Journal Issue
Share
Article

The New Normal

Author(s):
Era Dabla-Norris, Si Guo, Vikram Haksar, Minsuk Kim, Kalpana Kochhar, Kevin Wiseman, and Aleksandra Zdzienicka
Published Date:
March 2015
Share
  • ShareShare
Show Summary Details

I. Context

1. A new mediocre? Global growth has been anemic in the wake of the balance sheet recessions that rocked advanced economies during 2008–12. The malaise has spread from advanced economies to many emerging markets. Not only are recoveries more protracted in spite of large policy stimulus, questions are being raised about future sources of growth. This, in turn, raises uncertainty about the understanding of potential output. As a result, the policy debate on dealing with the weak recovery has been fraught. Attention has shifted from an early focus on demand stimulus, which is still needed in some key economic areas, to measures to boost global supply amid slowing trend growth and large persistent output losses (IMF, forthcoming).

2. Should we be surprised? The current debate on the slow recovery could benefit from a longer-term perspective. While variation exists across countries and time, one clear mega-trend emerges. Underlying output and productivity growth in advanced economies had already been slowing well before the global financial crisis. The ups and downs associated with the financial cycle, along with periods of productivity gains from the adoption of new technologies and efficiency-enhancing reforms, may have masked this trend. Still, the finding is in line with economic theory that suggests that some deceleration in economic growth in advanced economies is inevitable as countries get closer to the technology frontier. Indeed, the experiences of advanced economies that exhibited convergence in the past (Japan in the 1970s) suggest that during the transition to higher income levels, catch-up gains from capital deepening, learning and technology transfer, and structural transformation diminish, and productivity and growth slows. Moreover, working-age population growth has slowed and old-age dependency ratios have increased. Prospects for boosting medium-term growth in advanced economies via increasing factor accumulation alone are likely limited.

3. The challenge to enhance productivity. Increasing productivity growth must remain a priority for advanced economies to revive growth and maintain living standards. This implies tackling market and institutional rigidities, and boosting infrastructure and innovation to spur productivity growth and maintain competitiveness. This is reflected in the increased emphasis in domestic policy debates on reforms to support productivity and efficiency. This was echoed at the global level in the Fall 2014 Global Policy Agenda presented by the IMF Managing Director, which emphasized measures to help increase productive capacity (IMF 2014a). This is a complex and multidimensional task, requiring reforms and, hence, a granular approach to policy analysis and application.

4. Focus of this Note. Cross-country narratives of the growth slowdown and future prospects have focused on aggregate trends in growth, factor accumulation, and productivity. This study focuses on sector-level productivity developments before the global financial crisis, examining the period from 1970 to 2007 in order to highlight trends already in progress before the onset of the crisis and to provide evidence on underlying productivity developments. This allows for identifying key sector sources of total factor productivity (TFP) gaps that existed even prior to the crisis, exploring issues of misallocation of resources across sectors, and, importantly, more robustly identifying policy requirements than would be possible with aggregate cross-country data. The work here is based on a unique and rich industry-level dataset across advanced economies.2 The database provides harmonized input and output series’ suitable for cross-country productivity analysis. But as with any cross-country datasets, measurement issues and scope of country coverage, particularly at the detailed industry level, are important caveats.

5. Roadmap. Section II examines trends in aggregate and sectoral growth, productivity, and structural transformation across advanced economies. This is a useful starting point to see how economies and the associated growth bottlenecks evolved even prior to the crisis. Section III presents a quantitative thought experiment on how improved factor allocation across sectors could boost output levels—the scope for gains by reducing such inefficiencies is large, especially in slow-growing economies. Section IV then examines how policies—which run the gamut from product market and labor market reforms to boosting technology development and adoption—matter for productivity growth across countries. The section also assesses the dynamic effects of reforms on output and employment as a way of examining potential trade-offs. Policy priorities are derived from this work to boost productivity and offset the secular slowdown apparent in advanced economies.

II. Growth and Structural Transformation

A slowing pace of human and physical capital accumulation and declining TFP contributed to lower trend growth in the years immediately preceding the global financial crisis. The aggregate decline in TFP reflected the reallocation of resources to sectors where productivity growth was slower, as well as declining productivity growth within those sectors, which increasingly account for the bulk of employment and economic activity. Large sectoral TFP-level gaps across countries, however, point to opportunities for boosting growth.

A. The Aggregate Story

6. Secular decline in output and productivity growth. Advanced economies have experienced a secular decline in growth and average labor productivity over the last four decades. These trends are evident even in pre-crisis data, and were further magnified by the global financial crisis. Figure 1 shows long-run trend declines in output growth and labor productivity growth before the crisis.3 Output growth declined from an annual average of more than 3½ percent in the 1970s to less than 2½ percent in the mid-2000s just before the crisis. This decline was largely driven by labor productivity. Trend growth in output per hour worked declined from about 3 percent annually in 1970 to less than 1½ percent in 2007. Trend growth extended to encompass the period of the crisis is, not surprisingly, even gloomier.

Figure 1.GDP and Labor Productivity Trend Growth

(In percent; purchasing power parity weighted)

Sources: Penn World Tables; and IMF staff calculations.

Note: H. spread EA stands for euro area countries with high borrowing spreads.

7. Cross-country heterogeneity.4 This secular decline in labor productivity growth applies to a range of advanced economies, albeit with significant variation. The slowing trend was most marked in core Europe and in euro area countries with high borrowing spreads (van Ark, O’Mahony, and Timmer 2008; Blanchard 2004), where growth fell from the high levels achieved in the recovery following World War II. The decline moderated in the 1980s due to significant reforms, but resurfaced again by the mid-1990s. Japan saw similar high and sustained catch-up growth through the end of the 1980s, but trend labor productivity growth fell to the advanced economy average by the late 1990s. The United States saw a mini-boom as information and communication technologies (ICT) sparked a productivity surge (Jorgenson, Ho, and Stiroh 2008). The surge, which occurred primarily from 1995–2003 (Fernald 2014), temporarily offset the long-run decline and was waning by the time of the crisis. After an extended period of lackluster growth, labor productivity in the United Kingdom grew faster than in most countries from the mid-1990s.

8. Labor supply played a secondary role. Narratives of long-run growth decline in advanced economies often focus on labor supply issues—moribund fertility rates, aging populations, and a declining employment ratio (falling labor force participation—reported below as labor utilization—and rising unemployment). These, however, did not significantly slow the long-run growth trend. Figure 2 shows a clear secular decline in the contribution of labor productivity to total growth as driving the broader trend.

Figure 2.Average Labor Supply Growth Contributions

(In percent; purchasing power parity weighted)

Sources: Penn World Table; World Bank, World Development Indicators; and IMF staff calculations.

Note: H. spread EA stands for euro area countries with high borrowing spreads.

9. Demographic headwinds loom. The right-side panel of Figure 2 indicates the labor supply components that have impeded growth for different country groups. Observations below the 45 degree line contributed to the growth slowdown, exerting a greater drag in more recent periods. The rising dependency ratio (more retirees and children per working-age person) is a clear example of a factor exerting downward pressure for most countries. Japan, in particular, has seen a major increase in old-age dependency since the 1990s, with population growth also turning negative. Changes in labor utilization (the fraction of the working-age population in the work force) were less of a drag after 1990 than before, reflecting increasing female labor participation in some countries. Demographic factors will, however, bear down on future growth prospects as population growth follows the dependency ratio down, and as aging populations dissave in retirement and seek lower-risk investments.

10. Capital deepening struggled. The decline in the growth of human and physical (both ICT and non-ICT) capital per worker since the 1990s has been an important contributor to the overall labor productivity decline in advanced economies (Figure 3).5 ICT-capital deepening, widely regarded as giving the United States a labor productivity edge over European economies, retreated from the peak reached in the mid-1990s, particularly in the United States. The decline in non-ICT capital deepening in many advanced economies also predated the crisis, reflecting in part a long-run decrease in public investment (IMF 2014b). Moreover, educational attainment, measured by average years of schooling, also appears to be close to plateauing in most economies. In fact, for cohorts born after 1950 in the United States, the rise in educational attainment slowed markedly and has ceased for the most recent cohort (Fernald and Jones 2014).

Figure 3.Contribution of Capital Deepening to Labor Productivity

(In percent)

Sources: Conference Board; EU KLEMS database; and IMF staff calculations.

11. TFP growth stagnated or declined in most of the developed world.6 TFP growth fell dramatically throughout the crisis but was stagnant even at the pre-crisis peak across all economies (Figure 4).7 TFP growth fell most dramatically in euro area countries with high borrowing spreads, registering a half percentage point decline on average every decade. The slowdown in TFP growth in the United States, commonly regarded as the world technology frontier, also started well before the crisis (Fernald 2014). Japan was a rare success story, with TFP growth bouncing back from negative territory reached in the 1990s, reflecting firm-level restructuring. Core European countries were also rationalizing their business models in the 1990s and 2000s, facilitated by labor market and other reforms in Germany and smaller economies (e.g., Sweden, Finland). Natural resource economies, which experienced commodity booms, experienced a big drop in aggregate TFP growth in the pre-crisis 2000s.8

Figure 4.Trend Total Factor Productivity Growth

Sources: EU KLEMS database; World KLEMS database; and IMF staff calculations.

Note: Trends were estimated with an HP filter on the pre-crisis (through 2007) and full sample (through 2011) total factor productivity data. PPP = purchasing power parity.

12. Technological progress versus efficiency. An alternative way to analyze TFP growth across advanced economies is to examine contributions from technological progress and improvements in efficiency. We apply stochastic frontier analysis (SFA) to estimate a TFP frontier across all advanced economies, representing the maximum possible TFP levels under current technologies (see Annex 1). We also estimate a distance from the frontier for each country, which represents the efficiency with which global technological know-how is being implemented domestically. In this framework, technological progress shifts the production frontier upward for all industries and all countries, while an improvement in technical efficiency moves countries towards the production frontier.

Figure 5.Sources of Total Factor Productivity Growth

(Stochastic Frontier Analysis)

Sources: EU KLEMS database; World KLEMS database;

University of Groningen; and IMF staff calculations.

13. Advances in both the technology frontier and the pace of catch-up to the frontier stalled. Our results show the pace of technological change in secular decline, while technical efficiency (the average country’s distance to the frontier), after some initial convergence, also stalled (Figure 5). Rolling window estimates of the SFA model during the period 1983–2007 suggest that the growth rate of the productivity frontier declined at about a third of a percentage point per decade (the blue line in Figure 5). Estimated technical efficiency also stalled in our estimates over time, with the average country moving away from the frontier (the red line in Figure 5). These trends warrant a more in-depth examination of the factors underlying TFP growth.

B. Sectoral Productivity Dynamics

14. Sectoral perspective. Stagnant and declining TFP growth at the aggregate level, reflected in declining technological progress, can be indicative of slowing human and physical capital accumulation, a falling pace of sector-specific innovation, and structural shifts to lower productivity sectors. At the same time, the stalled pace of catch-up to the frontier can be reflective of limited technological spillovers from the frontier, the waning impact of past reforms, and institutional and regulatory rigidities that stifle competition and induce slower uptake of existing technologies. This subsection takes a sectoral perspective to assess the drivers of underlying aggregate productivity performance before the crisis.

15. Structural transformation—the reallocation of factors of production across sectors and activities and sectors—can lead to higher growth if driven by technical change and efficient factor allocation. Yet, different forces have been driving this process within advanced economies. Globalization, for instance, has been accompanied by the outsourcing of some high-productivity manufacturing in many advanced economies. Consumer preferences also change as incomes rise, increasing demand for services (Herrendorf, Rogerson, and Valentinyi 2013). At the same time, technological changes, such as the ICT revolution and the shift into more skill-intensive output, have had profound implications for production structures. Indeed, over the last four decades, advanced economies experienced a well-documented shift away from agriculture and manufacturing and into services. Currently the services sector is dominant, accounting for more than three-quarters of the total labor share. Hence the dynamics of output and productivity within the sector are more important quantitatively (Jorgenson and Timmer 2011).

16. Shifts within the services sector. An analysis of employment and value-added shares within the services sector and the underlying shifts between 1980 and 2007 points to a number of commonalities across advanced economies (Figure 6). The increase in value-added and employment shares in finance and business services—sectors that experienced a pre-crisis boom—was universal, and particularly marked in the United Kingdom, where the labor share in finance and business services was the highest among advanced economies. Personal services (e.g., hotels, restaurants, social, and personal services) and nonmarket services (e.g., public administration, education, health, and real estate) also gained labor share in most economies. Productivity growth in these services was much lower than in the rest of the economy due to limited scope for innovation and technical change (Baumol, Blackman, and Wolff 1985). The growth in labor shares in nonmarket services in euro area countries with high borrowing spreads was particularly striking in this regard. Yet other services, such as distribution and ICT services, experienced falling prices and rising labor productivity growth, gaining real value-added share while shedding labor.

Figure 6.Value Added and Employment, 1980–2007

(In percent; purchasing power parity weighted)

Sources: EU KLEMS database; World KLEMS database; and IMF staff calculations.

Note: ICTGS = information communication technology goods and services; LCU = local currency unit.

17. TFP growth within individual sectors. The aggregate decline in TFP growth in many advanced economies prior to the crisis was driven not only by the reallocation of resources to sectors where productivity growth was slower, but, more importantly, by a decline in productivity growth within those sectors, which increasingly account for the bulk of employment and economic activity. Figure 7 shows TFP growth within selected sectors averaged over 1990–99 and 2000–07. Relatively protected services sectors that are often closed to competition (OECD 2014), such as nonmarket, personal, and business services (which accounted for a growing majority of jobs), were the heaviest drags in terms of TFP growth for many countries (and strongly negative in Italy and Spain).

Figure 7.Total Factor Productivity Growth Rates by Countries and Selected Sectors

Sources: EU KLEMS database; World KLEMS database; and IMF staff calculations.

Note: ICTGS = information and communication goods and services.

18. ICT productivity growth on the wane? ICT has had a broad-based effect on aggregate TFP through its role as a general-purpose technology that fosters complementary innovations, including business reorganization in other services. As shown in Figure 7, average TFP growth in ICT goods and services (ICTGS) led all sectors during the 2000–07 period, particularly in the United States and the Nordic countries, while Japan, core Europe, and euro area countries with high borrowing spreads lagged. The role in spurring productivity growth in other sectors was even more significant, particularly in the United States. Indeed, slower ICT adoption and diffusion within the services sector in Europe (Finland and Sweden were exceptions) is widely cited as a key driver of the productivity gap between the United States and Europe, reflecting, in part, differences in flexibility of product markets (left-side panel of Figure 8). Yet, the positive effect of ICT on aggregate productivity growth may have already peaked. Figure 8 (right-side panel) shows a slowdown of TFP growth in IT-producing sectors in the United States in advance of the crisis (Fernald 2014). These trends have led some commentators to note that the productivity frontier is likely to return to slower growth rates last seen in the early 1990s (Gordon 2012).

Figure 8.Diffusion of Information and Communication Technologies in Services and Total Factor Productivity Growth in ICT-Producing Sectors

* Value-added weighted average of Distribution, Finance and Business, and Personal Services.

Note: Simple average for Core Europe excluding Sweden.

Sources: Fernald (2014); EU KLEMS database; World KLEMS database; Organization for Economic Cooperation and Development; and IMF staff calculations.

19. Distribution services rode the coattails of the ICT revolution. Distribution services (e.g., wholesale and retail trade and transportation), one of the largest sectors in terms of aggregate employment and output shares that benefited from innovations in ICT (e.g., in sales, inventories, and supply chains), recorded above-average TFP growth in most economies (Figure 8).9 However, the United States, among the leaders in the 1990s, and euro area countries with high borrowing spreads (e.g., Italy, Spain), saw a marked TFP growth slowdown in the sector on the eve of the crisis. The slowdown in the United States could reflect the fact that once retailing was reorganized to take advantage of faster information processing, the gains may have become more incremental.

20. Other “boom” sectors. While measured TFP growth can have a strong procyclical element—likely exacerbated in the upswing of financial cycles such as that before the global financial crisis—the performance of sectors commonly associated with the pre-crisis boom was mixed. For instance, finance was a high-productivity sector, with the United Kingdom and some European countries, in particular, registering strong TFP growth in the run-up to the crisis. Considering the events of the crisis, however, it remains to be seen how durable these gains will be. In contrast, TFP growth in the construction and real estate sectors disappointed across the board even prior to the crisis, particularly in the United States, Japan, and euro area countries with high borrowing spreads. Similarly, productivity growth in extractive industries slowed prior to the crisis, possibly as increasingly marginal and less productive resource extraction opportunities were pursued in response to booming commodity prices.

21. Large gaps in TFP levels across industries. The similarity among advanced economies in aggregate TFP levels masks dramatic variations across individual sectors.10 As suggested by Figure 9, even the most technologically advanced countries may have had TFP levels in some sectors that were more than half that of the leader. Indeed, the frontier economy varies from sector to sector. Germany, for instance, led in manufacturing, but lagged in services sectors. A similar story emerges for the United States, which represented the technology frontier in finance, business, and personal services. Those sectors accounted for the bulk of the aggregate productivity gap with the euro area (see also O’Mahony and Timmer 2009), but the United States lagged in non-ICT manufacturing.

Figure 9.Total Factor Productivity Levels

(Percent of frontier, 2000–07 average)

Sources: EU KLEMS database, GGDC Productivity Level database; and IMF staff calculations.

*These ICT subindustries are presented as deflators for other subindustries (e.g., IT and other information services) are not available across all countries. The box plot shows the minimum, first quartile, median, third quartile, and maximum TFP levels in each industry.

** Other production includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

22. Room for reforms. Natural advantages in some sectors (e.g., mining in Australia) and agglomeration spillovers in others (e.g., manufacturing in Germany) play a role in explaining the observed TFP gaps across countries. Measurement issues are also an important caveat, as errors due to industry misclassification, cross-country comparability of hours worked, and capital services used can be larger in the levels data. As such, differences in TFP levels reported in Figure 9 should be viewed with care. Still, the data suggest the existence of large gaps in sectoral TFP levels across countries prior to the crisis, pointing to room for policy and structural reforms to facilitate catch-up by improving resource allocation (allocative efficiency) and the use of productive inputs (productive efficiency). For countries already at, or close to, the production possibility frontier (e.g., the United States in personal services, the Netherlands in distribution), innovations in products and processes could help boost productivity.

III. Resource Allocation and Productivity

Significant policy-induced distortions prevented efficient allocation of factors of production across diverse industries and economic sectors even prior to the global financial crisis. Reducing these distortions and achieving more efficient resource allocation could have a substantial impact on boosting productivity.

23. Resource allocation. Aggregate TFP in an economy depends not only on the efficiency of individual firms or industries but also on how inputs are allocated across them. Indeed, economic theory suggests that more productive firms should have a greater incentive to innovate, and be more able to attract labor and capital inputs, relative to less efficient firms. Over time, less productive firms are forced to become more efficient or exit.11 This process brings about capital and labor reallocation, which shows up in measured TFP and output. Misallocation of resources, however, can arise if impediments exist to the movement of factors between heterogeneous firms and sectors. This can give rise to persistent rates-of-return differentials across firms and sectors, undermining aggregate TFP growth.

24. Impediments to efficient allocation. Impediments to the efficient allocation of capital and labor in advanced economies take various forms. Exemptions of certain producers from regulations or taxes, heavily-subsidized agricultural sectors, and “size-dependent policies” (e.g., size of retail establishments in Japan, and employment protection policies in Italy and France that only take effect beyond a certain size threshold) are distortions that induce a misallocation of factors. Labor and product market rigidities, market structure, and imperfect competition also influence factor reallocation, innovation, and firm restructuring.12 Examples of capital misallocation include financial frictions that increase the cost of capital, and distort investment decisions, including in technology-enhancing capital.13 These distortions lead firms to make inefficient capital and labor choices, thereby lowering aggregate TFP (Restuccia and Rogerson 2008; Hsieh and Klenow 2009).

25. Within-sector productivity. Substantial TFP gains can accrue from reducing resource misallocation and improving efficiency even within narrowly-defined industries of the economy. Indeed, industry and firm-level evidence confirms the presence of substantial misallocation of factor inputs as well as cross-country differences in establishment-level productivity and points to their potentially important role in generating large losses in aggregate productivity (Restuccia and Rogerson 2013; Hopenhayn 2014). For instance, Dias, Marques, and Richmond (2014) find that a better allocation of resources across firms within services industries in Portugal could boost measured TFP by about 40 percent. Similarly, Bartelsman, Haltiwanger, and Scarpetta (2013) find large effects on aggregate TFP from the presence of idiosyncratic distortions, predicting that output could be improved by as much as 15 percent in some European countries by improving the allocation of resources across manufacturing establishments.

26. Measuring misallocation across sectors. While within-industry misallocation accounts for the bulk of cross-country productivity differences, reducing misallocation across broad economic sectors can also spur aggregate productivity and output. Following Aoki (2012), and using data on sector-level value-added outputs, factor inputs, and input payments from the EU KLEMS database for 12 advanced economies over 2000–07, we assess the size of the potential labor and capital frictions (or equivalently “wedges”) that must be in place for the data to be an equilibrium outcome of the distorted economy (see the Technical Appendix for the theoretical model). We then undertake a counterfactual exercise to assess aggregate productivity and growth gains from eliminating these distortions (Box 1). Given stylized model assumptions, the estimates are indicative of what can be achieved from reducing distortions, rather than precise productivity payoffs.

Box 1.Estimating TFP Gains from Improved Across-Sector Resource Allocation

This exercise calculates the TFP gain that could accrue if each country were to adopt a better allocation of its productive inputs across sectors. In an environment without any frictions, the optimal allocation of factor inputs across productive sectors requires the equalization of marginal revenue. Deviations from this outcome represent a misallocation of resources and translate into lower aggregate productivity and output.

The quantitative exercise includes the following steps:

  • Building a static general equilibrium model with multiple sectors (see Technical Appendix), where each sector is subject to frictions. Our framework abstracts from specific frictions by assuming that the distortions ultimately affect the effective cost of capital and labor.
  • For each country, measuring the sectoral TFP for each year from the data. We also measure labor and capital frictions in each sector by estimating the cross-sector differences in marginal returns to inputs from the data. With these frictions, the equilibrium sectoral input allocation and output in our stylized model should exactly replicate the allocation and output measured in the data.
  • Fixing other elements of the model (technology, aggregate labor and capital stock, consumer preferences, etc.), calculating the counterfactual aggregate TFP in an economy with identical frictions in all sectors. This allows us to derive the potential TFP gain if the country were to adopt the (Pareto) efficient allocation.
  • Also considering changes only to capital (or labor) frictions. This allows us to identify the contributions of reallocating labor (capital) to the aggregate TFP gain.

One noteworthy issue arises from the measurement of labor inputs. The easiest way to measure labor input is to consider the total hours worked by employees without distinguishing workers’ skill levels (our baseline measure of labor inputs). However, this measure could be problematic, as high-skilled workers tend to provide more effective labor services than low-skilled workers. Therefore, we also constructed a labor index as a weighted sum of hours worked by different skill groups. We report the results of our counterfactual exercise for both the baseline (without labor skill heterogeneity) and the scenario with labor skill heterogeneity.

27. Sectors with the highest distortions. In most advanced countries, the agriculture sector is “favored,” with lower marginal returns to capital and labor than in other sectors. This can be seen in Figure 10, which indicates that the labor and capital wedges in the agriculture sector are positive in selected economies (i.e., the user cost of inputs is lower in the agriculture sector than in other sectors), particularly in Japan and Portugal. The underlying distortion stems from subsidies or other preferential policies favoring the agriculture sector. This pattern holds even after accounting for cross-sector heterogeneity in labor skill composition. Furthermore, labor distortions in the personal services sector are usually higher than in other sectors, indicating low marginal returns to labor in this sector. This could be reflective of lack of competition or labor market regulations, as policies, market failures, or location advantages favor this sector for reasons other than relative productivity (Duarte and Restuccia 2010).

Figure 10.Sectoral Distortions (Wedges) in Selected Countries

Sources: EU KLEMS database; World KLEMS database; IMF staff calculations.

*Average of Australia, Austria, Denmark, Finland, Germany, Italy, Japan, the Netherlands, Portugal, Spain, Sweden, United Kingdom, and United States. Higher capital (labor) wedges indicate smaller frictions for capital (labor) to move into the sector (i.e., lower marginal returns).

28. Cross-country heterogeneity. Apart from the “favored” agriculture and personal services sectors, the pattern of factor input misallocation varies across countries. For example, in Italy, a positive capital wedge in the business sector indicates an over-accumulation of capital, and low capital returns to the sector. In Portugal, low labor wedges in the financial and ICTGS sectors indicate significant barriers to entry and/or other frictions, which limit the size of employment in these sectors. In the United States, misallocation stems from capital distortions, particularly in mining.

29. TFP gains from eliminating distortions. The payoffs from improved factor allocation across sectors, while illustrative, are potentially large. Table 1 shows that a reallocation of capital and labor such that distortions are eliminated could substantially increase TFP and output.14 TFP gains average around 9 percent across all countries, ranging from 3.8 percent in Germany to nearly 20 percent in Italy (Column 1). If labor inputs are measured as skill-weighted hours worked, TFP gains are even larger (Column 2), indicating that labor misallocation includes not only the allocation of working hours but also skill mismatches in the labor force. For instance, accounting for labor skill composition, efficiency gains for the United States could be as high as 6 percent. This is lower than in other advanced economies, potentially reflecting a relatively more flexible labor market and lower skill mismatches than in other countries, but still sizable. Moreover, the relative contributions of labor and capital allocations vary across countries. For Austria and Portugal, the contribution from adjusting labor allocation is substantially higher than the gain from adjusting capital allocation alone. By contrast, the benefit from the adjustment of capital allocation is larger in Italy.15

Table 1.TFP Level Gain from Adjusting Input Allocation across Sectors(Percent)
Full Adjustment(1) Baseline (Pareto)

Adjust Labor

Allocation Only
Adjust Capital

Allocation Only
(2) With Labor Skill

Composition1
Austria9.96.33.310.2
Australia7.23.73.38.7
Denmark9.85.14.410.7
Finland10.95.25.417.9
Germany3.81.91.98.1
Italy19.54.514.336.4
Netherlands7.45.32.012.1
Portugal13.712.01.5
Sweden4.42.12.2
Spain6.23.22.911.0
United Kingdom5.33.12.112.9
Japan9.24.54.413.7
United States6.73.23.46.2
Sources: EU KLEMS database; World KLEMS database; and IMF staff calculations.Note: Calculated as TFP after counterfactual inputs allocation adjustment/”actual TFP” -1. Results are reported as averages for 2000–07 (or the latest year upon data availability) average. Country coverage reflects data availability.

Results are reported as 2000–05 average.

Sources: EU KLEMS database; World KLEMS database; and IMF staff calculations.Note: Calculated as TFP after counterfactual inputs allocation adjustment/”actual TFP” -1. Results are reported as averages for 2000–07 (or the latest year upon data availability) average. Country coverage reflects data availability.

Results are reported as 2000–05 average.

30. Growth implications. A better allocation of productive inputs could thus have sizable growth effects. Although input reallocation only explains a part of the measured TFP gap across countries, the potential impact on growth could still be significant. For example, assuming Italy and Portugal eliminate distortions (achieve the Pareto efficient allocation) within the next 10 years, their annual growth rates of TFP (and GDP) could potentially increase by 1.8 and 1.3 percent per annum, respectively, over the decade (Figure 11). While these potential gains appear substantial, it is useful to recall that the underlying distortions themselves are significant and that the exercise assumes that these are completely eliminated. Moreover, since this calculation only considers cross-sector reallocation of inputs (and not cross-industry or cross-firm level reallocation), our estimates only capture a lower bound of the overall gains from reducing misallocation.

Figure 11.Annual Total Factor Productivity Growth from Better Input Allocation

Source: EU KLEMS database; World KLEMS database; and IMF staff calculations.

Note: Assuming the adjustment to Pareto optimal allocation will take 10 years.

*Scenario with labor skill heterogeneity not available.

IV. What Reforms Will Enhance Productivity?

The examination of sector-level productivity developments and the distance of each country and sector from the “technology frontier” allows for identifying country-specific reform priorities. The analysis of how reform payoffs evolve over time also allows for assessing short-run costs to reforms, thus pointing to where compensatory measures may be needed to make reforms palatable.

31. Which reforms? There is an extensive literature documenting how policy and structural reforms can lead to improvements in resource allocation, productivity, and growth both in the near term and longer term (Nicoletti and Scarpetta 2003; Barkbu and others 2012; Mc Morrow, Werner, and Turrini 2010; Bourlès and others 2013; Bouis and others 2012). In addition to the economy-wide misallocation of resources discussed above, TFP growth of an economy can be influenced by various within-industry structural settings that constrain the allocation of factors and returns to innovation. The attempt here is not to provide an exhaustive list of reform priorities. Instead, the focus is to step away from one-size-fits-all policy messages and empirically explore the determinants of TFP growth, and to assess the dynamic effects of policy and structural reforms.

A. Drivers of Total Factor Productivity Growth

32. Conceptual framework. The conceptual framework of “distance to the technology frontier” (Aghion and Howitt 2006, 2009; Acemoglu, Aghion, and Zilibotti 2006) is employed to empirically assess the determinants of TFP growth across different industries and countries.16 Economic theory distinguishes between two types of innovation activity: adoption of existing vintage technologies (“imitation”) and introduction of new state-of-the-art technology (“innovation”). The key intuition is that countries that are farther away from the global technological frontier tend to grow mainly through technology adoption and imitation, whereas countries closer to the frontier rely more on innovation. Therefore, the set of policies aimed at sustaining productivity growth across industries and countries could vary depending on their locations vis-à-vis their technological frontiers.

33. Structural drivers of TFP growth. The focus here is on assessing how regulatory characteristics affect efficiency and convergence and other factors that drive expansion of the technological frontier and foster technology adoption, such as education, innovation, and ICT use. The degree of market flexibility, entry barriers, and competition in an economy is captured by the stringency of product market and employment protection regulations.17 As can be seen in Figure 12, rigidities in product markets vary across industries and countries, with services among the most heavily regulated industries in most countries, especially in Italy and Japan. The wide dispersion in services regulation across peers suggests a potential upside to reforms. An economy’s capacity for innovation and absorption/adoption of new technologies is captured by the intensity of use of high-skilled labor, ICT capital inputs, and research and development (R&D) spending. The United States appears to lead its peers in most sectors in innovation and the intensive-use of high-skilled labor (Figure 12). Moreover, as can be seen in Figure 12, unlike the manufacturing and ICT sectors, innovation in services does not stem from R&D spending. Instead, investment in tangibles (particularly automation and ICT) and nontangibles (e.g., business processes, organizational structure) are important sources of productivity growth in many services (Uppenberg and Strauss 2010).

Figure 12.Regulation, Skill, ICT, and R&D Indicators across Industries and Countries, 2007

Sources: EU KLEMS database; Indicators of Regulation Impact (OECD); ANBERD database (OECD); and IMF staff calculations.

1 High index value indicates more anti-competitive regulations in selected upstream nonmanufacturing sectors. The box plot shows the minimum, first quartile, median, third quartile, and maximum levels in each industry.

34. Catch-up convergence and technology spillovers. Tables 2 and 3 illustrate how changes in structural indicators have a material impact on TFP growth, also depending on the distance from the technological frontier.18 We find evidence that productivity growth across industries is driven by a catching-up process associated with the gradual adoption of newer technologies (Table 2). In particular, the pace of convergence in “follower” industries increases with the distance to the technological frontier (measured by the negative coefficient of the TFP gap). Furthermore, TFP growth across industries is also spurred by developments at the technological frontier (captured by the positive coefficient of TFP growth at the frontier), suggestive of significant productivity-enhancing knowledge spillovers from the technological leaders.

Table 2.Impact of Product Market Frictions on Total factor Productivity Growth
Dependent Variable: TFP Growth RateAll IndustriesManufacturingICT-related 1Services
(Percent, Annual)123456
TFP Growth Rate at the Frontier*************
TFP Gap w.r.t. the Frontier**************
Product Market Regulation****
Labor Market Regulation
Product Market Regulation X TFP Gap***
Labor Market Regulation X TFP Gap
Product Market Regulation X MFG**
Product Market Regulation X SVC******
Product Market Regulation X TFP Gap X MFG
Product Market Regulation X TFP Gap X SVC***
Number of observations4,6464,6464,6462,4241,6161,414
R20.20.20.20.20.30.2
Source: IMF staff calculations.Note: The symbols *, **, and *** represent panel-corrected robust standard errors at 10, 5, and 1 percent statistical significance. W.r.t. = with respect to.

Industries that produce ICT goods or employ ICT goods intensively.

Source: IMF staff calculations.Note: The symbols *, **, and *** represent panel-corrected robust standard errors at 10, 5, and 1 percent statistical significance. W.r.t. = with respect to.

Industries that produce ICT goods or employ ICT goods intensively.

Table 3.Impact of ICT, Human Capital, and R&D
Dependent Variable: TFP Growth RateAll IndustriesManufacturinICT-related 1Services
(Percent, Annual)12345678
TFP Growth Rate at the Frontier*******************
TFP Gap w.r.t. the Frontier*******************
ICT Capital*********
High-skilled Labor*************
R&D Expenditure***********
ICT Capital X TFP Gap****
*********
High-skilled Labor X TFP Gap*********
R&D Expenditure X TFP Gap***
Number of observations6,9236,4022,9502,6852,6851,707849487
R20.10.10.10.10.10.20.20.2
Source: IMF staff calculations.Note: The symbols *, **, and *** represent panel-corrected robust standard errors at 10, 5, and 1 percent statistical significance. W.r.t. = with respect to.

Industries that produce ICT goods or employ ICT goods intensively.

Source: IMF staff calculations.Note: The symbols *, **, and *** represent panel-corrected robust standard errors at 10, 5, and 1 percent statistical significance. W.r.t. = with respect to.

Industries that produce ICT goods or employ ICT goods intensively.

35. Product market reforms. Removing product market rigidities can lift productivity growth, particularly in the more heavily regulated services sectors (e.g., distribution, business, and personal services). Indeed, our empirical results suggest that higher product market regulation in the services sector is associated with lower TFP growth (Table 2). Moreover, the negative impact of stringent product market regulations on TFP growth is more severe the closer the industry is to the technology frontier. Differences in the flexibility of product markets and the business environment also explain observed differences in the uptake and diffusion of ICT between advanced economies (Collecia and Shreyer 2002). This suggests that increasing competition in the services sector by eliminating entry barriers, constraints to business operations, and red tape and administrative burdens would provide a needed productivity boost to services sectors. Indeed, the empirical results suggest that the productivity gains associated with these reforms could be significant. For example, if Italy were to reduce its product market regulation to the level in the Netherlands (e.g., the frontier economy in distribution services in Figure 9), the TFP growth gain within the services sector could amount to about 0.4 percent per year.

36. Labor market rigidities. As is common in many similar studies, the extent of economy-wide employment protection is not found to have significant explanatory power for industry-level TFP growth. Among other factors, this reflects the failure to capture the sheer diversity of institutions, underlying distortions, and misallocations in labor markets across countries and the fact that the effect of reforms may take time to materialize (as seen below). But in general, improving labor market institutions can have beneficial effects, with some evidence suggesting that mandatory dismissal regulations have a depressing impact on productivity growth in industries where layoff restrictions are more likely to be binding (Bassanini, Nunziata, and Venn 2009). Other studies found that longer and more generous unemployment benefits, high labor tax wedges, and collective bargaining systems that are more favorable to wages than employment affect employment negatively (Nickell, Nunziata, and Ochel 2005; Bassanini and Duval 2009). However, studies do not always agree on the role of specific institutions, and cross-country evidence on their productivity impact is less clear-cut (OECD 2007). In general, firm-level evidence points to stronger productivity effects, with less stringent labor market institutions and a more efficient use of human capital (e.g., reducing labor skills mismatches) facilitating the movement of labor to more productive firms, and fostering creative destruction (firm entry and exit) (Henrekson 2014).

37. Productivity impact of knowledge capital and innovation. More intensive use of high-skilled labor and ICT capital inputs and higher spending on R&D activities is associated with higher productivity growth (Table 3). The channels through which these factors affect TFP growth, however, differ. A highly skilled labor force can foster the development of skill-intensive industries and the diffusion of new technologies. For example, the results in Table 3 suggest that increasing the share of high-skilled workers in the United Kingdom to the level of Japan could boost productivity growth by an additional 0.7 percent per year. More intensive and widespread use of ICT can support more efficient organizational structures and business processes within firms, leading to efficiency gains. Greater R&D spending creates new efficiency-enhancing technologies. These results clearly support an important public policy role for improving educational attainment, fostering ICT diffusion, and encouraging innovation in advanced economies.

38. Varying sectoral effects. Investments in human capital and R&D can play a significant role in spurring TFP growth in ICT-related industries (and services in the case of skilled labor), with the strongest productivity effects for countries closer to the frontier.19 Trade and technology spillovers imply that reforms in leading economies not only have large productivity effects within these countries, but also generate important positive spillovers. Within the services sector, however, the impact on TFP growth of more intensive use of ICT is highest for countries with lagging productivity in the sector. This suggests an important role for fostering greater ICT investment and diffusion in services, including by eliminating barriers to trade and foreign direct investment, particularly in countries that have important productivity gaps with respect to peers.20

39. Bringing it together. What is the economic significance of these results? Using the TFP level estimates in Figure 9, Table 4 presents the potential TFP growth gain associated with improving each structural variable from the 25th percentile to the best-practice level. For the median country, improvements in structural variables could generate additional TFP growth gains of ¼ to ½ a percentage point annually across all industries. Moreover, the table suggests that the sectors through which these improvements affect TFP growth vary. In the case of product market regulations, TFP gains accrue largely within the services sector, while for R&D spending the largest gains are seen in the ICT sector. Better human capital (labor skills) and more intensive use of ICT appear to have broader productivity-enhancing effects, indicating their relative importance in driving aggregate TFP growth.

Table 4.TFP Growth Gains from Improvement in Structural Determinants(Percentage points, annually)
SectorManufacturingServicesICTAll
Product Market Regulations0.10.1
Labor Skill1.41.20.4
ICT Capital0.70.70.3
R&D Expenditure1.90.2
< 0.5< 1> =1
Source: IMF staff calculations.Note: For each indicator, an “improvement” is defined as closing the gap between the highest-ranked country and the 25th percentile country. In each sector, we assume the median TFP level of sample countries over 2000–07.
Source: IMF staff calculations.Note: For each indicator, an “improvement” is defined as closing the gap between the highest-ranked country and the 25th percentile country. In each sector, we assume the median TFP level of sample countries over 2000–07.

B. Dynamic Effects of Reforms on Productivity

40. Reform payoffs evolve over time. The preceding section considered how institutional rigidities and investments in knowledge capital and innovation are associated with productivity performance. As reforms could result in adjustment costs that potentially vary across sectors, it is useful to assess their dynamic (short- and medium-term) impact. Moreover, the overall productivity gains are likely to depend on the magnitude of reforms undertaken. As a result, this section focuses on large changes in structural indicators and examines their impact on productivity, employment, and output over various time horizons (see Technical Appendix for details).21 We also investigate how the productivity impact depends on the distance from the technological frontier, initial settings, and prevailing economic conditions.

41. Identifying reform “shocks.” For the purposes of better identifying cause and effect, it is useful to look at “shocks” to various structural indicators. Large reforms episodes are identified as a significant improvement (in absolute terms) of a structural indicator above a certain threshold (Bouis and others 2012).22 Reforms are identified using the same set of variables described in the previous section with two additions. First, we include a measure of infrastructure, as improvements can have a positive impact on productivity over a longer horizon that goes beyond the effect of increases in the industry-specific capital stock. This is a result of economies of scale, the existence of network externalities, and competition-enhancing mechanisms. Second, we include a measure of labor tax wedges as an additional measure of labor market rigidities, as these can create disincentives to work in some segments of the labor market and affect the direction and pace of technological change.23

42. Shock episodes. Figure 13 presents the number of identified reforms per country. Large infrastructure shocks that took place mainly in the 1970s have since waned, reflecting in part the declining stock of public capital as a share of output in advanced economies (IMF 2014). Product market reforms saw a spurt in the 1990s and continued at a similar pace in the 2000s (Duval 2008). However, labor market reforms, as measured by a reduction in job protection or a reduction in the labor tax wedge, have been less frequent. The frequency of positive ICT “shocks” increased since the 1970s but slowed on the eve of the crisis, as did human capital and R&D shocks.

Figure 13.Reform Episodes per Country

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: Number of reforms has been adjusted by the number of years.

43. Productivity impact typically grows over time. The effects of reforms on productivity growth over the short and medium term are generally positive (Figure 14), consistent with our findings above. Product market reforms, innovation, and infrastructure shocks have a positive productivity impact that increases over time. For instance, a 10 percent shock in R&D spending boosts TFP across all industries by 0.7 percent after three years and 2 percent over five years. In contrast with our previous results, we find that a reduction in employment protection is associated with a decline in TFP after three years, potentially reflecting weaker incentives for investment in firm-specific skills through excessive job turnover.24 A reduction in the labor tax wedge, however, is associated with higher productivity three years after the reform, suggesting that the efficiency impact of different labor market reforms can vary, rendering a one-size-fits-all reform recipe unsuitable. Hence, as stressed in IMF (2013), policies should attempt to stay on the “plateau” and avoid the “cliffs” of either excessive regulations or extreme disregard for labor conditions.

Figure 14.Short- and Medium-Term Impact of Structural Reforms on Total Factor Productivity

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show the cumulative three-year (ST) and five-year (MT) level gains in TFP. Brown and shades of red indicate a negative and statistically significant impact; shades of blue indicate a positive and statistically significant one. For instance, an ICT capital shock leads to a cumulative TFP level increase in ICT-related industries of about 30 percent after five years (i.e., two standard deviations of the average cumulative five-year change in TFP in the sample). “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

44. The impact of reforms varies across industries and countries. We confirm many of the insights from the previous section, including those regarding gains from product market reforms, innovation, and knowledge capital. We also confirm that effects tend to be both sector- and country-specific (in terms of distance to the frontier). Productivity gains from product market liberalization over the medium term are highest in ICT and personal services, sectors where many European countries and Japan have mostly lagged behind the United States. Lower tax wedges, higher R&D spending, and better human capital boost productivity growth in innovation-intensive manufacturing and ICT sectors. Reforms can also have short-term negative productivity effects (e.g., a negative impact of product market regulation on ICT and personal services, and of labor tax wedges on personal services), reflecting varying adjustment costs across sectors during the reform process (Blanchard and Giavazzi 2003; Gomes and others 2013). Importantly, we find a positive productivity impact from infrastructure for all countries, irrespective of their distance to the frontier.25

45. Pre-reform policy and structural settings. The impact of reforms could differ depending on initial policy and structural settings. To assess this, we examined the productivity effects in industries with low (below the 25th percentile of the distribution), medium, and high (above the 75th percentile) initial policy and structural settings (Figure 15). We find that the largest productivity gains accrue from reducing product market regulations and labor tax wedges in highly regulated services sectors, while the impact is not statistically significant for lightly regulated industries, a result consistent with previous findings (Bourlès and others 2013). Moreover, the medium-term productivity impact of knowledge capital, innovation, and infrastructure, on average, is highest for industries with higher initial levels of innovation and infrastructure (e.g., manufacturing, ICT). Some differences, however, can be gleaned across industries, especially in ICT and personal services, where, the lower the initial levels of R&D spending and ICT capital use, the higher the productivity gains. This again points to the policy role of fostering greater diffusion of ICT services, particularly in economies where productivity growth has been lagging (e.g., many European countries).

Figure 15.Medium-Term Impact of Reforms on Total Productivity Growth: Conditional on Initial Settings

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: This reports the average cumulative five-year impact of a shock on TFP levels as a function of the initial level of the structural indicator.

46. Reform impact across the business cycle. The impact of reforms could also depend on the business cycle position, in particular on how changes in institutional and structural settings affect supply and demand. For instance, weak demand and excess capacity conditions risk limiting the short-term output response to reforms. At present, balance sheet concerns and low confidence weigh on private sector decisions, including technology-enhancing investment choices. To assess the influence of overall economic conditions, we examined reform impacts during upturns and downturns.26 The results, reported in Figure 16, confirm that infrastructure shocks are associated with significantly higher productivity gains during downturns (IMF 2014b), suggesting an important policy role for infrastructure investment in spurring productivity. In contrast, productivity gains from product market liberalization and, not surprisingly, innovation shocks tend to be higher during upturns. However, as seen below, product market deregulation could still be a useful tool for propelling growth. Liberalizing entry into regulated sectors, for instance, can be a significant source of investment and job creation (Figures 17 and 18).

Figure 16.Impact of Reform Shocks on Total Factor Productivity across the Business Cycles

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show the average cumulative five-year gains in TFP levels as a function of overall economic conditions.

Figure 17.Impact of Reform Shocks on Sectoral Employment

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show the cumulative three-year (ST) and five-year (MT) level gains in employment. Brown and shades of red indicate a negative statistically significant impact of the reform; shades of, blue indicate a statistically significant positive one. “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

Figure 18.Impact of Reform Shocks on Sectoral Output

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show the cumulative three-year (ST) and five-year (MT) level gains in employment. Brown and shades of red indicate a negative statistically significant impact of the reform; shades of, blue indicate a statistically significant positive one. “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

C. Dynamic Effects of Reforms on Employment and Output

47. Short term costs of reforms? Reforms could have a small, and in some cases even negative, short-term effect on output and employment because of costly reallocation of labor and capital and firm restructuring, with a temporary rise in unemployment and potentially high fiscal and social costs. But any potential short-term drags from these factors could be offset even in the near term to the extent that reforms enhance credibility and confidence. Fiscal policy can also play a facilitating role by bringing forward employment gains and overcoming resistance from stakeholders who risk short-term losses (IMF 2014c). Nonetheless, while the focus is on the medium-term productivity impact, it is useful to consider short term trade-offs, especially in the current context of weak recoveries in many advanced economies. To shed more light on this issue, we extend the dynamic analysis of reforms on sectoral productivity to sectoral employment and output (value added).

48. Impact on employment. The short- and medium-term employment impact of reforms is mostly positive (Figure 17), but can vary across type of reform and by sector. Overall, product market liberalization and infrastructure shocks have a positive and persistent impact on employment. Intensive use of skilled labor has a positive effect on employment, particularly in manufacturing and ICT industries, but is associated with lower employment in other production sectors (e.g., agriculture, utilities, and mining). Labor market liberalization boosts employment in personal services in the short term, reflecting the fact that lower dismissal costs increase employment because firms can hire more freely. ICT shocks are associated with lower employment in manufacturing and personal service sectors, and R&D shocks in distribution services, likely reflecting the labor displacement effects of increasing automation.

49. Impact on output. Output gains from reforms are large and materialize predominantly in the medium term (Figure 18). In particular, product market liberalization and more intensive use of high-skilled labor raise output across all industries by about 2 percent over the medium term. Effects of infrastructure and innovation shocks are even larger, increasing output across all industries by around 3 and 5 percent, respectively. Labor market liberalization, as captured by lower employment protection, has a negative aggregate short-term impact on output (though not on employment—see above), similar to its effect on TFP. Some differences are also discernible across industries, but the overall productivity impact across all industries is largely significant. Infrastructure shocks have a positive short- and medium-term output impact in almost all sectors, highlighting the role of infrastructure investment in supporting productivity, employment, and output.

50. No one size fits all. Productivity and growth effects of reforms depend on country-specific policy and institutional settings, suggesting the need to prioritize reforms according to individual conditions. Given the nature of statistical relationships, the empirical results should be interpreted as highlighting associations rather than revealing causation. Moreover, they are illustrative of the type of productivity-enhancing reforms that would be effective, and are not intended to suggest that these specific reforms should be implemented by all countries. For instance, the nature of product market rigidities varies across countries, reflecting outright entry barriers (e.g., in business services in Europe) in some countries or monopolistic structures (e.g., network industries) in others. Labor market institutions also vary considerably across advanced economies, suggesting that reforms should take into account different work models. Political economy considerations (e.g., how social consequences of reforms are dealt with) and implementation costs critically influence reform design strategies and effectiveness. Complementarities (e.g., between product and labor market reforms) and sequencing of reforms in different areas are also material for the realization of potential gains. These factors make it difficult to identify the causal effects of reforms on productivity and growth, suggesting that there is neither a single reform path nor a silver bullet.

V. Final Remarks

51. Headwinds. This Staff Discussion Note has documented the secular headwinds that advanced economy policymakers face as a result of slowing pre-crisis trend growth in output and productivity. These headwinds reflect structural changes that continue to unfold as economies mature and relative factor costs adjust, with implications for the mix of inputs and structure of production. However, slowing innovation and human and physical capital accumulation, alongside policy distortions, also contribute to the challenges faced.

52. Back to basics. To increase growth potential, advanced economies must boost the productive capacity of their workforces, make needed investments in physical and IT infrastructure, and encourage innovation. Removing binding product and labor market rigidities that impede resource allocation toward the most dynamic sectors and firms is crucial to boosting productivity. So is augmenting human capital, including reducing skills gaps and improving the quality of education. The case for refocusing decisively on these fundamental supply drivers is even more urgent in an environment in which many economies are reaching the limits of the policy support they can provide to shore up demand. Moreover, even as some of the crisis-related effects are expected to fade, demographic headwinds loom, further strengthening the case for enhancing productivity.

53. Investment in knowledge capital and innovation is critical. Innovation and investments in equipment, ICT, and new ideas are vital for raising long-term trend growth. Investing in capital allows firms to adopt new technologies and can be an important part of their strategies to reorganize production processes toward global best practices. Fostering a supportive environment for investment in innovation and technology adoption is central to having a dynamic and productive economy. Competition can increase the supply of new ideas and push frontier firms to innovate and produce new technology. Well-functioning markets diffuse this technology, prompting its adoption more widely, and resources can be quickly reallocated from failing to growing firms so that the economy can reap the returns of innovation and entrepreneurs have the incentives to undertake risky projects. Since there are often market failures in innovation, the government also has a role in providing incentives for R&D and for creating wider policies to support innovation.

54. Tailoring to economic conditions. Where demand is weak and there is slack in the labor market, and where investment efficiency is high, product market reforms and increased investment in productive infrastructure could be prioritized to reap immediate employment and output benefits. More generally, well-designed structural reforms that minimize potential short-run contractionary effects could translate into rising output and employment.

55. Some other key takeaways include:

  • Services reform is a priority. This is the most important sector—but also the most highly regulated one—across advanced economies. The variance in performance and regulations across peers suggests significant upside scope from determined efforts, especially from product market reforms, to increase competition. The potential to boost service productivity through investment in information technology in lagging regions is also sizable. Labor market policies should attempt to avoid either excessive regulations or extreme disregard for labor conditions.
  • Spillovers can be large. Efforts to enhance performance by technology leaders (e.g., Germany in manufacturing, the United States in ICT services) are global goods that spill over to support productivity gains in other countries. There is a lot of low hanging fruit in this area. Increasing returns to scale are evident in knowledge capital and create many opportunities for countries to boost their own and global prospects. This will support countries that are further from the frontier, but these are just the countries that need to work doubly hard to remove impediments to productivity growth.
  • Infrastructure investment has clear benefits. Investments in transport, energy, and telecommunications are essential to improving productivity, boosting employment, and increasing output. Because they tend to be large-scale and long-term, these investments require high levels of coordination to maximize the wider benefits to society as a whole. In many cases, governments will need to play a vital role in planning, delivering, and financing these projects. At present, debt-financed projects could have large productivity and output effects, particularly if clearly identified infrastructure needs are met through efficient investment (IMF 2014b).
  • Improving resource allocation across sectors. Reducing policy distortions that affect resource allocation across broad economic sectors could yield significant productivity payoffs. In this regard, rationalizing large agricultural subsidies and increasing competition in personal services could be important, as they create a misallocation of productive resources to relatively low-value activities.
  • Calibrating reforms. The productivity and growth effects of reforms depend on a variety of country-specific policy and institutional settings. While some policies are applicable across all advanced economies, granularity is necessary, as there is no one-size-fits-all strategy. The divergence in productivity performance and the underlying drivers across sectors suggest that reform priorities need to be appropriately calibrated.
Annex 1. Stochastic Frontier Analysis

We use stochastic frontier analysis (SFA) to illustrate the patterns of total factor productivity (TFP) growth across countries at the sector level. The basic idea of SFA in our context is that there is a world frontier technology for each given industry or sector, which represents the most productive technology within that industry or sector. The TFP of industry (or sector) i in country c depends on the world frontier TFP level and how efficient the country can make use of the frontier technology. We choose a model specification close to Pitt and Lee (1981), that is,

TFPc,ti is the TFP level of industry or sector in country c at time t. αi + βit represents the world frontier productivity level, which is determined by a constant αi and a trend component βit. vc,ti is an i.i.d white noise. The term uci represents the inefficiency level, which is the distance of country to the world frontier technology. See Belotti and others (2012) for alternative model specifications. Most specifications differ either in the main equation (1) or the distribution of the inefficiency term.

The TFP level estimates used for the calculation were obtained by linking the TFP growth estimates from the EU KLEMS database with the Groningen Growth and Development Centre) Productivity-Level database, which provides the PPP-adjusted industry TFP levels in 1997 for our sample countries (Inklaar and Timmer 2008). We constructed productivity levels series for 23 sectors spanning the period from 1970–2007.

We use the maximum likelihood method to estimate the model specified by (1) and (2) on 10-year rolling window sectoral TFP data. For each sector i, we report the trend of frontier TFP growth (βi) and the average TFP level in percentage of world frontier (cross-sectional mean of (1euci)).

References

    Acemoglu, D., P.Aghion, and F.Zilibotti. 2006. “Distance to Frontier, Selection, and Economic Growth.” Journal of the European Economic Association4(1): 37–74.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Aghion, P., and P.Howitt. 2006. “Joseph Schumpeter Lecture—Appropriate Growth Policy: A Unifying Framework.” Journal of the European Economic Association4(2-3): 269–314.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Aghion, P., and P.Howitt. 2009. The Economics of Growth. Cambridge, Massachusetts: MIT Press.

    Aoki, Shuhei. 2012. “A Simple Accounting Framework for the Effect of Resource Misallocation on Aggregate Productivity.” Journal of the Japanese and International Economies26(4): 473–94.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Arnold, J., G.Nicoletti, and S.Scarpetta. 2008. “Regulation, Allocative Efficiency and Productivity in OECD Countries: Industry and Firm-level Evidence.” OECD Economics Department Working Paper No. 616, Organization for Economic Cooperation and Development, Paris.

    • Search Google Scholar
    • Export Citation

    Baily, M. N., C.Hulten, and D.Campbell. 1992. “Productivity Dynamics in Manufacturing Plants.” Brookings Papers on Economic Activity. Washington, Brookings Institution.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Barkbu, B., J.Rahman, R.Valdés, and others. 2012. “Fostering Growth in Europe Now.” IMF Staff Discussion Note SDN/12/07, International Monetary Fund, Washington.

    • Search Google Scholar
    • Export Citation

    Barnett, A., A.Chiu, J.Franklin, and M.Sebastia-Barriel. 2014. “The Productivity Puzzle: A Firm-Level Investigation into Employment Behavior and Resource Allocation over the Crisis.” Bank of England Working Paper 495.

    • Search Google Scholar
    • Export Citation

    Bartelsman, E., J.Haltiwanger, and S.Scarpetta. 2013. “Cross-Country Differences in Productivity: The Role of Allocation and Selection.” American Economic Review103(1): 305–34.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Bassanini, A., L.Nunziata, and D.Venn. 2009. “Job Protection Legislation and Productivity Growth in OECD Countries.” Economic Policy24: 349–402.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Baumol, W.J., S A.B.Blackman, and E.N.Wolff. 1985. “Unbalanced Growth Revisited: Asymptotic Stagnancy and New Evidence.” American Economic Review75: 806–17.

    • Search Google Scholar
    • Export Citation

    Belotti, F., S.Daidone, G.Ilardi, and V.Atella. 2012. “Stochastic Frontier Analysis Using Stata.” The Stata Journal13(4): 719–58.

    Blanchard, O,2004. “The Economic Future of Europe.” Journal of Economic Perspectives18(4): 3–26.

    Blanchard, O., and F.Giavazzi. 2003. “Macroeconomic Effects of Regulation and Deregulation in Goods and Labor Markets.” The Quarterly Journal of Economics118(3): 879–907.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Blanchard, O., F.Jaumotte, and P.Loungani. 2013. “Labor Market Policies and IMF Advice in Advanced Economies during the Great Recession.” Staff Discussion Note 13/02, International Monetary Fund, Washington.

    • Search Google Scholar
    • Export Citation

    Boius, R., O.Causa, L.Demmou, R.Duval, and A.Zdzienicka. 2012. “The Short-Term Effects of Structural Reforms: An Empirical Analysis.” OECD Economics Department Working Paper 949, Organization for Economic Cooperation and Development, Paris.

    • Search Google Scholar
    • Export Citation

    Bourlès, R., G.Cette, J.Lopez. J.Mairesse, and G.Nicoletti. 2013. “Do Product Market Regulations in Upstream Sectors Curb Productivity Growth? Panel Data Evidence for OECD Countries.” The Review of Economics and Statistics95(5): 1750–768.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Collecia, A., and P.Shreyer. 2002. “ICT Investment and Economic Growth in the 1990s: Is the United States a Unique Case? A Comparative Study of Nine OECD Countries.” Review of Economic Dynamics5: 408–42.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Dabla-Norris, E., and others. 2013. “Anchoring Growth: The Importance of Productivity-Enhancing Reforms in Emerging Market and Developing Economies.” Staff Discussion Note 13/08, International Monetary Fund, Washington.

    • Search Google Scholar
    • Export Citation

    Dias, D., C.Robalo Marques, and C.Richmond. 2014. “Misallocation and Productivity in the Lead Up to the Eurozone Crisis.” Banco de Portugal Working Paper.

    • Search Google Scholar
    • Export Citation

    Duarte, M., and D.Restuccia. 2010. “The Role of the Structural Transformation in Aggregate Productivity.” Quarterly Journal of Economics125(1): 129–73.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Duval, R.2008. “Is There a Role for Macroeconomic Policy in Fostering Structural Reforms? Panel Evidence from OECD Countries over the Past Two Decades.” European Journal of Political Economy24(2): 491–502.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Fernald, J.2014. “Productivity and Potential Output Before, During, and After the Great Recession.” NBER Macroeconomics Annual 2014, Vol. 29. Chicago: University of Chicago Press.

    • Search Google Scholar
    • Export Citation

    Fernald, J., and C. I.Jones. 2014. “The Future of US Economic Growth.” American Economic Review104(5): 44–49.

    Foster, L., J.Haltiwanger, and C.Syverson. 2008. “Reallocation, Firm Turnover, and Efficiency: Selection on Productivity or Profitability.” American Economic Review98(1): 394–425.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Garicano, L., C.Lelarge, and J.Van Reenen. 2013. “Firm Size Distortions and the Productivity Distribution: Evidence from France.” NBER Working Paper 18841, National Bureau of Economic Research, Cambridge, Massachusetts.

    • Search Google Scholar
    • Export Citation

    Gilchrist, S., J. W.Sim, and E.Zakrajsek. 2013. “Misallocation and Financial Market Frictions: Some Direct Evidence from the Dispersion in Borrowing Costs.” Review of Economic Dynamics16(1): 159–76.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Gomes, S., P.Jacquinot, M.Mohr, and M.Pisani. 2013. “Structural Reforms and Macroeconomic Performance in the Euro Area Countries: A Model-Based Assessment.” International Finance16: 23–44.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Gordon, R. J.2012. “Is US Economic Growth Over? Faltering Innovation Confronts the Six Headwinds.” NBER Working Paper 18315, National Bureau of Economic Research, Cambridge, Massachusetts.

    • Search Google Scholar
    • Export Citation

    Henrekson, M.2014. “How Labor Market Institutions Affects Job Creation and Productivity Growth.” IZA World of Labor38.

    Herrendorf, B., R.Rogerson, and A.Valentinyi. 2013. “Growth and Structural Transformation.” NBER Working Paper 18996, National Bureau of Economic Research, Cambridge, Massachusetts.

    • Search Google Scholar
    • Export Citation

    Hopenhayn, H.2014. “On the Measure of Distortions.” NBER Working Paper 20404, National Bureau of Economic Research, Cambridge, Massachusetts.

    • Search Google Scholar
    • Export Citation

    Hopenhayn, H., and R.Rogerson. 1993. “Job Turnover and Policy Evaluation: A General Equilibrium Analysis.” Journal of Political Economy101(5): 915–38.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Hsieh, C.-T., and P.Klenow. 2009. “Misallocation and Manufacturing TFP in India and China.” Quarterly Journal of Economics124(4): 1403–448.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Inklaar, R., and M.Timmer. 2008. “GGDC Productivity Level Database: International Comparisons of Output, Inputs and Productivity at the Industry Level.” Paper No. GD-104, Groningen Growth and Development Centre, University of Groningen.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2013. “Jobs and Growth: Analytical and Operational Considerations for the Fund.” International Monetary Fund, Washington, 14March. http://www.imf.org/external/np/pp/eng/2013/031413.pdf

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2014a. The Managing Director’s Global Policy Agenda to the International and Finance Committee. International Monetary Fund, Washington. http://www.imf.org/external/np/pp/eng/2014/041214.pdf

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2014b. World Economic Outlook. Washington, October.

    International Monetary Fund (IMF). 2014c. Fiscal MonitorWashington, October.

    International Monetary Fund (IMF). Forthcoming. “World Economic Outlook. Washington, April2015.

    Jorgenson, D., and M.P.Timmer. 2011. “Structural Change in Advanced Nations: A New Set of Stylised Facts.” The Scandinavian Journal of Economics113(1): 1–29.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Jorgenson, D., M.Ho, and K. J.Stiroh. 2008. “A Retrospective Look at the U.S. Productivity Growth Resurgence.” Journal of Economic Perspectives22(1): 3–24.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Lagos, R.2006. “A Model of TFP.” Review of Economic Studies73: 983–1007.

    Lewis, W.2005. The Power of Productivity: Wealth, Poverty, and the Threat to Global Stability. Chicago: University of Chicago Press.

    Mc Morrow, K., R.Werner, and A.Turrini. 2010. “Determinants of TFP Growth: A Close Look at Industries Driving the EU-US TFP Gap.” Structural Change and Economic Dynamics21(3): 165–80.

    • Search Google Scholar
    • Export Citation

    Ngai, L.R., and R.M.Samaniego. 2011. “Accounting for Research and Productivity Growth across Industries.” Review of Economic Dynamics14(3): 475–95.

    • Crossref
    • Search Google Scholar
    • Export Citation

    NickellS., L.Nunziata, and W.Ochel. 2005. “Unemployment in the OECD since the 1960s: What Do We Know?” Economic Journal115: 1–27.

    Nicoletti, G., and S.Scarpetta. 2003. “Regulation, Productivity and Growth: OECD Evidence.” Economic Policy18(36): 9–72.

    O’Mahony, M., and M.P.Timmer. 2009. “Output, Input and Productivity Measures at the Industry Level: The EU Klems Database.” The Economic Journal119(538): F374–F403.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Organization for Economic Cooperation and Development (OECD). 2001. The New Economy: Beyond the Hype. Paris: Organization for Economic Cooperation and Development.

    • Search Google Scholar
    • Export Citation

    Organization for Economic Cooperation and Development (OECD). 2007. OECD Employment Outlook. Paris: Organization for Economic Cooperation and Development.

    • Search Google Scholar
    • Export Citation

    Organization for Economic Cooperation and Development (OECD). 2014. Economic Policy Reforms 2014: Going for Growth. Paris: Organization for Economic Cooperation and Development.

    • Search Google Scholar
    • Export Citation

    Restuccia, D., and R.Rogerson. 2008. “Policy Distortions and Aggregate Productivity with Heterogeneous Establishments.” Review of Economic Dynamics11(4): 707–20.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Restuccia, D., and R.Rogerson. 2013. “Misallocation and Productivity.” Review of Economic Dynamics16(1): 1–10.

    Syverson, C.2011. “What Determines Productivity?” Journal of Economic Literature49(2): 326–65.

    Uppenberg, K., and H.Strauss. 2010. “Innovation and Productivity Growth in the EU Services Sector.” European Investment Bank.

    van Ark, B., M.O’Mahony, and M. P.Timmer. 2008. “The Productivity Gap Between Europe and The United States: Trends and Causes.” The Journal of Economic Perspectives22(1): 25–44.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Zheng, S., and H.Bloch. 2014. “Australia’s Mining Productivity Decline: Implications for MFP Measurement.” Journal of Productivity Analysis41(2): 201–12.

    • Crossref
    • Search Google Scholar
    • Export Citation
The New Normal: A Sector-Level Perspective on Productivity Trends in Advanced Economies—Technical Appendix

Prepared by Era Dabla-Norris, Si Guo, Minsuk Kim, and Aleksandra Zdzienicka1

1. This Technical Appendix describes the theoretical model underlying the counterfactual exercise that assesses potential total factor productivity (TFP) gains from reducing resource misallocation across sectors (Section I). Section II outlines the conceptual framework and data sources used for the empirical analysis and presents detailed findings summarized in the Staff Discussion Note. We first describe the empirical methodology to examine the drivers of industry-level growth, before turning to the dynamic effects of structural reforms on sectoral TFP (Section III). The analysis also investigates the dynamic impact of structural reforms on sectoral employment and output.

1. Theoretical Framework for Assessing Misallocation

2. The model is a revised version of Aoki (2012). We assume that there are I (i=1,2,…I) sectors in a closed economy. Each sector i produces one kind of good and the price is pi. Residents consume a basket of I goods and sector shares depends on prices pi and consumer preferences. Production requires capital and labor as inputs. The aggregate stock of capital and labor is fixed, which implies that the measured aggregate TFP is proportional to output in this economy. There is no capital depreciation or investment, and therefore output equals consumption. There are N kinds of intermediate goods, and one final good. Production of the final good is given by:

where γ > 0 is the elasticity of substitution2 and θi > 0 represents the weight in the production function. Each intermediate goods sector i has a representative firm i with the production function:

Each sector is competitive, therefore firm i takes output price pi and input prices w, r as given. Firm i‘s maximization problem can then be written as

where τK,i and τL,i are capital and labor frictions that influence the effective rental cost of capital and labor in sector i. The market clearing conditions are Σi Ki = K, ΣiLi = L.

3. Firm’s optimization problem. Following Aoki (2012), from firms’ optimality conditions we can derive

where MRK (or MRL) represents the marginal revenue of one more unit of capital (or labor). Using the market clear conditions, we can derive

where si is the valued-added share of sector i and α¯=Σiαisi is the weighted average capital share.

4. Calculating wedges (distortions). Defining MRK¯=1/[Σjαjsjα¯1MRKj] as the average marginal revenue to capital (and defining MRL similarly), we can calculate capital wedges from the observed sectoral capital and value added data:

In the Pareto optimal allocation, capital (or labor) frictions should be identical across sectors (i.e., capital and labor wedges are zero). However, deviating from a Pareto optimal allocation, if a sector i is relatively more “favored” in capital markets (i.e., τi is relatively smaller compared with other sectors), more capital will be allocated to sector i and the capital wedge of sector i will be positive.

5. Data and Parameterization. We measure Li, Ki, σi, and αi from the EUKLEMS dataset. For labor, we first use “total hours worked by persons engaged” in sector as the relevant measure of labor input Li. One drawback with using total hours as the measure of labor input is that it puts an even weight on skilled and unskilled labor. Therefore, we also construct an alternate index, which puts a higher weight on high- and medium-skilled labor. The weights of high- and medium-skilled labor are proportional to their hourly wages in 1995. For capital stock, we use “real fixed capital stock, 1995 prices” of “all assets” in sector i as the measure of Ki. For capital share, we use “capital compensation” as a share of total value added. The remaining parameters to be estimated are θt (the weight of sector in final goods production). The first order condition from the maximization problem of the firm that produces the final good is:

This implies

Using data on value added (pY and piYi) and price indices (p and pi), we can calculate θi. Note that we do not have to specify the unit of the price and volume index for each sector. This is because the variable of interest is the real value-added volume Y and its changes in various counterfactual exercises. Changing the units of sectoral prices and volumes will change the measured θi but will not change the aggregate value-added volume Y. To see this, suppose we change of a unit of output of sector i by m times. That is, Y’i = Yi/m and pi = mpi. The measured weight of sector i in the final good production θi=m11γθi and the aggregate value-added Y calculated from (1) remains unchanged.

6. Counterfactual exercise. Suppose that capital and labor frictions are changed to tK,i and tL,i. We can then calculate output under this counterfactual set of frictions and examine differences with output observed in the data. Notice that if tK,i=tk¯ and tL,i=tL¯, there is no misallocation across sectors.

II. Drivers of Productivity Growth

7. Conceptual framework. The analysis in this section follows the conceptual framework of the neo-Schumpeterian growth theory as put forward in Aghion and Howitt (2006, 2009) and Acemoglu, Zilibotti, and Aghion (2006). Economic theory distinguishes between two types of innovation activity: adoption of existing vintage technologies (“imitation”) and introduction of new “state-of-the-art” technology (“frontier innovation”). The key intuition is that countries farther away from the global technological frontier tend to grow mainly through imitation, whereas countries closer to the frontier rely more on innovation for growth. From a policy design perspective, an important implication of this theory is that the set of policies and reforms aimed at enhancing productivity can have varying effects depending on each country’s relative stage of technological development.

8. Empirical approach. Our empirical setup is similar to other recent studies in the literature (e.g., Scarpetta and Tressel 2002; Nicoletti and Scarpetti, 2003; Griffith, Redding, and Van Reenen 2004; Mc Morrow, Werner, and Turrini 2010). Specifically, we model TFP growth using the following baseline specification:

where subscripts i, j, t denote country, industry, and year, respectively. In particular, we use the subscript L to denote the “global frontier,” the country with the highest level of TFP in industry j in a given year t. The TFP growth rate (Δyijt) is the difference in the log-level of TFP between year t and t-1 regressed on the following explanatory variables: (i) the TFP growth rate in the global frontier (ΔyLit), (ii) the TFP level gap with respect to the global frontier measured as log-difference (yijt–1yLit-1), (iii) a set of structural variables (Xijt1k) and the interaction terms with the TFP gap, and (iv) country, industry, and year dummy variables. Conceptually, the TFP growth rate (ΔyLit) at the global frontier captures potential knowledge spillovers from cutting-edge innovation at the frontier, while the TFP gap represents the distance to the frontier, and hence the potential for “catch-up” for the follower country i. The dummy variables control for the effects of unobserved country and industry characteristics, as well as common shocks across time.

9. Policy and structural variables. We consider two groups of variables that are widely viewed as boosting productivity. The first group is comprised of measures of product and labor market regulations, which capture the degree of market flexibility and competition in an economy. Stricter regulations in these markets can curb TFP growth by impeding the efficient reallocation of resources across plants, firms, and industries. The second group consists of variables that proxy for the economy’s absorptive capacity for innovation and technology diffusion, namely, the intensity of high-skilled labor and ICT capital inputs and R&D expenditures. Beyond their direct contribution to output growth, high-skilled workers and ICT capital can create positive externalities for the overall economy by facilitating adoption of technologies. Similarly, R&D activities can directly contribute to productivity growth through the introduction of new technologies, but also have indirect effects by facilitating the adoption of existing technologies (Griffith, Redding, and Van Reenen 2004).

10. Data used. The sample for our analysis covers 23 market industries3 in 11 advanced economies (Australia, Austria, Denmark, Finland, Germany, Italy, Japan, Netherlands, Sweden, United States, and United Kingdom) over the period 1970–2007. The data definitions and sources for the variables used in the analysis are as follows:

  • Total factor productivity. TFP growth and level estimates are taken from the EU KLEMS database. The TFP level estimates used for the calculation of TFP gaps are obtained by linking the TFP growth estimates from the EU KLEMS database with the Groningen Growth and Development Centre Productivity Level database, which provides the PPP-adjusted industry TFP levels in 1997 for our sample countries (see Inklaar and Timmer 2008).
  • Product market regulation (PMR). We use the Organization for Economic Cooperation and Development’s Indicators of Regulation Impact to capture the extent of market regulation at the industry level. These indicators are calculated as the weighted-average of indicators that measure anti-competitive regulation in non-manufacturing upstream sectors (electricity, gas, air, rail, road, road, post and telecommunications, retail trade, professional services, and banking), where the weight for each indicator is the input requirement of these upstream sectors for each downstream industry derived from the OECD input-output tables (for details, see Conway and Nicoletti 2006).
  • Employment protection legislation (EPL). We use the OECD’s employment protection legislation indicator as a proxy for the degree of overall labor market regulations in an economy. The indicator measures the procedures and costs involved in dismissing individuals or groups of workers and the procedures involved in hiring workers on fixed-term or temporary work agency contracts. Unlike the other explanatory variables used in the empirical analysis, this indicator is only available at the aggregate economy level.
  • Share of high-skilled labor. The EU KLEMS database provides industry-by-industry information on different types of labor inputs classified by age, gender, and skill levels measured by educational attainment (primary, secondary, and tertiary). In our analysis, we use the share of total working hours provided by workers with tertiary education as the relevant measure of high skilled labor.
  • Share of ICT capital. The EU KLEMS database also provides information on various types of capital inputs (3 ICT - office and computer equipment, communication equipment, software; 4 non-ICT: transport equipment, other machinery and equipment, residential buildings, and nonresidential structures) at the industry level. For our analysis, we use the ICT capital share of total capital services as a proxy for the intensity of ICT capital usage in each industry.
  • R&D expenditure. The R&D expenditure data come from the OECD’s Analytical Business Enterprise Research and Development (ANBERD) database, which provides annual industry-level data (ISIC rev. 3). For our analysis, we divide the expenditure estimates by industry value-added.

A. Empirical Findings: Impact of Product and Labor Market Regulations

11. Baseline results. Table A1 presents the ordinary least squares (OLS) regression results showing the impact of product and labor market regulations on TFP growth across industries. Column 1 shows that, within each industry, the coefficients for TFP growth in the frontier country of the industry and the TFP gap with respect to the frontier country are both significant and are of the signs expected from neo-Schumpeterian growth theory. In particular, our analysis confirms the existence of significant productivity-enhancing knowledge spillovers from the frontier as well as the notion that the rate of convergence in follower countries increases with the distance to the frontier.4 The coefficients for product and labor market regulations, however, have unexpected positive signs (i.e., higher regulation is associated with higher TFP growth) but are statistically insignificant.

Table A1.Regression Results – Product and Labor Market Regulation
Dependent Variable: TFP Growth RateAll industriesManufacturingICT-related 1Service
(Percent, Annual)123456
C−5.387−5.238−5.081−4.313−2.387−4.306
[1.641]***[1.646]***[1.698]***[2.429]*[2.083][1.590]***
TFP Growth Rate at the Frontier0.0540.0530.0520.1150.0250.013
[0.014]***[0.014]***[0.014]***[0.031]***[0.013]*[0.011]
TFP Gap w.r.t. the Frontier−0.113−0.110−0.099−0.093−0.053−0.060
[0.023]***[0.023]***[0.027]***[0.037]**[0.029]*[0.026]**
Product Market Regulation0.3850.7170.9450.892−0.199−1.315
[0.431][0.460][0.516]*[0.786][0.776][0.445]***
Labor Market Regulation0.8280.8250.6450.8950.3950.451
[0.570][0.569][0.624][0.954][0.814][0.640]
Product Market Regulation0.006−0.006−0.010−0.017
X TFP Gap[0.007][0.008][0.010][0.005]***
Labor Market Regulation−0.008−0.007−0.014−0.012
X TFP Gap[0.008][0.012][0.011][0.011]
Product Market Regulation−0.638−1.255
X Manufacturing Dummy[0.424][0.536]**
Product Market Regulation−0.537−1.461
X Service Dummy[0.192]***[0.366]***
Product Market Regulation−0.014
X TFP Gap X Manufacturing Dummy[0.012]
Product Market Regulation−0.021
X TFP Gap X Service Dummy[0.007]***
# of obs.4,6464,6464,6462,4241,6161,414
R20.190.200.200.240.290.21
Source: Organization for Economic Co-Operation and Development; and IMF staff estimates.Note: The symbols *, **, *** denote statistical significance at 10, 5, and 1 percent, respectively. The figures in brackets are panel-corrected standard errors robust to heteroskedasticity and possible contemporaneous cross-section correlation. TFP growth at the frontier is the annual TFP growth rate of the country with the highest TFP level in industry j. TFP gap with respect to the frontier is defined as the percent deviation of the TFP level in country i from the frontier country in year t-1 (with minus sign). Product market regulation is an indicator measuring the potential costs of anti-competitive regulation in select nonmanufacturing sectors (electricity, gas, air, rail, road transport, post and telecommunications, retail, and professional services) on the industries that use the outputs of these sectors as intermediate inputs, normalized to have mean zero and unit standard deviation for this analysis. Labor market regulation is the employment protection legislation index, which we use as a proxy for labor market regulations.

Industries that produce ICT goods or employ ICT goods intensively (ISIC3 code: 21t22, 29, 30t33, 36t37, G, 64, J, 71t74).

Source: Organization for Economic Co-Operation and Development; and IMF staff estimates.Note: The symbols *, **, *** denote statistical significance at 10, 5, and 1 percent, respectively. The figures in brackets are panel-corrected standard errors robust to heteroskedasticity and possible contemporaneous cross-section correlation. TFP growth at the frontier is the annual TFP growth rate of the country with the highest TFP level in industry j. TFP gap with respect to the frontier is defined as the percent deviation of the TFP level in country i from the frontier country in year t-1 (with minus sign). Product market regulation is an indicator measuring the potential costs of anti-competitive regulation in select nonmanufacturing sectors (electricity, gas, air, rail, road transport, post and telecommunications, retail, and professional services) on the industries that use the outputs of these sectors as intermediate inputs, normalized to have mean zero and unit standard deviation for this analysis. Labor market regulation is the employment protection legislation index, which we use as a proxy for labor market regulations.

Industries that produce ICT goods or employ ICT goods intensively (ISIC3 code: 21t22, 29, 30t33, 36t37, G, 64, J, 71t74).

12. Varying impact of product market regulations by sector. To examine whether PMRs in different sectors have varying productivity effects, we introduce two additional explanatory variables by interacting the PMR index with manufacturing and service sector dummies (Column 2). In this specification, the coefficient for the interaction term between the PMR and the service sector dummy turns out to be negative and significant, indicating that greater product market regulation in the services sector is associated with lower TFP growth. When the interaction terms with TFP gaps are added for these sectors (Column 3), the coefficient of the interaction term for the service sector is negative and significant. This indicates that the negative productivity impact of PMR is more costly for countries closer to the frontier, consistent with findings from other studies (Mc Morrow, Werner, and Turrini 2010; Bourlès and others 2013).

13. Effects across manufacturing and services sectors. The PMR for the manufacturing sector also turns out to have a negative and significant effect on overall TFP growth, but unlike the services sector, the impact does not vary with the size of the TFP gap. Moreover, the regression results for the manufacturing subsample (Column 4) do not show a significant role of PMR for TFP growth in the manufacturing sector. This could be interpreted as an indication that the PMR in the manufacturing sector affects overall TFP growth mainly through its indirect impact on other nonmanufacturing sectors.

14. Impact of labor market regulations. The labor market regulation measure turns out to be insignificant in all the specifications. There could be several reasons for this result. First, it could reflect the fact that the aggregate-level employment protection legislation indicator exhibits little variation across countries and over time, failing to capture diverse industry-level TFP growth dynamics. Second, this is a partial measure of labor market rigidities, focusing only on the formal procedural costs of firing and hiring in an economy. For example, the flexibility of reallocating workers within firms is not captured by these indicators. Similarly, other labor market institutions, such as the degree of centralization in wage bargaining or the extent of labor tax wedges could significantly affect productivity growth. Finally, as is seen in the next section, it is possible that changes in labor market indicators have delayed effects on TFP growth and over a longer horizon.

B. Empirical Findings: Impact of Labor Skills, ICT Capital, and R&D

15. Baseline estimates. We next turn to the empirical estimates of the impact of labor skills, ICT capital, and R&D expenditure on TFP growth (Table A2). Columns 1–4 show that more intense use of high-skilled labor and ICT capital inputs, as well as higher spending on R&D activities, all csontribute positively and significantly to productivity growth. The channels through which they affect TFP growth differ. In the case of labor skills and ICT capital, TFP growth estimates, by construction, already take into account the direct contribution of skills and ICT capital services to value-added growth. Therefore, positive coefficients for these variables should be viewed as indicating the indirect returns to output beyond their direct contributions. In the case of labor skills, the additional return could be achieved through an improvement in the absorptive capacity for innovation (i.e. a higher share of more educated workers allows for faster acquisition of new technologies). In the case of ICT capital, more intensive use could bring about broader changes in the production process, the organizational structure, and other business practices within firms that lead to unobserved efficiency gains.

Table A2.Regression Results – ICT, Human Capital, R&D
Dependent Variable: TFP Growth RateAll industriesManufacturingICT-related 1Service
(Percent, Annual)12345678
C−0.579−1.7591.68−0.89−1.59−0.905−6.921−4.205
[0.508]*[0.760]**[0.606]***[1.101][0.691]**[1.005][1.875]***[1.719]**
TFP Growth Rate at the Frontier0.0190.0180.0430.0430.0460.0890.0280.005
[0.005]***[0.005]***[0.013]***[0.013]***[0.013]***[0.030]***[0.016]*[0.012]
TFP Gap w.r.t. the Frontier−0.019−0.021−0.009−0.008−0.026−0.043−0.076−0.038
[0.004]***[0.005]***[0.005]*[0.005][0.007]***[0.010]***[0.016]***[0.014]***
ICT Capital0.0250.0240.0230.1460.000−0.063
[0.007]***[0.014]**[0.022][0.053]***[0.037][0.037]*
High-skilled Labor0.0550.0470.120.0770.1830.236
[0.020]***[0.024]*[0.028]***[0.053][0.041]***[0.057]***
R&D Expenditure0.1180.0840.1950.1000.4800.387
[0.045]***[0.048]*[0.056]***[0.082][0.119]***[0.731]
ICT Capital0.0000.0020.000−0.002
X TFP Gap[0.000][0.001]**[0.001][0.001]**
High-skilled Labor0.0020.0020.0030.003
X TFP Gap[0.001]***[0.001][0.001]***[0.001]***
R&D Expenditure0.0020.0010.0060.013
X TFP Gap[0.001][0.001][0.002]***[0.013]
# of obs.6,9236,4022,9502,6852,6851,707849487
R20.110.110.110.110.110.150.210.24
Source: Organization for Economic Co-Operation and Development; and IMF staff estimates.Note: The symbols *, **, *** denote statistical significance at 10, 5, and 1 percent, respectively. The figures in brackets are panel-corrected standard errors robust to heteroskedasticity and possible contemporaneous cross-section correlation. TFP growth at the frontier is the annual TFP growth rate of the country with the highest TFP level in industry j. TFP gap with respect to the frontier is defined as the percent deviation of the TFP level in country i from the frontier country in year t-1 (with minus sign). ICT capital and high-skilled labor are the ICT capital share of total capital compensation and the hours worked by workers with tertiary education, respectively, all from the EU KLEMS. R&D expenditure is the R&D expenditure as percent of each industry’s value-added.

Industries that produce ICT goods or employ ICT goods intensively (ISIC3 code: 21t22, 29, 30t33, 36t37, G, 64, J, 71t74).

Source: Organization for Economic Co-Operation and Development; and IMF staff estimates.Note: The symbols *, **, *** denote statistical significance at 10, 5, and 1 percent, respectively. The figures in brackets are panel-corrected standard errors robust to heteroskedasticity and possible contemporaneous cross-section correlation. TFP growth at the frontier is the annual TFP growth rate of the country with the highest TFP level in industry j. TFP gap with respect to the frontier is defined as the percent deviation of the TFP level in country i from the frontier country in year t-1 (with minus sign). ICT capital and high-skilled labor are the ICT capital share of total capital compensation and the hours worked by workers with tertiary education, respectively, all from the EU KLEMS. R&D expenditure is the R&D expenditure as percent of each industry’s value-added.

Industries that produce ICT goods or employ ICT goods intensively (ISIC3 code: 21t22, 29, 30t33, 36t37, G, 64, J, 71t74).

16. TFP growth impact and distance to the frontier. Column 5 in Table A2 suggests that, within each industry, the productivity impact of high-skilled labor is larger for countries that are closer to the technology frontier. This provides empirical support for policies aimed at boosting higher-level education in countries with already-high productivity levels, and is also in line with recent studies (Mc Morrow, Werner, and Turrini 2010; Vandenbussche, Aghion, and Meghir 2006). Innovation also appears to improve productivity, but the impact does not vary with the TFP gap. As in the case of labor skills, the coefficient estimates should be viewed as capturing the additional returns to productivity growth beyond the direct contribution to output.

17. Results for industry subsamples. The results in Columns 6–7 of Table A2 for manufacturing, ICT, and services sector subsamples show that the effects of labor skill, ICT capital service, and R&D are both sector- and country-specific (in terms of the distance to the frontier). The detailed results from sector-level regressions are as follows:

  • Manufacturing. The coefficient for ICT capital service is positive and significant, suggesting that a more intensive use of ICT capital generates positive returns in addition to directly increasing output (Column 6). Furthermore, this impact is larger as a country gets closer to the global frontier. The effects of labor skills and R&D expenditure, on the other hand, are not statistically significant. While this is somewhat unexpected, one potential explanation could be that many manufacturing industries are relatively “low-tech” in nature, while the use of workers with tertiary education (our definition of “high-skilled” workers) and R&D activities are likely to matter more for “high-tech” industries that require developing and adopting frontier cutting-edge technologies (e.g., in the ICT sector). Another reason could be the potentially long time horizon over which R&D affects TFP growth. Introduction of new innovative products and production processes, for example, may not necessarily lead to immediate productivity gains, requiring time for their technological potential to be sufficiently exploited.
  • Information and communication technologies (ICT). The coefficient on high-skilled labor for TFP growth in the ICT sector is positive and significant.5 Furthermore, the coefficient on the interaction terms with the TFP gap variable suggests that the productivity growth gain is larger for countries closer to the frontier. Similarly, the impact of R&D expenditure is also positive and significant, and larger for countries that are closer to the frontier, as indicated by the positive sign for the interaction term with the TFP gap.6 The result for R&D could, in part, reflect the innovative nature of activities in this sector, which are likely to be more cutting-edge in nature and therefore matter more for countries at higher stages of technological development.
  • Services. As in the ICT sector, the empirical results indicate that a higher share of high-skilled labor has a significant and positive impact on TFP growth, particularly for countries closer to the frontier. The coefficient for ICT capital is negative, indicating that intense use of ICT capital in the services sector may actually have an unobserved dampening effect on TFP growth. It should be noted, however, that the coefficient on the interaction term with the TFP gap has a negative sign and is significant, implying that the net impact for countries sufficiently farther away from the frontier may actually be positive. Given that the average service sector TFP gap for our sample countries is about 48 percent, the average country is expected to benefit from more intensive use of ICT services (0.002 x 48 - 0.063 = 0.033).

18. Robustness checks. The results presented in Tables A1 and A2 are broadly robust under alternative specifications, sample periods, and the inclusion of other control variables that capture macroeconomic conditions, such as the output gap and the systemic banking crisis dummy variable. First, to control for possible country-specific developments, we used a joint time-country fixed effect in Table A2 instead of considering country and time fixed effects separately.7 In this case, both the significance and the signs of the coefficients remain the same as in the baseline specification in the full sample (Column 5), but the interaction term between the R&D and the TFP gap becomes significant and positive, implying greater productivity gains for countries closer to the frontier. Second, we also obtained broadly similar results when using the post-1995 subsamples. Notably we find that while the coefficient of the interaction term for manufacturing PMR becomes significant and negative in the all-industry sample, the coefficients for R&D expenditure and its interaction term with the TFP gap become significant and positive in the manufacturing subsample. Finally, the main baseline results remain unchanged even after controlling for macroeconomic conditions, which are captured using the output gap and the systemic banking crisis dummies.

III. Dynamic Effect of Reforms

This section describes the framework to assess the dynamic effects of structural reforms on sectoral TFP, employment, and output, then discusses the main findings. We also investigate how the effects of reform depend on the distance from the frontier, initial structural and institutional settings, and prevailing economic conditions.

19. Identification of reforms. Reforms are identified using the same set of policy and structural indicators described in the previous section. We also include a measure of the level of infrastructure and the labor tax wedge. The infrastructure measure is the principal component of roads, phones lines, and electricity generation capacity (IMF 2014). the labor tax wedge is taken from the OECD’s Taxing Wages database and measured as a percentage of taxes and transfers paid in the share of total labor costs. An increase (in absolute terms) of a structural indicator above a certain threshold is assumed to signal a reform, while any other change—including in the opposite direction—is not. This definition is directional in the sense that large changes in indicators in the opposite direction (e.g., an increase in product and labor market rigidities) are ignored. The threshold is constructed by industry, country, and year to capture the catch-up effect to the best performers. More specifically, a reform variable is equal to 1 if the change in R&D, ICT capital, and high-skilled labor is above (or below for labor tax wedge) two standard deviations of the average annual change in the indicator, and 0 otherwise. Given little variation in some structural indicators over time, a reform variable takes the value of 1 if it is below (product and labor market regulations) or above (infrastructure) one standard deviation of the average annual change in the indicator, and 0 otherwise. Table A3 presents the descriptive statistics underlying the analysis.

Table A3.Descriptive Statistics
Policy/ Structural IndicatorsMeanStandard

Deviation
Number of

Reform

Shocks per

Country
Average

Cumulative

5-year

Change in TFP

Following the
Percent of Shocks and

Average Cumulative 5-year

TFP Change
Shocksduring upturnsduring downturns
Product Market Regulation2.771.161366%55%45%
Job Protection Legislation1.780.96155%3%67%
Labor Tax Wedge0.071.141110%82%18%
Infrastructure0.331.91353%91%9%
R&D2.995.167518%56%44%
ICT capital0.130.1619714%71%32%
High-Skill Labor8.948.773296%40%60%
Memorandum Items: Average Cumulative 5-year Changes
TFP level in the sample4%19%
Employment level in the sample−1%16%
Output (value added) level in the sample11%18%
Source: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.
Source: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

20. Empirical specification. A dynamic version of the empirical specification outlined in the previous section is used to assess the temporal effects of reform shocks on changes in the TFP level. In particular, the following equation is estimated for each of the three or five years after the reform episode:8

where tfpijt is the log of real TFP in country i, industry j, and year t and Si,j,t denotes reform dummies (equal to 1 in case a reform has been identified and zero otherwise). The log of real TFP at the frontier industry j and technological gap from the frontier are indicated by tfpL,j,t and tfpgapi,j,t, respectively. Xit is a set of controls and Di, Dj, and Dk are country, industry and time dummies, respectively.9 As economic crises can facilitate the introduction of reforms, controls include recession and financial crises dummies. GDP growth is also included to further control for the procyclical nature of TFP measures. The estimated coefficients β1 and β2 capture the unconditional and conditional (taking into account the technological gap) effect of reforms at horizon k and yield the impulse response function. In the baseline specification, the impact of a reform is assessed at the average technological gap: β1+β2tfpgap¯.

21. Impact on TFP across all industries. Figure A1 presents the effect of reforms on the cumulative short-term (three years) and medium-term (five years) changes in the TFP level. It indicates that PMR reforms, a large decline in the labor tax wedge, and significant increases in R&D, ICT capital, and infrastructure generally have a positive impact on TFP growth. A reduction in employment protection legislation, however, tends to be associated with a short-term reduction in aggregate TFP level.10 More specifically, Figure A1 shows that an average product market regulation shock, on average, increases the TFP level across all industries by about 1 percent after five years (i.e., 0.05 standard deviation of the average cumulative five-year TFP change in the sample). Further, a 10-percent increase in R&D spending, on average, increases the TFP level across all industries by about 2 percent in the medium term (0.1 standard deviation of the average cumulative five-year TFP change in the sample). The results appear to confirm previous findings,11 particularly those that find that deregulation of network industries is positively associated with higher efficiency in sectors that more intensively use these industries as intermediate inputs (Bourlès and others, 2013; IMF 2014b).

Figure A1.Dynamic Impact of Reforms on Total Factor Productivity

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show cumulative three-year (ST) and five-year (MT) level gains in TFP. Brown and shades of red indicate a negative statistically significant impact of the reform; shades of blue indicate a positive statistically significant one. For instance, following an average reform shock in ICT capital, the cumulative increase in the TFP level in ICT-related industries is about 30 percent after five years (i.e., about two standard deviations of an average cumulative five-year change in the TFP level in the sample). “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

22. Reform impact tends to vary across different industries. PMR reforms increase TFP in manufacturing and distribution both in the short and medium term, and over the medium term in ICT and personal services. Large infrastructure shocks have a positive and increasing impact on productivity in ICT, finance-business, and other production.12 R&D shocks steadily boost TFP in manufacturing and ICT, and, over the medium term, in distribution services. ICT shocks have a positive permanent impact on manufacturing and personal services, while a large increase in high-skilled labor affects medium-term TFP growth in ICT and other production sectors (e.g., agriculture, mining, and utilities). Some adjustment costs can be discerned in the ICT, finance-business, and personal services sectors following product and market liberalization.

23. Reforms and distance to the frontier. As the impact of structural reforms may depend on the degree of technological advancement, we extend the baseline analysis by focusing on technological gaps with the frontier. In particular, we estimate how the effect of reforms varies between an industry with a lower technological gap (75th percentile of the distribution) and an industry with a higher technological gap (25th percentile of the distribution) with respect to the frontier. Consistent with the results presented in the previous section, we find that the impact of product market liberalization, R&D, ICT, and human capital is larger the closer a country is to the frontier (Figure A2). In particular, the TFP increase in industries with a lower technological gap, on average, is about 2 percent higher following PMR and infrastructure reforms, and about 10 percent higher following R&D and ICT shocks, than for industries farther away from the frontier. A significant increase in high-skilled labor raises the TFP level by about 7–10 percent more in the distribution and ICT-related industries that are closer to the frontier. The positive productivity impact of infrastructure, however, does not appear to be affected by the distance to the technological frontier.

Figure A2.Dynamic Impact of Reforms on TFP: Conditional on the TFP Gap

(Percent)

Sources: EU KLEMS database: World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show cumulative five-year level gains in TFP as a function of the distance to the technological frontier. For instance, following an average R&D shock, the cumulative five-year increase in the aggregate TFP level is about 15 percent higher (i.e., about one standard deviation of an average cumulative five-year change in TFP level in the sample) for industries closer to the technological frontier (75th percentile of the distribution) than farther from it (25th percentile). “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

24. Pre-reform policy and structural settings. The impact of reforms may also differ depending on initial structural and policy settings. Therefore, the baseline specification is extended as follows:

where the estimated coefficients δ1<Xth and δ2<Xth capture the impact of reforms for industries with initial policy and structural settings below the 25th, 50th, and 75th percentile of the distribution. Conversely, δ1>Xth and δ2>Xth denote the average reform effects for industries with initial settings above the 75th, 50th, and 25th percentile of the distribution. The results (reported in Figure A3) appear broadly in line with previous findings.13 In particular, a large decline in PMR, on average, results in the highest productivity gains for industries with initially higher regulatory burdens, while the impact is not statistically significant for industries with the lowest regulations. Similarly, the impact of R&D, ICT, and infrastructure shocks on TFP, on average, is higher for industries with higher initial levels of innovation and infrastructure. The impact of labor tax wedge shocks is higher for industries in countries with an initial level of the tax wedge above the 75th percentile and not significant for industries in countries with lower tax wedges. Some differences, however, can be gleaned across industries, especially in ICT and personal services, where productivity gains tend to be higher the lower are initial levels of R&D and ICT capital use.

Figure A3.Dynamic Impact of Reforms on TFP: Conditional on Initial Settings

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show the cumulative five-year level gains in TFP as a function of initial policy and structural settings. “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

25. Reform impact across the business cycle. The impact of structural reforms may also depend on the business cycle position, in particular on how changes in institutional and structural settings affect supply and demand factors. In particular, the impact of reforms may be stronger if they increase demand more than supply during a downturn. To capture the importance of economic conditions, the baseline specification is extended as follows:14

where the coefficients γ1 and θ1 capture the impact of reforms during downturns and γ2 and θ2 during upturns.15 The results (Figure A4) confirm that infrastructure shocks are associated with significantly higher productivity gains during downturns (Abiad, Furceri, and Topalova, forthcoming), while product market liberalization and innovation shocks have a more significant impact during upturns.

Figure A4.Dynamic Impact of Reforms on TFP across the Business Cycle

(Percent)

Source: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show the cumulative five-year level gains in TFP across overall economic conditions. For instance, following an average reform shock in ICT capital during upturns, the cumulative five-year increase in the average TFP across all industries increases by about 15 percent (i.e., 0.8 standard deviation of the average cumulative five-year TFP level change in the sample). “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

A. Dynamic Impact on Employment and Output Growth

26. Empirical approach. Previous analyses have shown that although structural reforms can have substantial medium-term benefits, there could be short-term costs associated, for instance, with adjustment, reallocation, or loss of physical and human capital (Blanchard and Giavazzi, 2003). To shed more light on this issue, this section extends the dynamic analysis of sectoral productivity on employment and output. In particular, the baseline specification is modified as follows:

where vaijt is the log of real value added or employment in country i, industry j, and year t. Si,j,t denotes the reform variables. Xit is a set of controls and Di, Dj, and Dk are country, industry, and time dummies, respectively. As the analysis focuses on the overall (i.e., unconditional) impact of reform shocks, no transition channel is specified.

27. Impact on employment. The results (reported in Figure A5) suggest that product market liberalization and infrastructure shocks have a positive and persistent impact on employment, while ICT capital shocks seem to reduce it, likely reflecting the effect of automation on labor shedding. More intensive use of high-skilled labor has a positive effect on employment in manufacturing and ICT industries, but negatively affects employment in other production sectors. Labor market liberalization, proxied here by employment protection legislation, boosts employment in personal services in the short term.

Figure A5.Dynamic Impact of Reforms on the Sectoral Employment Level

(Percent)

Source: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates

Note: The results show cumulative three-year (ST) and five-year (MT) level gains in employment. Brown and shades of red indicate a negative and statistically significant impact of the reform; shades of blue indicate a positive and statistically significant one. For instance, following an average R&D shock, the cumulative increase in the employment level in finance/business-related industries is about 10 percent after five years (i.e., about 0.7 standard deviations of the average cumulative five-year change in employment level in the sample). “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

28. Impact on output. In general, structural reforms have positive and increasing effects on output (Figure A6). In particular, product market liberalization and more intensive use of high-skilled labor increase overall industrial value added by about 2 percent over the medium term. Effects of infrastructure and innovation shocks are even larger, boosting output, on average, by around 3 and 5 percent, respectively. Labor market liberalization has a short-term negative impact on output, similar to that on TFP. Some differences can be noticed across industries. For instance, high-skilled labor and R&D shocks have larger positive effects on manufacturing and ICT-related industries. ICT shocks increase output in ICT, distribution and personal services. Infrastructure shocks have a positive short-term impact on almost all sectors.

Figure A6.Impact of Reform Shocks on Sectoral Output

(Percent)

Sources: EU KLEMS database; World KLEMS database; OECD and IMF databases; and IMF staff estimates.

Note: The results show cumulative three-year (ST) and five-year (MT) level gains in output. Brown and shades of red indicate a negative impact of the reform; shades of blue indicate a positive one. For instance, following an average shock in high-skilled labor, a cumulative increase in output level in manufacturing-related industries is about 6 percent after five years (i.e., about 0.3 standard deviations of an average cumulative five-year change in output level in the sample). “Other production” includes agriculture, forestry, fishing, mining, quarrying, electricity, gas, and water-related industries.

References

    Abiad, A., D.Furceri, and P.Topalova. Forthcoming. “Macroeconomic Effects of Public Investment: Evidence from Advanced Economies.” IMF Working Paper.

    • Search Google Scholar
    • Export Citation

    Acemoglu, D., F.Zilibotti, and P.Aghion. 2006. “Distance to Frontier, Selection and Economic Growth.” Journal of the European Economic Association4(1): 37–74.

    • Crossref
    • Search Google Scholar
    • Export Citation

    AghionP., and P.Howitt. 2006. “Joseph Schumpeter Lecture Appropriate Growth Policy: A Unifying Framework.” Journal of the European Economic Association4(2-3): 269–314.

    • Crossref
    • Search Google Scholar
    • Export Citation

    AghionP., and P.Howitt. 2009. The Economics of Growth. Cambridge, Massachusetts: MIT Press.

    AhrendR., J.Arnold, and C.Moeser. 2011. “The Sharing of Macroeconomic Risk: Who Loses (and Gains) from Macroeconomic Shocks.” OECD Economics Department Working Paper No. 877, Organization for Economic Cooperation and Development, Paris.

    • Search Google Scholar
    • Export Citation

    Aoki, S.2012. “A Simple Accounting Framework for the Effect of Resource Misallocation on Aggregate Productivity.” Journal of the Japanese and International Economies26(4): 473–94.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Blanchard, O., and F.Giavazzi. 2003. “Macroeconomic Effects Of Regulation And Deregulation In Goods And Labor Markets.” The Quarterly Journal of Economics118(3): 879–907.

    • Crossref
    • Search Google Scholar
    • Export Citation

    BouisR., O.Causa, L.Demmou, R.Duval, and A.Zdzienicka. 2012. “The Short-Term Effects of Structural Reforms: An Empirical Analysis.” OECD Economics Department Working Paper No. 949, Organization for Economic Cooperation and Development, Paris.

    • Search Google Scholar
    • Export Citation

    Bourlès, R., G.Cette, J.Lopez, J.Mairesse, and G.Nicoletti. 2013. “Do Product Market Regulations In Upstream Sectors Curb Productivity Growth? Panel Data Evidence for OECD Countries.” The Review of Economics and Statistics95(5): 1750–768.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Conway, P., and G.Nicoletti. 2006. “Product Market Regulation in the Non-Manufacturing Sectors of OECD Countries: Measurement and Highlights.” OECD Economics Department Working Paper No. 530, Organization for Economic Cooperation and Development, Paris.

    • Search Google Scholar
    • Export Citation

    Dabla-Norris, E., and others. 2013. “Anchoring Growth: The Importance of Productivity-Enhancing Reforms in Emerging Market and Developing Economies.” Staff Discussion Note 13/08, International Monetary Fund, Washington.

    • Search Google Scholar
    • Export Citation

    Duval, R.2008. “Is There a Role for Macroeconomic Policy in Fostering Structural Reforms? Panel Evidence from OECD Countries over the Past Two Decades.” European Journal of Political Economy24(2): 491–502.

    • Crossref
    • Search Google Scholar
    • Export Citation

    GriffithR., S.Redding, and J.Van Reenen. 2004. “Mapping the Two Faces of R&D: Productivity Growth in a Panel of OECD Industries.” The Review of Economics and Statistics86(4): 883–95.

    • Crossref
    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2014a. World Economic Outlook, Washington, October.

    International Monetary Fund (IMF). 2014b. “Macroeconomic and Reform Priorities.” Paper presented at the meeting of G-20 Finance Ministers and Central Bank Governors, 22-23 February, Sydney, Australia.

    • Search Google Scholar
    • Export Citation

    Inklaar, R., and M. P.Timmer. 2008. “GGDC Productivity Level Database: International Comparisons of Output, Inputs and Productivity at the Industry Level.” GGDC Research Memorandum GD-104, Groningen Growth and Development Centre, University of Groningen.

    • Search Google Scholar
    • Export Citation

    Jorda, O.2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review95(1): 161–82.

    Mc Morrow, K., R.Werner, and A.Turrini. 2010. “Determinants of TFP Growth: A Close Look at Industries Driving the EU-US TFP Gap.” Structural Change and Economic Dynamics21(3): 165–80.

    • Search Google Scholar
    • Export Citation

    Nicoletti, G., and S.Scarpetta. 2003. “Regulation, Productivity and Growth: OECD Evidence.” Economic Policy18(36): 9–72.

    Organization for Economic Cooperation and Development (OECD). 2007. OECD Employment Outlook. Paris: Organization for Economic Cooperation and Development.

    • Search Google Scholar
    • Export Citation

    Teulings, C., and N.Zubanov. 2014. “Is Economic Recovery A Myth? Robust Estimation Of Impulse Responses.” Journal of Applied Econometrics29(3): 497–514.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Vandenbussche, J., P.Aghion, and C.Meghir. 2006. “Growth, Distance to Frontier and Composition of Human Capital.” Journal of Economic Growth11(2): 97–127.

    • Crossref
    • Search Google Scholar
    • Export Citation
1

The authors would like to thank Jovana Sljivancanin, Marijn Willem Otte, Tushara Ekanayake, and Xin Guo for excellent research assistance, and Ricardo Reinoso for editorial assistance.

2

The EU KLEMS database is compiled as a product of a research project financed by the European Commission (O’Mahony and Timmer 2009).

3

Estimating the trend on the whole sample, and ending roughly at the trough of the economic cycle, biases the trend growth rates down. By the same token, the pre-crisis sample ends at a cycle peak, meaning that the trends are biased upward. The true underlying trend is likely to fall somewhere between these two estimates. Figure 1 plots the growth rates of output- and productivity-level trends, estimated with an HP filter using a smoothing parameter of 100.

4

Countries in the sample are Japan, the United States, the United Kingdom, Core Europe (Austria, Belgium, France, Germany, and the Netherlands); euro area countries with high borrowing spreads (Greece, Italy, Ireland, Portugal, and Spain); and Natural Resource Producers (Norway, Canada, and Australia).

5

Australia was an exception, as contributions from capital deepening, particularly in the mining sector, have been a key feature of labor productivity growth since the early 1990s, although prospects here may be more mixed with the ending of the commodity super-cycle.

6

TFP growth depends on advances in technology and also captures the efficiency with which labor and capital are combined to generate output. This depends not only on businesses’ ability to innovate, but also on the extent to which they operate in an environment that fosters competition, imposes less onerous administrative burdens, provides modern and efficient infrastructure, and allows easy access to finance (Svyerson 2011). TFP is measured as a residual, and any measurement errors in the labor and capital series will be captured in the estimates. Extensive robustness checks were performed using alternative data sources to assess whether the broad patterns across countries and over time were broadly consistent.

7

Controlling for capacity utilization, which tends to cyclical, makes the post 2000 decline in TFP across advanced economies even more apparent (IMF, forthcoming).

8

Cross-country productivity estimates often find extractive industry productivity to be negatively correlated with commodity prices, due to decreasing returns to scale, and challenges with measuring the quality of resource inputs and outputs (Zheng and Bloch 2014).

9

For instance, companies such as Walmart and Amazon transformed retailing in the United States, pushing through operational and technological improvements as they gained market share (Lewis 2005).

10

The TFP level estimates were obtained by linking the TFP growth estimates from the EU KLEMS database with the Groningen Growth and Development Centre (GGDC) Productivity-Level database, which provides the PPP-adjusted industry TFP levels in 1997 for our sample countries (Inklaar and Timmer 2008). The use of sector-specific price indices allows for meaningful comparisons of the level of productivity in different sectors across countries.

11

For example, Baily, Hulten, and Campbell (1992) find that 50 percent of the growth in manufacturing productivity in the United States in the 1970s and 1980s can be attributed to the reallocation of factors across plants and firm entry and exit (see Foster, Haltiwanger, and Syverson 2008). Similarly, Barnett and others (2014) find that labor reallocation across firms explained 48 percent of labor productivity growth for most sectors in the U.K. economy in the five years prior to 2007.

12

Hopenhayn and Rogerson (1993) showed that labor market firing taxes distort the allocation of labor across establishments and that empirically reasonable values for this tax could generate TFP losses on the order of about 5 percent. In a related study, Lagos (2006) shows how policies such as unemployment insurance and employment protection affect TFP via selection effects. See also Garicano, Lelarge, and Van Reenen (2013).

13

Gilchrist, Sim, and Zakrajsek (2013) relate financial frictions, measured as an increased dispersion in borrowing costs, to capital misallocation across U.S. manufacturing firms even prior to the crisis.

14

The Pareto efficient allocation requires that the marginal return to capital (or labor) be identical across sectors. A larger cross-sector dispersion of input marginal products thus implies a less efficient input allocation, and hence lower aggregate TFP.

15

Note that our analysis ignores the heterogeneity of capital and its effect on measurement of frictions. Systemic differences in the composition of capital stock across countries could affect our results.

16

We conduct an industry-level panel study for 23 market industries in 11 advanced economies using annual data over 1970–2007 (see Technical Appendix for details). Our empirical specification controls for country-industry specific characteristics and common factors affecting TFP growth, as well as the TFP gap with respect to the “global frontier—defined as the highest level of TFP in a specific industry in a given year.

17

For the product market regulation measure, we use the industry-specific OECD Indicators of Regulation Impact, which capture the extent of anti-competitive regulations in upstream industries. For labor market regulations, the OECD’s Employment Protection Legislation indicator is used as the proxy. Unlike other explanatory variables used in the empirical analysis, the employment protection indicator is only available at the economy-wide level. See Technical Appendix for more details.

18

The baseline results reported in Tables 2 and 3 are broadly robust to alternative specifications and the sample periods considered (see Technical Appendix).

19

This finding is consistent with previous studies that find that differences in TFP growth rates across industries are linked to persistent cross-country differences in R&D intensity. The empirical literature has identified three set of factors as potential determinants of industry variation in research intensity and productivity growth: technological factors affecting the efficiency of research; the extent to which R&D benefits the innovator; and demand, which influences the returns to research (Ngai and Samaniego 2011).

20

OECD (2001) finds that firms in the United States and Canada enjoyed considerably lower costs of ICT investment goods in the 1990s than firms in European countries and Japan. Barriers to trade, in particular nontariff barriers related to standards, import licensing, and government procurement, may partly explain cost differentials. The difference in the timing of the technological boom in ICT and the regulatory reform process in Europe is also considered a key factor hindering effective accumulation of ICT capital, particularly in services industries where deregulation has been much slower to take place (Arnold, Nicoletti, and Scarpetta 2008).

21

In particular, a dynamic version of the empirical specification outlined in the previous section is used to assess the effects of reforms on TFP levels over time.

22

Specifically, a shock is identified if the change in R&D, ICT capital, and high-skilled labor is above two standard deviations of the average annual change in the indicator. Because product and labor market regulations and infrastructure do not vary as much over time, a major reform is assumed to have been undertaken when the change in the policy indicators in a given year is below (above for infrastructure) one standard deviation.

23

The infrastructure measure is the principal component of roads, phones lines, and electricity generation capacity. For more details see IMF (2014b). The labor tax wedge is taken from the OECD Taxing Wages database, measured as a percentage of taxes and transfers paid in the share of total labor costs.

24

Reallocation is important for productivity growth, but much productivity growth also comes from stable employment relationships, as they allow for investment in firm-specific skills (Blanchard, Jaumotte, and Loungani 2013). However, excessive protection that takes the form of complex legal and administrative restrictions on the separation process can increase the cost of adapting to new technologies and hamper the reallocation of labor into higher productivity activities, firms, or industries.

25

While the large TFP gains associated with infrastructure shocks are likely to be driven by the fact that our measure of TFP embodies public infrastructure capital, the results are also indicative of complementarities between private and public capital (Dabla-Norris and others 2013).

26

Reform shocks are broadly equally distributed between periods of economic upturns and downturns in the data. As economic crises often facilitate reform introduction (Duval 2008), recession and crises dummies are included in the model specification.

1

The authors would like to thank Jovana Sljivancanin, Tushara Ekanayake, Xin Guo, Monica Devi, and Ricardo Reinoso for their excellent research and editorial assistance.

2

Aoki (2012) does not specify the form of the final good production function. However, he implicitly assumes that sector shares do not respond changes in frictions, which is equivalent to the special case with unitary elasticity of substitution. In our counterfactual analysis, we allow for a potential change of sector shares.

3

Industry coverage includes food, beverage, and tobacco (15t16); textiles, textile products, leather, and footwear (17t19); wood and products of wood and cork (20); pulp, paper, printing and publishing (21t22); chemicals and chemical products (24); rubber and plastics (25); other non-metallic mineral products (26); basic metals and fabricated metal (27t28); machinery, nec (29); electrical and optical equipment (30t33); transport equipment (34t35); manufacturing nec; recycling (36t37); electricity, gas and water supply (E); construction (F); wholesale and retail trade (G); hotels and restaurants (H); transport and storage (60t63); post and telecommunications (64); real estate activities (70); renting of machinery & equipment and other business activities (71t74); financial intermediation (J).

4

We conducted further robustness checks without industry/country fixed effects, by sector, and for the post-1995 period, all of which confirmed that our results still hold under these different specifications.

5

We define the ICT sector as consisting of manufacturing industries that produce ICT goods and service industries that use ICT goods intensively (e.g., telecommunications).

6

The latter result is in contrast to Griffith, Redding, and Van Reenen (2004), who find an increasing positive impact of R&D spending for countries farther away from the frontier in select manufacturing industries over 1974–90.

7

In Table A1, the labor market regulation measure is a country-level index and hence including the joint fixed effect leads to perfect collinearity.

8

Robust standard errors are clustered at country and industry levels. The empirical setup consists of derivate impulse response functions from local projections (Jorda, 2005; Teulings and Zubanov 2014) by estimating a set of k (with k=1…3 or k=1…5) independent equations. This approach has been applied, for instance, by Ahrend, Arnold, and Moeser (2011) and Bouis and others (2012).

9

To minimize potential omitted variable bias resulting, for instance, from country-wide developments, an alternative specification includes a country-time dummy. The results remain qualitatively similar.

10

Partial labor protection reforms have been associated with lower GDP, consumption, and investment growth in the short term (Bouis and others 2012). However, the cross-country productivity effects of labor market reforms are less clear-cut in the literature (OECD 2007).

11

The results are qualitatively similar when all reforms and structural change shocks are jointly estimated. In particular, the coefficients associated with product market reforms, R&D, and ICT shocks remain statistically significant. In addition, while the coefficients associated with infrastructure and high-skilled labor shocks become statistically insignificant due to multicollinearity between the various reform variables, a test of joint significance validates the inclusion of these variables in the model.

12

While the large gains in TFP associated with infrastructure shocks are likely to be driven by the fact that our measure of TFP embodies public infrastructure capital, the results are also indicative of complementarities between private and public capital (Dabla-Norris and others 2013).

13

For instance, Bourlès and others (2013) find that the highest productivity gains accrue from adopting best practices in terms of PMR regulations in highly-regulated sectors with lower gains resulting when best practices are adopted in sectors with initially low regulatory burdens.

14

Reform shocks are broadly equally distributed between periods of economic upturns and downturns. As economic crises can facilitate the introduction of reforms (Duval 2008), recession and crises dummies are included in the model specification.

15

Overall economic conditions are identified using the output gap and, alternatively, downturn and expansion normalized indicators (IMF 2014a).

Other Resources Citing This Publication