0.557*** 0.902*** (0.000) (0.000) (0.000) (0.000) Urban population share -0.066** (0.031) R&D spending as percent of GDP -0.092 (0.752) Number of observations 778 698 706 497 Number of countries 125 125 125 109 Within R-square 0.15 0.15 0.28 0.34 Source: World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. Estimations are based on standard errors clustered around countries. Sample includes unbalanced panel of 33 advanced economies and 92 EMDEs for 1983-2020. p-statistics are shown in parentheses. 50 TABLE 7 Regression results for labor force participation rates, baseline 15-19 years old 20-24 years old 25-49 years old 50-64 years old 65+ years old Female Male Female Male Female Male Female Male Female Male Fertility 0.734*** 0.057* 0.000 (0.000) (0.000) (0.945) Secondary enrollment 0.197*** 0.127*** (0.000) (0.000) Tertiary enrollment -0.114*** -0.180*** (0.000) (0.000) Completion of tertiary 0.039 -0.023 0.235*** 0.130*** 0.406*** 0.063 education (0.249) (0.394) (0.000) (0.000) (0.000) (0.221) Completion of tertiary 0.158** -0.099* 0.323*** 0.313*** 0.486** 0.426** education (0.002) (0.045) (0.000) (0.000) (0.003) (0.002) Life expectancy 0.569*** -2.679** 0.101*** 0.227*** (0.000) (0.003) (0.000) (0.000) Cycle 16.14*** 21.43*** 1.04 11.54*** 1.504 -0.591** 0.590 -2.329** 1.435 21.76 (0.000) (0.000) (0.144) (0.000) (0.182) (0.008) (0.495) (0.008) (0.394) (0.399) Cycle * life expectancy -0.031 -0.192 (0.216) (0.584) Fertility * EMDE -0.669*** -0.066** (0.000) (0.006) Secondary enrollment * -0.337*** EMDE (0.000) Completion of secondary -0.027 -0.038 education * EMDE (0.495) (0.238) Completion of tertiary -0.127 0.153* education * EMDE (0.056) (0.000) Life expectancy * EMDE -0.143*** -0.608*** (0.000) (0.000) Secondary enrollment * -0.337*** EMDE (0.000) Completion of secondary -0.027 -0.038 education * EMDE (0.495) (0.238) 51 TABLE 7 Regression results for labor force participation rates, baseline (continued) 15-19 years old 20-24 years old 25-49 years old 50-64 years old 65+ years old Female Male Female Male Female Male Female Male Female Male Completion of tertiary -0.127 0.153* education * EMDE (0.056) (0.000) Life expectancy * EMDE -0.143*** -0.608*** (0.000) (0.000) Cycle * EMDE -17.90*** -24.21*** -11.72*** -1.456* 16.46 (0.000) (0.000) (0.000) (0.038) (0.526) Cycle * life expectancy * 0.039 EMDE (0.912) Coefficient of fertility in 0.065*** -0.009 EMDEs (0.000) (0.234) Coefficient of secondary -0.133*** enrollment in EMDEs (0.000) Coefficient of secondary -0.012 -0.058*** education in EMDEs (0.570) (0.000) Coefficient of tertiary 0.031 -0.063 education in EMDEs (0.478) (0.189) Coefficient of cycle in -0.145** -2.78** -0.18 0.048** EMDE (0.008) (0.001) (0.801) (0.009) Country fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Cohort fixed effects No No No No No No Yes Yes Yes Yes County-cohort fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Age fixed effects No No No No Yes Yes Yes Yes Yes Yes Number of observations 4432 4484 3741 3789 21382 21654 12239 12261 5111 5111 Number of countries 163 165 151 154 158 160 145 145 168 168 Adjusted R-square 0.997 0.997 0.999 0.999 0.997 0.999 0.986 0.993 0.998 0.999 Source: Barro and Lee 2013; Key Indicators of the Labor Market (KILM), International Labour Organization; Labour Force Statistics, Organisation for Economic Co-operation and Development (OECD); UN Population Prospects; World Development Indicators, World Bank; and World Bank staff estimations. Note: Business cycles defined as deviation of real GDP from Hodrick-Prescott-filtered trend. Sample includes unbalanced panel of 35 advanced economies and 133 EMDEs for 1987-2020. p-statistics are shown in parentheses. 52 TABLE 8 Regression results for labor force participation rates, robustness test: 10-year moving average 15-19 years old 20-24 years old 25-49 years old 50-64 years old 65+ years old Female Male Female Male Female Male Female Male Female Male Fertility 0.706*** 0.076** 0.004** (0.000) (0.009) (0.002) Secondary enrollment 0.202*** 0.149*** (0.000) (0.000) Tertiary enrollment -0.112*** -0.171*** (0.000) (0.000) Completion of secondary 0.022 0.030 0.252*** 0.149*** 0.341*** -0.014 education (0.540) (0.296) (0.000) (0.000) (0.000) (0.786) Completion of tertiary 0.167** -0.070 0.354*** 0.335*** 0.570*** 0.145 education (0.002) (0.166) (0.000) (0.000) (0.000) (0.265) Life expectancy 0.621*** 1.127*** 0.101*** 0.227*** (0.000) (0.000) (0.000) (0.000) Cycle 26.37*** 34.46*** 5.54*** 19.59*** 0.336 0.663 -2.63* -0.789 1.74 51.62 (0.000) (0.000) (0.000) (0.000) (0.832) (0.077) (0.042) (0.566) (0.826) (0.127) Cycle * life expectancy -0.023 -0.594 (0.574) (0.193) Fertility * EMDE -0.664*** -0.067** (0.000) (0.005) Secondary enrollment * -0.332*** EMDE (0.000) Completion of secondary -0.023 -0.057 education * EMDE (0.565) (0.080) Completion of tertiary -0.127 0.153* education * EMDE (0.056) (0.000) Life expectancy * EMDE -0.143*** -0.608*** (0.000) (0.000) 53 TABLE 8 Regression results for labor force participation rates, robustness test: 10-year moving average (continued) 15-19 years old 20-24 years old 25-49 years old 50-64 years old 65+ years old Female Male Female Male Female Male Female Male Female Male Cycle * EMDE -17.83*** -23.82*** -11.46*** -2.51* -17.04 (0.000) (0.000) (0.000) (0.033) (0.526) Cycle * life expectancy * 0.057 EMDE (0.876) Coefficient of fertility in 0.070*** -0.008 EMDEs (0.000) (0.251) Coefficient of secondary -0.133*** enrollment in EMDEs (0.000) Coefficient of secondary -0.015 -0.046*** education in EMDEs (0.470) (0.000) Coefficient of tertiary -0.035 0.047 education in EMDEs (0.450) (0.322) Coefficient of cycle in -1.69* -2.09* 0.220 -1.00** EMDE (0.033) (0.039) (0.745) (0.006) Country fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Cohort fixed effects No No No No No No Yes Yes Yes Yes County-cohort fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Age fixed effects No No No No Yes Yes Yes Yes Yes Yes Source: Barro and Lee 2013; Key Indicators of the Labor Market (KILM), International Labour Organization; Labour Force Statistics, Organisation for Economic Co-operation and Development (OECD); UN Population Prospects; World Development Indicators, World Bank; and World Bank staff estimations. Note: Sample of countries is balanced across gender and age specific regressions. Business cycles defined as deviation of real GDP from Hodrick-Prescott-filtered trend. Sample includes balanced panel of 34 advanced economies and 104 EMDEs for 1987-2020. p- statistics are shown in parentheses. 54 TABLE 9 Regression results of labor force participation rates, robustness check: linear-quadratic trend 15-19 years old 20-24 years old 25-49 years old 50-64 years old 65+ years old Female Male Female Male Female Male Female Male Female Male Fertility 0.697*** 0.059* 0.000 (0.000) (0.011) (0.922) Secondary enrollment 0.202*** 0.125*** (0.000) (0.000) Tertiary enrollment -0.113*** -0.180*** (0.000) (0.000) Completion of secondary 0.040 -0.013 0.236*** 0.1340*** 0.403*** 0.064 education (0.233) (0.642) (0.000) (0.000) (0.000) (0.218) Completion of tertiary 0.158** -0.100* 0.321*** 0.311*** 0.490** 0.431** education (0.002) (0.041) (0.000) (0.000) (0.003) (0.001) Life expectancy 0.571*** 0.972*** 0.101*** 0.229*** (0.000) (0.000) (0.000) (0.000) Cycle 15.11*** 24.22*** 0.281 12.72*** 3.24** 0.156 -1.56 -2.12* 1.45 17.01 (0.000) (0.000) (0.684) (0.000) (0.003) (0.470) (0.101) (0.014) (0.512) (0.491) Cycle * life expectancy -0.027 -0.118 (0.275) (0.348) Fertility * EMDE -0.630*** -0.067** (0.000) (0.005) Secondary enrollment * -0.342*** EMDE (0.000) Completion of secondary -0.029 -0.048 education * EMDE (0.482) (0.133) Completion of tertiary -0.126 0.155* education * EMDE (0.058) (0.014) Life expectancy * EMDE -0.145*** -0.620*** (0.000) (0.000) 55 TABLE 9 Regression results of labor force participation rates, robustness check: linear-quadratic trend (continued) 15-19 years old 20-24 years old 25-49 years old 50-64 years old 65+ years old Female Male Female Male Female Male Female Male Female Male Cycle * EMDE -16.77*** -25.50*** -12.11*** -3.91** -16.58 (0.000) (0.000) (0.000) (0.001) (0.504) Cycle * life expectancy * 0.073 EMDE (0.829) Coefficient of fertility in 0.067*** -0.008 EMDEs (0.000) (0.285) Coefficient of secondary -0.138*** enrollment in EMDEs (0.000) Coefficient of secondary 0.011 -0.164*** education in EMDEs (0.556) (0.000) Coefficient of tertiary 0.032 -0.083 education in EMDEs (0.472) (0.253) Coefficient of cycle in -1.66** -1.28 0.35 -0.667* EMDE (0.007) (0.103) (0.740) (0.063) Country fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Cohort fixed effects No No No No No No Yes Yes Yes Yes County-cohort fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes effects Age fixed effects No No No No Yes Yes Yes Yes Yes Yes Number of observations 4428 4480 3741 3789 21382 21654 12239 12261 5107 5107 Number of countries 163 165 151 154 158 160 145 145 168 168 Adjusted R-square 0.997 0.997 0.999 0.999 0.997 0.999 0.986 0.993 0.998 0.999 Source: Barro and Lee 2013; Key Indicators of the Labor Market (KILM), International Labour Organization; Labour Force Statistics, Organisation for Economic Co-operation and Development (OECD); UN Population Prospects; World Development Indicators, World Bank; and World Bank staff estimations. Note: Business cycles defined as deviation of real GDP from linear-quadratic trend. Sample includes unbalanced panel of 35 advanced economies and 133 EMDEs for 1987-2020. p-statistics are shown in parentheses. 56 TABLE 10 Coverage for univariate and multivariate filter-based estimates Sample Sample Sample Economy Economy Economy period period period Australia 1981-2024 East Asia and Pacific Paraguay 1994-2024 Austria 1995-2024 China 1992-2024 Peru 1998-2024 Belgium 1995-2024 Indonesia 2001-2024 Uruguay 1997-2024 Canada 1981-2024 Malaysia 2005-2024 Middle East and North Africa Croatia 2000-2024 Mongolia 2010-2024 Bahrain 2008-2024 Cyprus 1995-2024 Philippines 1998-2024 Egypt, Arab Rep. 2007-2024 Czech Rep. 1996-2024 Thailand 1993-2024 Iran, Islamic Rep. 2012-2024 Denmark 1991-2024 Vietnam 2008-2024 Jordan 1992-2024 Estonia 1995-2024 Europe and Central Asia Kuwait 2010-2024 Finland 1981-2024 Albania 2008-2024 Morocco 1998-2024 France 1981-2024 Azerbaijan 2001-2024 Saudi Arabia 2010-2024 Germany 1981-2024 Bulgaria 2000-2024 Tunisia 2000-2024 Greece 1995-2024 Georgia 2003-2024 South Asia Hong Kong SAR, 1990-2024 Hungary 1998-2024 India 1997-2024 China Iceland 1995-2024 Kazakhstan 1996-2024 Sub-Saharan Africa Ireland 1995-2024 North Macedonia 2000-2024 Botswana 1994-2024 Israel 1995-2024 Poland 1996-2024 Cameroon 1999-2024 Italy 1981-2024 Romania 1995-2024 Kenya 2009-2024 Japan 1981-2024 Turkey 2001-2024 Lesotho 2007-2024 Korea 1981-2024 Latin America and Caribbean Namibia 2000-2024 Latvia 1995-2024 Argentina 2004-2024 Nigeria 2010-2024 Lithuania 1995-2024 Belize 1994-2024 South Africa 1981-2024 Luxembourg 1995-2024 Bolivia 1990-2024 Tanzania 2010-2024 Malta 2000-2024 Brazil 1990-2024 Netherlands 1981-2024 Chile 1996-2024 New Zealand 1988-2024 Colombia 2000-2024 Norway 1981-2024 Costa Rica 1991-2024 Portugal 1995-2024 Dominican Republic 2007-2024 Singapore 1981-2024 Ecuador 2001-2024 Slovak Republic 1995-2024 El Salvador 1990-2024 Slovenia 1995-2024 Guatemala 2001-2024 Spain 1995-2024 Honduras 2000-2024 Sweden 1981-2024 Mexico 2000-2024 Switzerland 1981-2024 Nicaragua 2006-2024 Taiwan 1982-2024 Panama 2007-2024 United Kingdom 1981-2024 United States 1981-2024 Source: World Bank. Note: Forecasts for 2022Q2-2024Q4 are based on the lag structure of the estimation. 57 TABLE 11 Coverage for production function approach, filter-based, and expectations-based estimates: Advanced economies Production Univariate and Economy WEO expectations function approach multivariate filters Advanced economies Australia 1998-2032 1981-2024 1990-2022 Austria 1998-2032 1995-2024 1990-2022 Belgium 1998-2032 1995-2024 1990-2022 Canada 1998-2032 1981-2024 1990-2022 Croatia 1998-2032 2000-2024 1994-2022 Cyprus 1998-2032 1995-2024 1990-2022 Denmark 1998-2032 1991-2024 1990-2022 Estonia 1998-2032 1995-2024 1993-2022 Finland 1998-2032 1981-2024 1990-2022 France 1998-2032 1981-2024 1990-2022 Germany 1998-2032 1981-2024 1990-2022 Greece 1998-2032 1995-2024 1990-2022 Hong Kong SAR, China 1998-2032 1990-2024 1990-2022 Ireland 1998-2032 1995-2024 1990-2022 Israel 1998-2032 1995-2024 1990-2022 Italy 1998-2032 1981-2024 1990-2022 Japan 1998-2032 1981-2024 1990-2022 Korea, Rep. 1998-2032 1981-2024 1990-2022 Latvia 1998-2032 1995-2024 1993-2022 Lithuania 2000-2032 1995-2024 1993-2022 Netherlands 1998-2032 1981-2024 1990-2022 Norway 1998-2032 1981-2024 1990-2022 Portugal 1998-2032 1995-2024 1990-2022 Slovak Republic 1998-2032 1995-2024 1994-2022 Slovenia 1998-2032 1995-2024 1994-2022 Spain 1998-2032 1995-2024 1990-2022 Sweden 1998-2032 1981-2024 1990-2022 Switzerland 1998-2032 1981-2024 1990-2022 United Kingdom 1998-2032 1981-2024 1990-2022 United States 1998-2032 1981-2024 1990-2022 Source: World Bank. Note: Forecasts for filter-based estimates for 2022Q2-2024Q4 are based on the lag structure of the estimation. Forecasts for production function-based estimates are derived as described in Kilic Celik, Kose, and Ohnsorge (2023). Univariate filters: Hodrick-Prescott, Baxter and King, Christiano and Fitzgerald, Butterworth, and unobserved component model. 58 TABLE 12 Coverage for production function approach, filter-based, and expectations-based estimates: EMDEs Production Univariate and Economy WEO expectations function approach multivariate filters EMDEs Albania 1998-2032 2008-2024 1993-2021 Argentina 1998-2032 2004-2024 1990-2021 Bolivia 1998-2032 1990-2024 1990-2021 Brazil 1998-2032 1990-2024 1990-2021 Bulgaria 2000-2032 2000-2024 2000-2021 Cameroon 1998-2032 1999-2024 1990-2021 Chile 1998-2032 1996-2024 1990-2021 China 1998-2032 1992-2024 1990-2021 Colombia 1998-2032 2000-2024 1990-2021 Costa Rica 1998-2032 1991-2024 1990-2021 Dominican Republic 1998-2032 2007-2024 1990-2021 Ecuador 1998-2032 2001-2024 1990-2021 Egypt, Arab Rep. 1998-2032 2007-2024 1990-2021 Guatemala 1998-2032 2001-2024 1990-2021 Honduras 1998-2032 2000-2024 1990-2021 Hungary 1998-2032 1998-2024 1990-2021 India 1998-2032 1997-2024 1990-2021 Indonesia 1998-2032 2001-2024 1990-2021 Iran, Islamic Rep. 1998-2032 2012-2024 1990-2021 Jordan 1998-2032 1992-2024 1990-2021 Kazakhstan 1998-2032 1996-2024 1993-2021 Kenya 1998-2032 2009-2024 1990-2021 Lesotho 1998-2032 2007-2024 1990-2021 Malaysia 1998-2032 2005-2024 1990-2021 Mexico 1998-2032 2000-2024 1990-2021 Mongolia 1998-2032 2010-2024 1993-2021 Morocco 1998-2032 1998-2024 1990-2021 Namibia 1998-2032 2000-2024 1994-2021 Nicaragua 1998-2032 2006-2024 1990-2021 Paraguay 1998-2032 1994-2024 1990-2021 Peru 1998-2032 1998-2024 1990-2021 Philippines 1998-2032 1998-2024 1990-2021 Poland 1998-2032 1996-2024 1990-2021 Romania 1998-2032 1995-2024 1993-2021 South Africa 1998-2032 1981-2024 1990-2021 Thailand 1998-2032 1993-2024 1990-2021 Tunisia 1998-2032 2000-2024 1990-2021 Turkey 1998-2032 2001-2024 1990-2021 Uruguay 1998-2032 1997-2024 1990-2021 Vietnam 2013-2032 2008-2024 1990-2021 Source: World Bank. Note: Includes only countries with available data from 2001. Forecasts for filter- based estimates for 2022Q2-2024Q4 are based on the lag structure of the estimation. Forecasts for production function-based estimates are derived as described in Kilic Celik, Kose, and Ohnsorge (2023). Univariate filters: Hodrick- Prescott, Baxter and King, Christiano and Fitzgerald, Butterworth, and unobserved component model. 59 TABLE 13 List of banking crises Regions Countries Advanced AUT (2008), BEL (2008), CHE (2008), CYP (2011), CZE (1996), DEU (2008), DNK economies (2008), ESP (2008), FIN (1991), FRA (2008), GBR (2007), GRC (2008), HRV (1998), IRL (2008), ISL (2008), ITA (2008), JPN (1997), KOR (1997), LTU (1995), LUX (2008), LVA (1995), LVA (2008), NLD (2008), NOR (1991), PRT (2008), SVK (1998), SVN (2008), SWE (1991), SWE (2008), USA (2007) Emerging market ALB (1994), ARG (1995), ARG (2001), ARM (1994), AZE (1995), BDI (1994), BFA and developing (1990), BOL (1994), BRA (1990), BRA (1994), CAF (1995), CHN (1998), CMR (1995), economies COD (1991), COD (1994), COG (1992), COL (1998), CPV (1993), CRI (1994), DJI (1991), DOM (2003), DZA (1990), ECU (1998), GIN (1993), GNB (1995), GNB (2014), GUY (1993), HTI (1994), HUN (1991), HUN (2008), IDN (1997), IND (1993), JAM (1996), KAZ (2008), KEN (1992), KGZ (1995), LBN (1990), LBR (1991), MDA (2014), MEX (1994), MNG (2008), MYS (1997), NGA (1991), NGA (2009), NIC (1990), NIC (2000), PHL (1997), POL (1992), PRY (1995), ROU (1998), STP (1992), TCD (1992), TGO (1993), THA (1997), TUN (1991), TUR (2000), UGA (1994), URY (2002), VNM (1997), YEM (1996) Sources: Laeven and Valencia 2018; World Bank. Note: The list of banking crises corresponding to the sample of potential growth measures. Country codes are available at https://www.iban.com/country-codes. TABLE 14 List of countries affected by epidemics Epidemics Countries SARS (2003) CAN, CHN, FRA, MYS, PHL, SGP, THA, VNM, ZAF, HKG, TWN. Swine flu (2009) AFG, ALB, ARE, ARG, ARM, AUS, AZE, BGD, BGR, BHR, BHS, BIH, BLR, BMU, BOL, BRA, BRB, BRN, CAN, CHE, CHL, CHN, COL, CRI, CUB, CZE, DEU, DOM, DZA, ECU, EGY, ESP, EST, FRA, GBR, GEO, GHA, GRC, GTM, HND, HRV, HUN, IDN, IND, IRL, IRN, IRQ, ISL, ISR, ITA, JAM, JOR, JPN, KHM, KOR, KWT, LAO, LBN, LBY, LKA, LTU, LUX, LVA, MAR, MDA, MDG, MDV, MEX, MHL, MLT, MNE, MNG, MOZ, MUS, MYS, NAM, NGA, NIC, NLD, NOR, NPL, NZL, OMN, PAK, PAN, PER, PHL, POL, PRY, PYF, QAT, ROU, RUS, SAU, SDN, SGP, SLB, SLV, SRB, SUR, SVK, SVN, SWE, SYR, THA, TON, TUN, TUR, TZA, UKR, URY, USA, VNM, WSM, YEM, ZAF. MERS (2012) ARE, AUT, DEU, DZA, FRA, GBR, GRC, IRN, JOR, KOR, KWT, MYS, OMN, QAT, SAU, TUN, TUR, YEM. Ebola (2014) MLI, NGA, GIN, LBR, SLE. Zika (2016) BOL, BRA, COL, DOM, GLP, MTQ, PRI, SUR, USA. Source: World Bank. Note: Country codes are available at https://www.iban.com/country-codes. 60 TABLE 15 Impulse responses of potential growth to recessions Recessions: Baseline definition Recessions: Alternative definition Definition of potential h World AEs EMDEs World AEs EMDEs output 0 -0.042 0.066 -0.138 -0.046 0.042 -0.123 1 -1.153*** -0.773*** -1.499*** -1.123*** -0.792*** -1.414*** Production-function 2 -1.573*** -1.407*** -1.738*** -1.432*** -1.402*** -1.454*** approach 3 -1.542*** -1.444*** -1.645*** -1.401*** -1.432*** -1.371*** 4 -1.521*** -1.421*** -1.639*** -1.348*** -1.386*** -1.308*** 5 -1.431*** -1.257*** -1.635*** -1.244*** -1.193*** -1.296*** 0 -0.355*** -0.354*** -0.352*** -0.348*** -0.342*** -0.352*** 1 -2.082*** -1.782*** -2.465*** -2.014*** -1.709*** -2.419*** 2 -1.298*** -1.485*** -0.947*** -1.215*** -1.372*** -0.91*** Multivariate filter 3 -0.734*** -1.033*** -0.192 -0.647*** -0.848*** -0.272 4 -0.442* -0.699** 0.06 -0.356* -0.488** -0.103 5 -0.133 -0.215 0.025 -0.123 -0.143 -0.089 0 -0.058 -0.06 -0.057 -0.04 -0.037 -0.042 1 -0.208** 0.055 -0.356*** 0.08 0.128* 0.052 2 -0.33** -0.143 -0.425** -0.036 -0.042 -0.032 Expectations (WEO) 3 -0.315* -0.144 -0.403 -0.282 -0.08 -0.395 4 -0.251 -0.072 -0.348 -0.282** -0.022 -0.433** 5 -0.262* -0.125 -0.336 -0.269** -0.078 -0.378* 0 -0.208*** -0.215*** -0.2*** -0.215*** -0.238*** -0.184*** 1 -1.83*** -1.605*** -2.102*** -1.794*** -1.597*** -2.037*** Unobserved component 2 -0.638*** -0.711*** -0.532*** -0.599*** -0.67*** -0.497*** model 3 -0.279*** -0.256** -0.316* -0.275*** -0.217** -0.362** 4 -0.3*** -0.298** -0.301** -0.297*** -0.262** -0.358*** 5 -0.198* -0.143 -0.288*** -0.19** -0.118 -0.314*** Source: World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. “Recessions: Baseline definition” are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years of output growth that is both negative and one standard deviation below the long-term average (as in Huidrom, Kose, and Ohnsorge 2016). “Mild Recessions: Alternative definition” are defined as years of negative output growth only, regardless of the depth of the output decline. Sample includes unbalanced panel of 33 advanced economies and 77 EMDEs for 1981-2020. 61 TABLE 16 Impulse Responses of Potential Growth to Recessions (Other Measures) Recessions: Recessions: Baseline definition Alternative definition Definition of potential h World AEs EMDEs World AEs EMDEs output 0 0.004 0.04 -0.056 0.012 0.041 -0.04 1 -0.084 -0.024 -0.189** -0.087* -0.058 -0.139* 2 -0.157** -0.127 -0.207* -0.135** -0.13 -0.145 Expectations (CF) 3 -0.114 -0.07 -0.171 -0.077 -0.083 -0.067 4 -0.215** -0.134* -0.361 -0.241*** -0.224*** -0.272 5 -0.19** -0.187* -0.203 -0.214** -0.26** -0.124 0 -0.165*** -0.194*** -0.128*** -0.16*** -0.181*** -0.132*** 1 -0.212*** -0.337*** -0.046 -0.2*** -0.298*** -0.066 2 -0.493*** -0.664*** -0.224 -0.412*** -0.512** -0.264 Hodrick-Prescott filter 3 -0.32 -0.544* 0.056 -0.232 -0.35 -0.053 4 -0.146 -0.321 0.17 -0.072 -0.132 0.006 5 0.058 -0.047 0.249 0.089 0.089 0.055 0 -0.691*** -0.575*** -0.8*** -0.673*** -0.524*** -0.826*** 1 -0.809*** -0.937*** -0.61*** -0.798*** -0.867*** -0.67*** Christiano-Fitzgerald 2 -1.299*** -1.572*** -0.795** -1.193*** -1.304*** -0.956** filter 3 -1.233*** -1.563*** -0.608 -1.061*** -1.215*** -0.749* 4 -1.029*** -1.419*** -0.257 -0.887*** -1.062*** -0.548 5 -0.685** -0.833* -0.406 -0.598** -0.579 -0.666 0 -2.161*** -1.983*** -2.388*** -2.113*** -1.932*** -2.351*** 1 -4.197*** -4.099*** -4.327*** -4.08*** -3.983*** -4.216*** 2 -3.413*** -3.607*** -3.071*** -3.132*** -3.295*** -2.843*** Baxter-King filter 3 -1.589*** -1.799*** -1.2** -1.42*** -1.512*** -1.254** 4 -1.469*** -1.614*** -1.166** -1.303*** -1.281*** -1.353*** 5 -1.333*** -1.298*** -1.396*** -1.167*** -1.047*** -1.417*** 0 -0.703*** -0.562*** -0.744*** -0.693*** -0.544*** -0.726*** 1 -1.507*** -1.27*** -1.672*** -1.461*** -1.212*** -1.626*** 2 -1.419*** -1.493*** -1.078*** -1.29*** -1.307*** -1.01*** Butterworth filter 3 -1.103*** -1.017*** -1.05** -0.979*** -0.813*** -1.044*** 4 -0.792*** -0.75** -0.784* -0.679*** -0.554** -0.834** 5 -0.443** -0.433 -0.425 -0.378** -0.293 -0.51* Source: World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. “Recessions: Baseline definition” are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years of output growth that is both negative and one standard deviation below the long-term average (as in Huidrom, Kose, and Ohnsorge 2016). “Mild Recessions: Alternative definition” are defined as years of negative output growth only, regardless of the depth of the output decline. Sample includes unbalanced panel of 33 advanced economies and 77 EMDEs for 1981-2020. 62 TABLE 17 Impulse Responses of Potential Growth to Banking Crises and Epidemics Banking crises Epidemics Definition of potential h World AEs EMDEs World AEs EMDEs output 0 -0.574*** -0.538* -0.763** -0.731*** -0.846*** -0.68*** 1 -1.605*** -1.508** -1.865*** -0.796*** -1.035*** -0.649*** Production-function 2 -1.75*** -1.979*** -1.402*** -0.77*** -0.911*** -0.655*** approach 3 -1.467*** -1.958*** -0.451 -0.872*** -1.057*** -0.77** 4 -1.286*** -1.929*** 0.031 -1.083*** -1.126*** -1.062*** 5 -1.169** -1.908*** 0.416 -0.866*** -0.849** -0.895*** 0 -0.349** -0.406** -0.209 -0.229** -0.247 -0.214 1 -0.746*** -0.981*** -0.119 -0.021 -0.198 0.12 2 -0.724** -1.25*** 0.743 0.195 0.169 0.215 Multivariate filter 3 -0.27 -0.81** 1.176** 0.305 0.531* 0.127 4 0.127 -0.279 1.183* 0.232 0.63** -0.081 5 0.4 0.052 1.339* 0.335 0.874** -0.121 0 -0.025 -0.044 -0.019 -0.421*** -0.173 -0.525*** 1 -0.08 0.065 -0.155 -0.334*** -0.287*** -0.358** 2 0.028 -0.035 0.076 -0.313* -0.176 -0.374 Expectations (WEO) 3 0.276 0.088 0.394 -0.479*** -0.175 -0.609*** 4 0.174 0.141 0.199 -0.519*** -0.19 -0.661*** 5 0.142 0.071 0.199 -0.623*** -0.208 -0.808*** 0 -0.573*** -0.736*** -0.278 -0.664*** -0.792*** -0.564*** 1 -1.399*** -1.731*** -0.806** 0.139* 0.133 0.146 Unobserved component 2 -0.364** -0.67*** 0.18 0.075 0.083 0.066 model 3 -0.133 -0.48*** 0.488*** -0.075 -0.059 -0.085 4 -0.356** -0.796*** 0.43** -0.198 -0.028 -0.335* 5 -0.299** -0.553*** 0.152 0.005 0.191 -0.156 Sources: Laeven and Valencia 2018; World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. Sample includes unbalanced panel of 33 advanced economies and 98 EMDEs for 1981-2020. 63 TABLE 18 Responses of Potential Growth to Banking Crises and Epidemics (Other Measures) Banking crises Epidemics Definition of potential h World AEs EMDEs World AEs EMDEs output 0 0.046 0.093** -0.046 -0.081 -0.105 -0.062 1 -0.33** -0.144 -0.753*** -0.005 -0.148* 0.179 2 -0.192 -0.163 -0.266 0.077 -0.082 0.275** Expectations (CF) 3 -0.094 0.186 -0.632*** -0.056 -0.142** 0.027 4 -0.212* -0.102 -0.4 0.003 -0.063 0.082 5 -0.285* -0.161 -0.5 -0.104 -0.141 -0.039 0 -0.132** -0.229*** 0.113 0.065** 0.163*** -0.01 1 -0.177 -0.431*** 0.456 0.297*** 0.546*** 0.104 2 0.002 -0.39 0.979 0.499*** 0.878*** 0.199 Hodrick-Prescott filter 3 0.258 -0.224 1.453* 0.554*** 1.037*** 0.17 4 0.497 -0.006 1.747* 0.509** 1.097*** 0.042 5 0.761* 0.299 1.913* 0.456* 1.146*** -0.124 0 -0.485*** -0.53*** -0.253 -0.451*** -0.444*** -0.421*** 1 -1.034*** -1.365*** -0.005 -0.396*** -0.21 -0.513** Christiano-Fitzgerald 2 -1.096*** -1.612*** 0.338 0.032 0.284 -0.151 filter 3 -0.757 -1.481*** 1.181 0.364 0.673** 0.12 4 -0.344 -1.083** 1.512 0.214 0.57* -0.086 5 0.166 -0.501 1.825 0.604** 1.091*** 0.174 0 -2.288*** -2.64*** -1.31* -0.666*** -0.739** -0.614*** 1 -3.877*** -4.73*** -1.525 0.415 0.492 0.341 2 -2.149*** -2.975*** 0.125 0.677** 0.833** 0.539 Baxter-King filter 3 -0.921 -1.768*** 1.427 0.173 0.428 -0.031 4 -1.198** -1.993*** 1.001 0.02 0.407 -0.284 5 -0.875* -1.59*** 1.114 0.249 0.88* -0.269 0 -0.899*** -0.739*** -0.597 -0.45 0.03 -0.553* 1 -1.382*** -1.429*** -0.515 0.196 0.665*** 0.116 2 -0.892** -1.085*** 0.249 0.295 0.876*** 0.095 Butterworth filter 3 -0.476 -0.745** 0.782 0.117 0.803*** -0.204 4 -0.212 -0.619* 1.073 0.214 0.809*** -0.164 5 0.117 -0.278 1.262 0.212 0.922** -0.318 Sources: Laeven and Valencia 2018; World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. Sample includes unbalanced panel of 33 advanced economies and 98 EMDEs for 1981-2020. 64 TABLE 19 Channels: Impulse Responses of TFP, Investment, Employment and Actual Growth Rates to Contractions Recessions: Recessions: Alternative Baseline definition definition Definition of potential h World AEs EMDEs World AEs EMDEs output 0 -0.066** -0.019 -0.108** -0.064** -0.041** -0.087* 1 -0.359*** -0.228*** -0.471*** -0.353*** -0.251*** -0.443*** 2 -0.626*** -0.476*** -0.743*** -0.577*** -0.495*** -0.64*** Expectations (CF) 3 -0.676*** -0.495*** -0.819*** -0.635*** -0.527*** -0.723*** 4 -0.759*** -0.497*** -0.985*** -0.69*** -0.519*** -0.842*** 5 -0.686*** -0.418*** -0.919*** -0.619*** -0.425*** -0.793*** 0 -1.842** -2.913*** -1.151 -2.469*** -3.515*** -1.7* 1 -15.501*** -12.809*** -17.097*** -15.483*** -12.99*** -17.006*** 2 -7.689*** -10.231*** -6.265** -7.37*** -9.332*** -6.151** Hodrick-Prescott filter 3 -3.348** -4.079** -2.936 -2.963* -3.696*** -2.484 4 -2.947* -2.897 -2.976 -1.814 -2.478* -1.414 5 -3.017** -2.838* -3.13 -3.601*** -2.588** -4.216** 0 -0.432*** -0.309 -0.497** -0.446*** -0.435*** -0.444** 1 -1.691*** -2.898*** -1.247*** -1.723*** -2.845*** -1.248*** Christiano-Fitzgerald 2 -1.29*** -3.4*** -0.471 -1.331*** -3.13*** -0.549* filter 3 -1.038*** -1.592*** -0.819** -1.025*** -1.509*** -0.817** 4 -0.717*** -1.046*** -0.586* -0.631*** -0.964*** -0.482 5 -0.398 -0.975*** -0.16 -0.393 -0.86*** -0.179 0 -0.039 -0.077 -0.017 -0.048 -0.055 -0.044 1 1.326*** 1.555*** 1.21*** 1.281*** 1.588*** 1.126*** 2 1.88*** 3.424*** 1.15*** 1.78*** 3.417*** 1.048*** Baxter-King filter 3 1.786*** 3.457*** 1.002*** 1.698*** 3.515*** 0.897*** 4 1.689*** 3.257*** 0.902*** 1.577*** 3.234*** 0.803** 5 1.656*** 3.34*** 0.811** 1.464*** 3.112*** 0.695** 0 0.019 -0.887*** 0.419 -0.02 -0.986*** 0.446 1 -8.809*** -7.157*** -9.597*** -8.474*** -6.843*** -9.305*** 2 -4.992*** -4.506*** -5.197*** -4.649*** -3.94*** -4.979*** Butterworth filter 3 -1.399** -2.503** -0.957 -1.337** -2.112** -0.988 4 -2.349*** -2.539*** -2.28** -2.095*** -2.012*** -2.144** 5 -1.124** -1.609** -0.903 -0.886* -1.209** -0.719 Source: World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. “Recessions: Baseline definition” are defined as the period from the peak preceding a business cycle trough to the trough, with the troughs defined as years of output growth that is both negative and one standard deviation below the long-term average (as in Huidrom, Kose, and Ohnsorge 2016). “Mild Recessions: Alternative definition” are defined as years of negative output growth only, regardless of the depth of the output decline. Sample includes unbalanced panel of 32 advanced economies and 79 EMDEs for 1981-2020. 65 TABLE 20 Channels: Impulse Responses of TFP, Investment, Employment and Actual Growth Rates to Banking Crises and Epidemics Banking crises Epidemics Definition of potential h World AEs EMDEs World AEs EMDEs output 0 -0.177*** -0.119*** -0.279** -0.235*** -0.241*** -0.223*** 1 -0.559*** -0.419*** -0.771*** -0.276*** -0.307*** -0.248*** Total factor 2 -0.627*** -0.566*** -0.748*** -0.296*** -0.306*** -0.278*** productivity 3 -0.562*** -0.619*** -0.531** -0.394*** -0.389*** -0.388*** 4 -0.54*** -0.655*** -0.446 -0.524*** -0.358*** -0.606*** 5 -0.375** -0.558*** -0.189 -0.315*** -0.093 -0.434*** 0 -4.451* -4.119 -4.576 -12.522*** -9.658*** -13.252*** 1 -14.031*** -16.744*** -12.31*** -3.487** -1.575 -4.275** 2 -1.649 -11.541*** 4.509 -2.762* 2.696** -4.803** Investment 3 3.182 -2.718 6.846* -3.202*** 0.203 -4.575*** 4 0.507 -6.409*** 4.781* -3.442** -0.446 -4.772** 5 -2.145 -6.08*** 0.303 -4.085*** 1.671 -6.537*** 0 -0.223 -0.677* -0.03 -1.662*** -2.784*** -1.167*** 1 -1.196*** -3.444*** -0.358 -0.951*** -1.419*** -0.764* 2 -0.501 -2.528*** 0.243 -0.866*** -0.584** -1.009** Employment 3 -0.166 -1.511*** 0.339 -0.574* -0.897*** -0.44 4 -0.198 -1.551*** 0.316 -0.926*** -0.662* -1.021** 5 0.12 -1.403*** 0.692** -0.828*** -0.377 -1.039*** 0 0.382** 0.473** 0.355 0.869*** 1.881*** 0.465*** 1 1.592*** 2.81*** 0.909*** 1.063*** 2.516*** 0.497** 2 1.891*** 3.574*** 0.928*** 1.089*** 2.402*** 0.599** Unemployment 3 1.828*** 3.822*** 0.663** 1.151*** 2.701*** 0.592** 4 2.1*** 4.494*** 0.694** 1.316*** 2.841*** 0.742*** 5 2.156*** 4.684*** 0.661** 1.033*** 2.401*** 0.51* 0 -0.629 -2.113** 0.026 -3.956*** -4.161*** -3.76*** 1 -2.026 -5.123*** -0.64 -0.362 0.903 -0.871 2 0.967 -0.462 1.609 -0.128 0.491 -0.403 Actual growth 3 1.809** 0.055 2.596** -1.124*** -0.51 -1.379*** 4 1.859** -1.334 3.292*** -1.137*** -0.287 -1.491*** 5 1.66* -0.419 2.603** -1.081*** 0.183 -1.731*** Source: Laeven and Valencia 2018; and World Bank. Note: *** indicates significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level. Sample includes unbalanced panel of 32 advanced economies and 100 EMDEs for 1981-2020. 66 References Abiad, A. G., D. Leigh, and A. Mody. 2007. “International Finance and Income Convergence: Europe is Different.” IMF Working Paper 64, International Monetary Fund, Washington, DC. ADB (Asian Development Bank). 2016. Asian Development Outlook 2016: Asia’s Potential Growth. Manila: Asian Development Bank. Aghion, P., and M. Schankerman. 2004. “On the Welfare Effects and Political Economy of Competition-Enhancing Policies.” The Economic Journal 114 (498): 800-24. Ali, M., A. Egbetokun, and M. H. Memon. 2018. “Human Capital, Social Capabilities and Economic Growth.” Economies 6 (2): 1-18. Anand, R., M. K. C. Cheng, S. Rehman, and M. L. Zhang. 2014. “Potential Growth in Emerging Asia.” IMF Working Paper 14/2, International Monetary Fund, Washington, DC. Angrist, N., S. Djankov, P. K. Goldberg, and H. A. Patrinos. 2021. “Measuring Human Capital Using Global Learning Data.” Nature 592 (7854): 403-08. Anzoategui, D., D. Comin, M. Gertler, and J. Martinez. 2019. “Endogenous Technology Adoption and R&D as Sources of Business Cycle Persistence." American Economic Journal: Macroeconomics 11 (3): 67-110. Azevedo, J. P., L. F. Lopez-Calva, and E. Perova. 2012. “Is the Baby to Blame? An Inquiry into the Consequences of Early Childbearing.” Policy Research Working Paper 6074, World Bank, Washington, DC. Baker, S. R., N. Bloom, and S. J. Davis. 2016. “Measuring Economic Policy Uncertainty.” Quarterly Journal of Economics 131 (4): 1593-636. Balakrishnan, R., M. Dao, J. Solé, and J. Zook. 2015. “Recent U.S. Labor Force Dynamics: Reversible or Not?” IMF Working Paper 76, International Monetary Fund, Washington, DC. Ball, L. M. 2009. “Hysteresis in Unemployment: Old and New Evidence.” NBER Working Paper 14818, National Bureau of Economic Research, Cambridge, MA. Ball, L. M. 2014. “Long-Term Damage from the Great Recession in OECD Countries.” NBER Working Paper 20185, National Bureau of Economic Research, Cambridge, MA. Barro, R. J., and J. W. Lee. 2013. “A New Data Set of Educational Attainment in the World, 1950-2010.” Journal of Development Economics 104 (September): 184-98. Barro, R. J., and X. Sala-i-Martin. 1997. “Technological Diffusion, Convergence, and Growth.” Journal of Economic Growth 2 (1): 1-26. Basu, S., and J. G. Fernald. 2009. “What Do We Know (and Not Know) about Potential Output?” Federal Reserve Bank of St Louis Review 91 (4): 187-213. 67 Basu, K. 2015. Quoted in “Continuing Slowdown in Emerging Markets Heralds Lengthy Era of Weak Growth.” Press Release 2016/207/DEC, World Bank, Washington, DC. Baxter, M., and R. King. 1999. “Measuring Business Cycles: Approximate Band-Pass Filters for Economic Time Series.” The Review of Economics and Statistics 81 (4): 575- 93. Beneš, J., K. Clinton, R. Garcia-Saltos, M. Johnson, D. Laxton, P. Manchev, and T. Matheson. 2010. “Estimating Potential Output with a Multivariate Filter.” IMF Working Paper 10/285, International Monetary Fund, Washington, DC. Benhabib, J., and M. M. Spiegel. 1994. “The Role of Human Capital in Economic Development: Evidence from Aggregate Cross-Country Data.” Journal of Monetary Economics 34 (2): 143-73. Benhabib, J., and M. M. Spiegel. 2005. “Human Capital and Technology Diffusion.” Handbook of Economic Growth 1 (A): 935-66. Bhoi, B. K., and H. K. Behera. 2016. “India’s Potential Output Revisited.” RBI Working Paper 05, Reserve Bank of India, New Delhi. Bijsterbosch, M., and M. Kolasa. 2010. “FDI and Productivity Convergence in Central and Eastern Europe: An Industry-level Investigation.” Review of World Economics 145 (4): 689-712. Blagrave, P., R. Garcia-Saltos, D. Laxton, and F. Zhang. 2015. “A Simple Multivariate Filter for Estimating Potential Output.” IMF Working Paper 15/79, International Monetary Fund, Washington, DC. Blanchard, O. 1991. “Wage Bargaining and Unemployment Persistence.” NBER Working Paper 3664, National Bureau of Economic Research, Cambridge, MA. Blanchard, O. J., and L. H. Summers. 1987. “Hysteresis in Unemployment.” European Economic Review 31 (1-2): 288-95. Bloom, D. E., D. Canning, G. Fink, and J. E. Finlay. 2009. “Fertility, Female Labor Force Participation, and the Demographic Dividend.” Journal of Economic Growth: 79- 101. Bloom, N. 2009. “The Impact of Uncertainty Shocks.” Econometrica 77 (3): 623-85. Bloom, N. 2014. “Fluctuations in Uncertainty.” Journal of Economic Perspectives 28 (2): 153-76. Bloom, N., P. Bunn, P. Mizen, P. Smietanka, and G. Thwaites. 2020. “The Impact of Covid-19 on Productivity.” NBER Working Paper 28233, National Bureau of Economic Research, Cambridge, MA. Bodnár, K., J. Le Roux, P. Lopez-Garcia, and B. Szörfi. 2020. “The Impact of COVID- 19 on Potential Output in the Euro Area.”Economic Bulletin Articles7, European Central Bank, Frankfurt. 68 Bordo, M. D., and P. L. Siklos. 2017. “Central Bank Credibility Before and After the Crisis.” Open Economies Review 28 (1): 19-45. Butterworth, S. 1930. “On the Theory of Filter Amplifiers.” Experimental Wireless and the Wireless Engineer 7: 536-41. Caballero, R. J., and A. Simsek. 2017. “A Risk-Centric Model of Demand Recessions and Macroprudential Policy.” NBER Working Paper 23614, National Bureau of Economic Research, Cambridge, MA. Candelon, B., A. Carare, and K. Miao. 2016. “Revisiting the New Normal Hypothesis.” Journal of International Money and Finance 66 (September): 5-31. Cantelmo, A., G. Melina, and C. Papageorgiou. 2019. “Macroeconomic Outcomes in Disaster-Prone Countries.” IMF Working Paper 19/217, International Monetary Fund, Washington, DC. Cerra, V., and S. C. Saxena. 2008. “Growth Dynamics: The Myth of Economic Recovery.” American Economic Review 98 (1): 439-57. Cette, G., S. Nevoux, and L. Py. 2022. “The Impact of ICTs and Digitalization on Productivity and Labor Share: Evidence from French Firms.” Economics of Innovation and New Technology 31(8): 669-92. Chalaux, T., and Y. Guillemette. 2019. “The OECD Potential Output Estimation Methodology.” OECD Economics Department Working Paper 1563, Organisation for Economic Co-operation and Development, Paris. Christiano, L. J., and T. J. Fitzgerald. 2003. “The Band Pass Filter.” International Economic Review 44 (2): 435-65. Claessens, S., and M. A. Kose. 2017. “Macroeconomic Implications of Financial Imperfections: A Survey.” BIS Working Paper 677, Bank for International Settlements, Basel. Claessens, S., and M. A. Kose. 2018. “Frontiers of Macrofinancial Linkages.” BIS Paper 95, Bank for International Settlements, Basel. Claessens, S., M. A. Kose, and M. E. Terrones. 2009. “What Happens During Recessions, Crunches and Busts?” Economic Policy 24 (60): 653-700. Claessens, S., M. A. Kose, and M. E. Terrones. 2012. “How Do Business and Financial Cycles Interact?” Journal of International Economics 87 (1): 178-90. Coe, D. T., E. Helpman, and A. W. Hoffmaister. 1997. “North-South R&D Spillovers.” Economic Journal 107 (440): 134-49. Colacito, R., B. Hoffman, and T. Phan. 2018. “Temperature and Growth: A Panel Analysis of the United States.” Working Paper 18-09, Federal Reserve Bank of Richmond, Richmond, VA. Derviş, K. 2018. “The Future of Economic Convergence.” Project Syndicate, February 69 12, 2018. Dieppe, A., and S. Kilic Celik. 2021. “Productivity: Conceptual Considerations and Measurement Challenges.” In Global Productivity: Trends, Drivers, and Policies, edited by A. Dieppe, 59-63. Washington, DC: World Bank. Dieppe, A., S. Kilic Celik, and C. Okou. 2021. “What Happens to Productivity During Major Adverse Events?” In Global Productivity: Trends, Drivers, and Policies, edited by A. Dieppe, 141-197. Washington, DC: World Bank. Dribe, M., M. Breschi, A. Gagnon, D. Gauvreau, H. A. Hanson, T. N. Maloney, S. Mazzoni, et al. 2017. “Socio-Economic Status and Fertility Decline: Insights from Historical Transitions in Europe and North America.” Population Studies 71 (1): 3-21. Duval, R., M. Eris, and D. Furceri. 2011. “The Effects of Downturns on Labour Force Participation: Evidence and Causes.” OECD Economics Department Working Paper 875, Organisation for Economic Co-operation and Development, Paris. Dwyer, G. P., J. Devereux and S. Baie. 2013. “Recessions, Growth and Banking Crises.” Journal of International Money and Finance 38 (November): 18-40. Economides, G., and A. Xepapadeas. 2018. “Monetary Policy under Climate Change.” CESifo Working Paper 7021, Munich Society for the Promotion of Economic Research, Munich. Eggleston, K. N., and V. R. Fuchs. 2012. “The New Demographic Transition: Most Gains in Life Expectancy Now Realized Late in Life.”The Journal of Economic Perspectives(3): 137-56. Elsby, M. W. L., B. Hobijn, and A. Şahin. 2015. “On the Importance of the Participation Margin for Labor Market Fluctuations.” Journal of Monetary Economics 72 (C): 64-82. Fallick, B., and J. Pingle. 2007. “A Cohort-Based Model for Labor Force Participation.” Finance and Economics Discussion Series 2007-09, Board of Governors of the Federal Reserve System, Washington, DC. Fatás, A. 2000. “Do Business Cycles Cast Long Shadows? Short-Run Persistence and Economic Growth.” Journal of Economic Growth 5 (2):-62. Fernald, J., and H. Li. 2021. “The Impact of COVID on Potential Output.” Working Paper 2021-09, Federal Reserve Bank of San Francisco, San Francisco, CA. Feyrer, J. 2007. “Demographics and Productivity.” The Review of Economics and Statistics 89 (1): 100-09. Fletcher, J. M., and B. L. Wolfe. 2009. “Education and Labor Market Consequences of Teenage Childbearing: Evidence Using the Timing of Pregnancy Outcomes and Community Fixed Effects.” The Journal of Human Resources 44 (2): 303-25. Fort, T., J. Haltiwanger, R. Jarmin, and J. Miranda. 2013. “How Firms Respond to Business Cycles: The Role of Firm Age and Firm Size.” NBER Working Paper Series 19134, National Bureau of Economic Research, Cambridge, MA. 70 Furceri, D., S. Kilic Celik, J. T. Jalles, and K. Koloskova. 2021. “Recessions and Total Factor Productivity: Evidence from Sectoral Data.”Economic Modelling94 (January): 130-38. Furceri, D., and A. Mourougane. 2012. “The Effect of Financial Crises on Potential Output: New Empirical Evidence from OECD Countries.”Journal of Macroeconomics(3): 822-32. Gal, P., G. Nicoletti, C. von Rüden. and T. Renault. 2019. “Digitalization and Productivity: In Search of the Holy Grail—Firm-Level Empirical Evidence from European Countries.” International Productivity Monitor 37: 39-71. Giroud, X., and H. M. Mueller. 2017. “Firm Leverage, Consumer Demand, and Employment Losses During the Great Recession.” The Quarterly Journal of Economics 132(1): 271-316. Goldin, C. 1994.“The U-Shaped Female Labor Force Function in Economic Development and Economic History.” NBER Working Paper 4707, National Bureau of Economic Research, Cambridge, MA. Gomez, V. 2001. “The Use of Butterworth Filters for Trend and Cycle Estimation in Economic Time Series.” Journal of Business and Economic Statistics 19 (3): 365-73. Growiec, J., A. Pajor, D. Pell, and A. Predki. 2015. “The Shape of Aggregate Production Functions: Evidence from Estimates of the World Technology Frontier.” Bank i Kredyt 46 (4): 299-326. Gylfason, T., and G. Zoega. 2006. “Natural Resources and Economic Growth: The Role of Investment.” World Economy 29 (8): 1091-115. Haider, S., and D. Loughran. 2001. “Elderly Labor Supply: Work or Play?” Center for Retirement Research Working Paper 2001-04, Boston College, Boston, MA. Haltmeier, J. 2012. “Do Recessions Affect Potential Output?” International Finance Discussion Paper 1066, Board of Governors of the Federal Reserve System, Washington, DC. Hamilton, J. D. 2018. “Why You Should Never Use The Hodrick-Prescott Filter.” Review of Economics and Statistics 100 (5): 831-43. Harvey, A. C. 1990. Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge: Cambridge University Press. Hayat, A. 2018. “FDI and Economic Growth: The Role of Natural Resources?” Journal of Economic Studies 45 (2): 283-95. Herrera, C., D. E. Sahn, and K. M. Villa. 2016. “Teen Fertility and Labor Market Segmentation: Evidence from Madagascar.” IZA Discussion Paper 10464, IZA Institute of Labor Economics, Bonn, Germany. Hodrick, R. J., and E. C. Prescott. 1997. “Postwar U.S. Business Cycles: An Empirical Investigation.” Journal of Money, Credit and Banking 29 (1): 1-16. 71 Huckfeldt, C. 2022. “Understanding the Scarring Effect of Recessions.” American Economic Review 112 (4): 1273-310. Huidrom, R., M. A. Kose, and F. Ohnsorge. 2016. “Challenges of Fiscal Policy in Emerging Market and Developing Economies.” CAMA Working Paper 34/2016, Crawford School of Public Policy, Sydney, Australia. ILO (International Labour Office). 2019. “Working on a Warmer Planet: The Impact of Heat Stress on Labor Productivity and Decent Work.” International Labour Office, Geneva. IMF (International Monetary Fund). 2015. “Where Are We Headed? Perspectives on Potential Growth.” In World Economic Outlook April 2015. Washington, DC: International Monetary Fund. IMF (International Monetary Fund). 2019. “Eastern Caribbean Currency Union.” IMF Selected Issues Paper, IMF Country Report 19/63, International Monetary Fund, Washington, DC. IMF (International Monetary Fund). 2021. Reaching Net Zero Emissions. Washington, DC: International Monetary Fund. Jordà, Ò. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review 95 (1): 161-82. Jordà, Ò., M. Schularick, and A. M. Taylor. 2013. “When Credit Bites Back.” Journal of Money, Credit and Banking 45 (s2): 3-28. Justino, P., M. Leone, and P. Salardi. 2011. “Education and Conflict Recovery: The Case of Timor Leste.” Policy Research Working Paper 5774, World Bank, Washington, DC. Kahn, M. E., K. Mohaddes, R. N. C. Ng, M. H. Pesaran, M. Raissi, and J.-C. Yang. 2019. “Long-Term Macroeconomic Effects of Climate Change: A Cross-Country Analysis.” IMF Working Paper 19/215. International Monetary Fund, Washington, DC. Kato, H. 2016. An Empirical Analysis of Population and Technology Progress. Tokyo: Springer. Kemp, J. H. 2015. “Measuring Potential Output for the South African Economy: Embedding Information About the Financial Cycle.” South African Journal of Economics 83 (4): 549-68. Kilic Celik, S., M. A. Kose, and F. Ohnsorge. 2020. “Subdued Potential Growth: Sources and Remedies.” In Growth in a Time of Change: Global and Country Perspectives on a New Agenda, edited by H.-Y. Kim and Z. Qureshi, 25-74. Washington, DC: Brookings Institution. Kilic Celik, S., M. A. Kose, and F. Ohnsorge. 2023. “Potential Growth Prospects: Risks, Rewards, and Policies.” Policy Research Working Paper 10355, World Bank, Washington, DC. Kinoshita, Y., and F. Guo. 2015. “What Can Boost Labor Force Participation in Asia?” 72 IMF Working Paper 56, International Monetary Fund, Washington, DC. Klasen, S., and J. Pieters. 2013. “What Explains the Stagnation of Female Labor Force Participation in Urban India?” Discussion Papers 146, Courant Research Centre: Poverty, Equity and Growth, Göttingen, Germany. Kydland, F. E., and E. C. Prescott. 1982. “Time to Build and Aggregate Fluctuations.” Econometrica 50 (6): 1345-70. Laeven, L., and F. Valencia. 2012. “Systemic Banking Crises Database: An Update.” IMF Working Paper 12/163, International Monetary Fund, Washington, DC. Laeven, L., and F. Valencia. 2018. “Systemic Banking Crises Revisited.” IMF Working Paper 18/206, International Monetary Fund, Washington, DC. Laeven, L., and F. Valencia. 2020. “Systemic Banking Crises Database II.” IMF Economic Review 68 (2): 307-61. Laubach, T., and J. C. Williams. 2003. “Measuring the Natural Rate of Interest.” Review of Economics and Statistics 85 (4): 1063-70. Li, D., M. Plagborg-Møller, and C. K. Wolf. 2022. “Local Projections vs. VARs: Lessons from Thousands of DGPs.” NBER Working Paper 30207, National Bureau of Economic Research, Cambridge, MA. Linacre, S. 2007. “Labour Force Participation: An International Comparison.” Australian Social Trends 4102: 1-7. Liu, K., K. G. Salvanes, and E.Ø. Sørensen. 2016. “Good Skills in Bad Times: Cyclical Skill Mismatch and the Long-Term Effects of Graduating in a Recession.” European Economic Review 84: 3-17. Loayza, N. and S. Pennings. 2022. The Long Term Growth Model: Fundamentals, Extensions, and Applications. Washington, DC: World Bank. Lockwood, B. 1991. “Information Externalities in the Labour Market and the Duration of Unemployment.”The Review of Economic Studies,58 (4): 733-53. Moran, T. and L. Oldenski. 2016. “How Offshoring and Global Supply Chains Enhance the US Economy,” Policy Brief 16-5, Peterson Institute for International Economics, Washington, DC. Missirian, A., and W. Schlenker. 2017. “Asylum Applications Respond to Temperature Fluctuations.” Science 358 (6370): 1610-14. Mourougane, A. 2017. “Crisis, Potential Output and Hysteresis.” International Economics 149 (May): 1-14. Nguyen, H., and R. Qian. 2014. “Demand Collapse of Credit Crunch to Firms? Evidence from World Bank’s Financial Crisis Survey in Eastern Europe.” Journal of International Money and Finance 47 (October): 125-44. OECD (Organisation for Economic Co-operation and Development). 2014. “Growth 73 Prospects and Fiscal Requirements over the Long Term.” In OECD Economic Outlook, Vol. 2014/1. Paris: Organisation for Economic Co-operation and Development. Okun, A. M. 1962. “Potential GNP and Its Measurement and Significance.” Proceedings of the Business and Economics Statistics Section of the American Statistical Association, 98-104. Ollivaud, P., and D. Turner. 2014. “The Effect of the Global Financial Crisis on OECD Potential Output.” OECD Journal: Economic Studies: 41-60. Oulton, N., and M. Sebastia-Barriel. 2016. “Effects of Financial Crises on Productivity, Capital and Employment.” The Review of Income and Wealth 63 (s1): S90-S112. Panton, A. J. 2020. “Climate Hysteresis and Monetary Policy.” CAMA Working Paper 76/2020. Centre for Applied Macroeconomic Analysis, The Australian National University. Plagborg-Møller, M., and C. K. Wolf. 2021. “Local Projections and VARs Estimate the Same Impulse Responses.” Econometrica 89 (2): 955-80. Queralto, A. 2013. “A Model of Slow Recoveries from Financial Crises." International Finance Discussion Papers 1097, Board of Governors of the Federal Reserve System, Washington, DC. Quian, C. Z., Y. Liu, and V. Steenbergen. 2022. “Global Value Chains in the Time of COVID-19.” In An Investment Perspective in Global Value Chains, edited by C. Z. Quian, Y. Liu, and V. Steenbergen. Washington, DC: World Bank. Rodrik, D. 2015. “Back to Fundamentals in Emerging Markets.” Project Syndicate, August 13, 2015. Shimer, R. 2013. “Job Search, Labor-Force Participation, and Wage Rigidities.” In Advances in Economics and Econometrics: Tenth World Congress, edited by D. Acemoglu, M. Arellano, and E. Dekel. Cambridge: Cambridge University Press. Siklos, P. L. 2013. “Sources of Disagreement in Inflation Forecasts: An International Empirical Investigation.” Journal of International Economics 90 (1): 218-31. Solow, R. M. 1962. “Technical Progress, Capital Formation, and Economic Growth.”The American Economic Review52 (2): 76-86. Stokke, H. E. 2008. “Resource Boom, Productivity Growth and Real Exchange Rate Dynamics—A Dynamic General Equilibrium Analysis of South Africa.” Economic Modelling 25 (1): 148-60. Syverson, C. 2011. “What Determines Productivity?” Journal of Economic Literature 49 (2): 326-65. Tansel, A. 2002. “Determinants of School Attainment of Boys and Girls in Turkey: Individual, Household and Community Factors.” Economics of Education Review 21 (5): 455-70. 74 Teulings, C. N., and N. Zubanov. 2014. “Is Economic Recovery a Myth? Robust Estimation of Impulse Responses.” Journal of Applied Econometrics 29 (3): 497-514. Torvik, R. 2002. “Natural Resources, Rent Seeking and Welfare.” Journal of Development Economics 67(2): 455-70. Turner, D., M. C. Cavalleri, Y. Guillemette, A. Kopoin, P. Ollivaud, and E. Rusticelli. 2016. “An Investigation into Improving the Real-Time Reliability of OECD Output Gap Estimates.” Economics Department Working Paper 1294, Organisation for Economic Co- operation and Development, Paris. Wilms, P., J. Swank, and J. de Haan. 2018. “Determinants of the Real Impact of Banking Crises: A Review and New Evidence.” The North American Journal of Economics and Finance 43 (January): 54-70. World Bank. 2018. Learning to Realize Education’s Promise. Washington, DC: World Bank. World Bank. 2021. Global Economic Prospects. June. Washington, DC: World Bank. World Bank. 2021. The Human Capital Index 2020 Update: Human Capital in the Time of COVID-19. Washington, DC: World Bank. United Nations. 2012. “World Population Monitoring. Adolescents and Youth.” United Nations, New York. Yagan, D. 2019. “Employment Hysteresis from the Great Recession.” Journal of Political Economy 127 (5): 2505-58. 75