key: cord-0777028-fz478658
authors: Furceri, Davide; Celik, Sinem Kilic; Jalles, João Tovar; Koloskova, Ksenia
title: Recessions and Total Factor Productivity: Evidence from Sectoral data()
date: 2020-10-07
journal: Econ Model
DOI: 10.1016/j.econmod.2020.09.025
sha: 48910397cf6fc5867fffc8e2bf24d56701da41ee
doc_id: 777028
cord_uid: fz478658

The recent COVID-19 crisis has generated concern that productivity (which was already at historically low levels) may further decline. From a theoretical standpoint, the recessions-total factor productivity (TFP) nexus is ambiguous à priori. This paper empirically examines the dynamic impact of recessions on TFP. We compute a new measure of utilization-adjusted productivity from a sample of 24 industries in 18 advanced economies between 1970 and 2014. Resorting to the local projection method we trace out the dynamic short to medium-term impact of such recessionary shocks. We find that deep recessions lead to a permanent deterioration in the level of total factor productivity. This effect is driven by the increase in resource misallocation across different sectors.

In many advanced economies, productivity growth has been declining since the late 1990s and is now at historically low levels (Adler et al., 2017) . Several (Tett, 2015) . 1 The recent COVID-19 crisis has generated concern that productivity may further decline (Mauro and Syverson, 2020 and Baldwin and Weder di Mauro, 2020). 2 From a theoretical standpoint, the recessions-total factor productivity nexus is ambiguous à priori. On the one hand, in models of knowledge accumulation (e.g., Grossman and Helpman, 1991) , a fall in the proportion of the labor force dedicated to research and development following a negative output shock could lead to a long-lasting negative impact on the level of productivity. On the other hand, in models of Schumpeter's creative destruction (e.g. Caballero and Hammour, 1994) , recessions can have a cleansing economic effect by eliminating inefficient firms, consequently generating higher productivity and GDP growth. 3 In particular, aggregate productivity can be affected by recessions in two separate manners: i) by means of their impact on productivity 1 A recent paper by Crats and Mills (2020) conjectures that the UK productivity slowdown is unprecedented in the past 250 years and reflects a combination of adverse circumstances, namely, a financial crisis, a weakening impact of ICT and impending Brexit. 2 For a recent paper relating productivity and the COVID-19 shock using industry-level data in Canada see Blit (2020) . 3 Theoretically, the different set of constraints forced by recessions should close business of the least productivity firms and rise aggregate productivity (Petrosky-Nadeau, 2013). Relatedly, we also have the "pit-stop" view of recessions in which these correspond to periods of reduced profits and times during which restructuring can be done due to temporarily lower opportunity costs (Aghion and Saint-Paul, 1998 ). More recently, Gropp et al. (2020) assessed the cleansing effects of the Global Financial Crises looking at US regions. 3 within each sector; ii) by inducing sectoral reallocations of input factors across different sectors. 4 The effect through sectoral reallocation is not clear, since labor can move between various lowand high-productivity sectors, with an ambiguous net effect on productivity. Counter-cyclical reallocation takes place when input factor reallocation during a downturn leads to less productive jobs being destroyed and labor moving into more productive uses (Mortensen and Pissarides, 1994) . Pro-cyclical reallocation occurs when more productive industries are disproportionately affected by recessions, for example due to credit constraints (Barlevy, 2003) . Therefore, the extent to which recessions impact total factor productivity (TFP) is ultimately an empirical question. Against this background, this paper offers new empirical evidence on the impact of recessionary periods on productivity, including through reallocation.

While several studies have looked into this issue, we improve upon previous studies in two ways.

First, we construct a new measure of TFP growth adjusted for time-varying unobserved use in capital and labor. 5 Correcting for unobserved input factor utilization is crucial to accurately assess the development of aggregate TFP across the different phases of the business cycle and in the aftermath of recessions. Second, we explore the role of within-and between-sectors productivity.

Our analysis is carried out in three stages. First, we calculate time series of cyclicallyadjusted TFP growth following the method used by Basu et al. (2006) The rest of the paper is organized in the following way. Section 2 discusses the empirical methodology and presents the data used to evaluate the impact of recessions on productivity. 

We follow the approach of Basu et al. (2006) and Fernald (2014a Fernald ( , 2014b to estimate a measure of TFP growth adjusted for time-varying unobserved use in capital and labor, based on sectoral data. Correcting for unobserved input factor utilization is crucial to accurately assess the development of aggregate TFP growth across the different phases of the business cycle and, specifically, in the aftermath of recessions-the focus of the paper.

The approach consists in the following. Let's start assuming that the production function of each industry i is given by:

where ‫ܣ‬ ,௧ is TFP of industry i at time t; K and L denote the capital stock and labor inputs, respectively; the symbol ෩ denote utilization-adjusted capital and labor effort:

J o u r n a l P r e -p r o o f 6 where ‫ܷܥ‬ ,௧ and ‫ܧ‬ ,௧ denote capacity utilization for capital stock and labor effort, respectively.

Following Basu et al. (2006) , the production function ‫ܨ(‬ ) is assumed to be homogeneous of degree ߙ in total inputs. Assuming a cost minimization objective, output growth is related to the growth of input factors for each industry i as follows:

where dy refers to the growth rate of gross output and ‫݂ݐ݀‬ is TFP growth. dx is the observed input growth-defined as ‫ݔ݀‬ = s , ݀݇ + s , ݈݀ + s , ݀݉ ୧ ; and du is the unobserved input utilization growth-defined as ‫ݑ݀‬ = s , ‫ݑܿ݀‬ + s , ݀݁ .

The aggregate TFP growth is then obtained as the aggregate Solow residual minus the aggregate input utilization:

This implies that TFP growth is equal to the Solow residual when there are no changes in utilization. The aggregate Solow residual and input utilization are calculated as weighted sums of industry Solow residuals and utilization measures, respectively:

7 ‫ݓ‬ is the industry' proportion in gross output, and ‫݉ݏ‬ is the proportion of input payments in total costs.

The measure of cyclically-adjusted TFP is computed for an unbalanced sample of 18 advanced economies from 1970 to 2014, using EU KLEMS sectoral data for 24 industries. 6 Table 1 shows some descriptive statistics of the measure of cyclically-adjusted TFP computed for the countries in our sample and it compares with those of the standard Solow residual. Figure   1 plots, for each of the 18 countries in the sample, the time profile of both measures (additional summary statistics are provided in Table A1 ).

[insert Table 1] [insert Figure 1 ]

Using the method proposed by McMillan and Rodrick´s (2014), we decompose aggregate TFP is into two components: the within part; and the between part. 7 Specifically, we decompose the cumulative changes in TFP between time t and time t+k as follows: 6 The list of countries in our sample is the following: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Korea, the Netherlands, Portugal, Spain, the UK and the US. 7 Labor reallocation from low-to high-productivity activities was a key determinant of productivity growth in several regions of the world. The impact of structural reallocation on aggregate productivity was studied by Carree (2003) for the advanced economies, Olley and Pakes (1996) and Timmer and Szirmai (2000) for the Asian region, , Bartelsman is the logarithm of the cyclically-adjusted TFP of sector i in country j, and ‫ݓ‬ ,,௧ is the proportion of sector i's value added. 8 The first right-hand side term corresponds to the within component while the second right-hand side term corresponds to the between component.

The econometric method we use to assess the dynamic impact of economic downturns on productivity is Jordà´s (2005) local projection approach. The method allows us to trace the dynamics of productivity growth in the aftermath of recessions, controlling for pre-crisis trends in productivity. The following equation is estimated for each horizon k = 0, . . ., 5:

where ‫ݔ‬ ௧ା, − ‫ݔ‬ ௧ିଵ, is the cumulative change of the dependent variable (cyclically-adjusted TFP and its within and between components) for country j; ߙ and ߛ ௧ denote respectively country and time fixed effects. D is a dummy variable taking the value of one at the beginning of a recession period (see definition below), and zero otherwise. Additional controls include: (i) two lags of past output growth, (ii) two lags of recession dates, (iii) country-specific trends. Note that 8 These weights are those we also use to compute our aggregate measure of cyclically-adjusted TFP growth.

Weights are given by ‫ݓ‬ /(1 − ‫݉ݏ‬ ) in which ‫ݓ‬ is the industry's proportion in gross output, and ‫݉ݏ‬ is the proportion of input payments in total costs (Basu et al, 2006) . The industry's proportion in aggregate value added is used as weight in TFP growth rates. 9 at some forecasting horizon k, the dependent variable may already be affected by the recession, even though the variable measuring recession is set equal to zero. In such case, the effect of the recession on the dependent variable will be absorbed by the fixed effects rather than being (Table A2) . We then extend the analysis to consider alternative recessions and compare their productivity effects with those associated with deep recessions.

Before moving to the analysis on the effect of deep recessions on cyclically-adjusted TFP it is useful to examine the effects on aggregate output and input of production. To do so we estimate equation (6), alternatively using as dependent variables total output, employment and capital. The results presented in Figure 2 show [insert Figure 4 ]

Digging further, the results-not shown but available from the authors upon requestindicate that following past deep recessions economic activity has re-allocated from high productivity sectors such as information and communication, wholesale and retail trade and transportation to low-productivity sectors such as social services and real estate (results not shown but available upon request). These results are consistent with evidence provided in other papers suggesting that high productivity sectors tend to contract more during downturns (e.g.

Results shown until now provided evidence that, on average, deep recessions have a significant and persistent negative effect on TFP. However, such losses are likely to vary with the severity and the type of recessions (real vs. financial crisis). To test for this possibility, weestimate equation (6) using two alternative recessions. In the first exercise, we construct a dummy that takes value 1 for all the other "non-severe" recessions identified in Blanchard et al.

(2015). In the second analysis, we consider, severe recessions that coincided with financial crises-for this purpose, we rely on the financial crisis database constructed by Laeven and Valencia (2018). 9 The results for normal recessions are displayed in Figure 5 . In contrast, to deep recessions, normal recessions have not statistically significantly effects on TFP and its components. This 9 There is a large body of literature on the output costs of a financial crises. Most seem to agree that these crises have severe output effects- Romer and Romer (2017) [insert Figure 5 ]

In Figure 6 , we present the results for deep recessions that are also financial crises. Overall, the results do not point to any systematic difference between "real" deep recessions and "financial" deep recession-if any the results for deep financial crises are less precisely estimated because of fewer episodes. In both cases, TFP declines by about 3 percent in the medium-term, with the persistent effect mostly driven by sectoral re-allocation (the between component).

[insert Figure 6 ]

We examine whether the previous cross-country average effects mask considerable heterogeneity across countries. In particular, given that several studies found that the allocation of resources worsened in southern Europe in the last two decades (e.g., Reis To address this issue, we re-estimate equation (6) alternatively including as control variables credit growth, the current account balance (in percent of GDP) and the debt-to-GDP ratio. Figure   7 shows the results for the cyclically-adjusted TFP within and between components and confirms that such augmentations of the vector of controls do not change the basic thrust of our results;

that is, TFP goes down following a deep recession and most of that fall is due to between-sector reallocation of input factors.

[insert Figure 8 ]

In many advanced economies, productivity growth has been declining since the late 1990s and is now at historically low levels. There are worries that the current COVID-19 crisis may exacerbate this negative trend and further reduce total factor productivity growth. While there are different channels through which the current and past crises can lead to permanent J o u r n a l P r e -p r o o f 14 productivity losses, one often-cited factor is through sectoral reallocation-that is, the betweensector component of productivity that reflect reallocation of resources across sectors.

This paper provided new evidence on the potential impact of the COVID-19 crisis by examining the effect of past deep recessions on a newly constructed measure of cyclicality adjusted TFP. We find that deep recessions are not only associated with a persistent decline in output and its input of production, but also with permanent TFP losses. In particular, we find that cyclically-adjusted TFP strongly decline in the aftermath of deep recessions with the amplitude of the effect lying between 3-4 percent after 3-5 years. This effect comes from both the within and between components of TFP growth, but sectoral reallocation (the between effect) explains most of the medium-term decline in productivity.

Will the COVID-19 lead to similar losses? It is clearly too early to say, and further research will be needed. The crisis will surely lead to some sectoral reallocation from tourism, restaurants and other services that require in-person contact to sectors such as communication and IT. But the effect on aggregate productivity will much depend on the level of productivity of each of these sectors, which may vary across countries.

Much will also depend on the ability of policies to soften the persistence of the crisis and to reduce scarring effects. In this context, stimulating demand through fiscal and monetary policy and ensuring flow of credit to household and business would be key to mitigate the initial cost of the crisis. Significant fiscal stimulus on high-return investment will be also needed to sustain the recovery and should be complemented with structural reforms aimed at improving efficiency and help ensure resources are reallocated appropriately. 

Gone with the Headwinds: Global Productivity

Virtues of Bad Times: Interaction between Productivity Growth and Economic Fluctuations

Default Risk, Sectoral Reallocation and persistent recessions

Labor Productivity: Structural Change and Cyclical Dynamics

Economics in the Time of COVID-19, a VoxEU.org eBook

Credit market frictions and the allocation of resources over the business cycle

Microeconomic Evidence of Creative Destruction in Industrial and Developing Countries

Are Technology Improvements Contractionary?

Inflation and Activity -Two Explorations and their Monetary Policy Implications, IMF Working Papers 15/230

Is increasing productivity COVID-19's silver lining?

The Cleansing Effect of Recessions

Technological progress, structural change and productivity growth: a comment

Growth Dynamics: The Myth of Economic Recovery

The Pre-Great Recession Slowdown in Productivity

Structural Change, Labour Productivity Growth and Convergence of BRIC Economies

How do business and financial cycles interact?

Is the UK productivity slowdown unprecedented?

On the Driving Forces Behind Cyclical Movements in Employment and Job Reallocation

Labor Market Flows in the Cross Section and Over Time

What Determines Output Losses after Banking Crises?

On the Measurement of Technical Change

Productivity and Potential Output Before, During, and After the Great Recession

A Quarterly, Utilization-Adjusted Series on Total Factor Productivity. FRBSF Working Paper

Market Selection, Reallocation and Restructuring in the U.S. Retail Trade Sector in the 1990s

The effect of financial crises on potential output: New empirical evidence from OECD countries

Capital Allocation and Productivity in South Europe

Good Booms, Bad Booms

The Cleansing Effect of Banking Crises

Quality Ladders and Product Cycles

Dissecting the Cycle: A Methodological Investigation

Estimation and Inference of Impulse Responses by Local Projections

When Credit Bites Back

Private credit and public debt in financial crises

Systemic Banking Crises Revisited

The Covid crisis and productivity growth

Job Creation and Job Destruction in the Theory of Unemployment

The Dynamics of Productivity in the Telecommunications Equipment Industry

Credit frictions and the cleansing effect of recessions

Long and short-term effects of the financial crisis on labour productivity, capital and output. Bank of England Working Paper no. 470 and Centre for Economic Performance

TFP during a credit crunch

A model of slow recoveries from financial crises

The Portuguese Slump and Crash and the Euro Crisis

The Causes and Costs of Misallocation

New Evidence on the Aftermath of Financial Crises in Advanced Economies

Structural Change and Economic Development in China and India

Productivity paradox deepens Fed's rate-rise dilemma

Is Economic Recovery a Myth? Robust Estimation of Impulse Responses

Productivity growth in Asian manufacturing. The structural bonus hypothesis examined

Exporting raises productivity in sub-Saharan African manufacturing firms

Global Economic Prospects. World Bank

J o u r n a l P r e -p r o o f• COVID-19 has generated concern that productivity may further decline.• This paper examines the dynamic impact of recessions on productivity.• The local projection method is used on a sample of 18 advanced economies.• We find that recessions permanently lower the level of productivity.• The effect is driven by the increase in resource misallocation across sectors.J o u r n a l P r e -p r o o f

The authors confirm that there is no conflict of interest.All financial support has been properly acknowledged.J o u r n a l P r e -p r o o f