2. Japan’s Potential Output and Productivity Growth
- Alessandro Zanello, and Daniel Citrin
- Published Date:
- November 2008
Japan’s economic re-awakening over the past five years raises the question of whether the country’s potential output growth rate may now be higher. Structural adjustments to the imbalances of the “bubble” years have strengthened fundamentals. At the same time, an aging population complicates the challenge of ensuring strong self-sustaining growth. With Japan’s birth rate well below the population’s replacement rate, the working-age population in fact has been contracting since 2000, and the elderly dependency ratio (the share in the working-age population of people at least 65 years old) is now the highest among industrial countries. With a declining labor force, per capita income growth will depend critically on higher productivity.1
This chapter tries to shed light on the drivers of Japan’s long-term growth—and speculates about future potential growth. More specifically, it seeks to estimate the current level of potential output growth, the determinants of productivity growth, and the likely impact on productivity growth and potential output of reforms undertaken in recent years.
The empirical results indicate that potential output growth in Japan is likely to have picked up and is likely to be in the range of 1¾ to 2 percent in the near term. Higher total factor productivity (TFP) growth has helped offset a declining contribution of capital inputs and a negative contribution of labor inputs. The results also confirm that product market competition, openness, and research and development (R&D) investment are key determinants of productivity growth. At the same time, structural unemployment, which remains historically high at around 4 percent, is directly related to the generosity of the unemployment insurance system (the level of out-of-work benefits relative to in-work wages and salaries) and the aging of the labor force.
The potential gains from reforms are significant. The findings suggest that, going forward, the removal of lingering labor and product market distortions—cutting excessive domestic regulation, strengthening the anti-trust framework, and further liberalizing trade (specifically, agricultural)—together with increased returns on R&D investment, could raise further total factor productivity and hence potential output growth. Additional efforts to liberalize the labor market to reduce structural unemployment (e.g. reducing the negative effects of unemployment benefits on work incentives) could also provide a significant boost to output.
What is potential output growth?
There is a plethora of studies on Japan’s potential output growth. Depending on the methodology used, results differ markedly both with respect to the estimated potential growth rate and the contributions of key structural factors.2 Estimates of potential output growth recently prepared by official agencies range between 1½ percent and 2 percent. This diversity in estimates stems mainly from the fact that potential output is an unobserved variable and can only be inferred indirectly.3
The methodology used here combines filtering techniques to estimate trends along with estimation of a structural model encompassing behavioral relationships from economic theory. Potential output is considered as the level of output that would emerge from an aggregate production function, given the current levels of fixed inputs and levels of variable inputs consistent with stable inflation (Appendix 2.1).
Econometric analysis produces parameter estimates that are significant and correctly signed (Table 2.1). The estimated system consists of four basic structural equations (see Appendix 2.2)—an aggregate production function, a Phillips curve, an equation to estimate the non-accelerating inflation rate of unemployment (NAIRU), and “Okun’s law” relating the unemployment gap to the output gap—as well as several identities.4 In particular:
|TFP||Unemployment||Phillips Curve||Okun’s law||Utilization|
|Explanatory Variables||TFP||NAIRU||CPI Inflation||Gap||Utilization|
|CPI inflation at (t-1)||0.18(39)a|
|CPI inflation at (t-2)||0.37 (3.6)a|
|CPI inflation at (t-3)||0.17 (3.7)a|
|CPI inflation at (t-4)||0.27 (n.a.)|
|Change in Import Prices||0.05 (2.3)a|
|Change in Oil Prices||0.01 (2.4)a|
|Output Gap||0.07 (2.3)a||0.01 (2.3)a||0.04 (2.8)a|
|Output Gap at (t-1)|
|Change in Output Gap||0.00 (n.a.)|
|Unemployment Gap at (t-1)||0.81 (23.8)a|
|TFP at (t-1)||1.00 (n.a.)|
|Change in R&D Intensity at (t-4)||1.08 (2.0)a|
|Change in Mark-up||-1.12(4.8)a|
|Import Penetration||0.02 (2.8)a|
|NAIRU at (t-1)||1.00 (n.a.)a|
|Change in Replacement Ratio||0.03 (2.5)a|
|Share of Old in Labor Force||0.10 (3.0)a|
|Capacity Utilization at (t-1)||0.97 (76.1)a|
5% significance level.Source: IMF staff estimates.
5% significance level.Source: IMF staff estimates.
In the aggregate production function, total factor productivity depends on R&D intensity (as measured by R&D spending in relation to GDP), mark-ups (as measured by the ratio of operating profits to sales net of cost of sales), and import penetration (as measured by the ratio of imports to domestic demand). R&D intensity suggests a positive impact of innovation; mark-ups indicate a positive effect of competition; import penetration attributes a positive effect to openness.5 These results are consistent with recent empirical evidence on the determinants of total factor productivity in Japan at the firm level (Okada, 2005).
The Phillips curve equation relates inflation to the past values of the output gap, terms of trade shocks (changes in import prices and oil prices), and expected inflation. The estimated inflation dynamics are related positively, but weakly, to the output gap, corroborating evidence of a flat Phillips curve. The sacrifice ratio—the cumulative change in output required to change inflation permanently by 1 percentage point—is estimated at 8.7, around five times that for the United States. The direct pass-through of oil prices to headline CPI inflation appears small at about 0.1 percent for every 10 percentage points’ change in oil prices, in line with Japan’s low energy intensity. The direct pass-through of import prices is five times that of oil prices, but still small and consistent with the share of imported goods in the CPI.
The NAIRU specification relates the structural unemployment rate to the share of the old in the labor force, the replacement ratio (defined as the ratio of unemployment benefits to wages and salaries), and the past unemployment rate. The generosity of the unemployment insurance system tends to increase search time between jobs and the reservation wage, and the aging of the labor force tends to worsen skills mismatches, increase rigidities through seniority based-pay scales, and hamper the reallocation of workers.
The Okun’s law relationship links changes in the unemployment rate to those in the output gap. Cyclical fluctuations in the product market bring about adjustments in the labor market, although very slowly in comparison with other advanced economies (the estimate for Japan is 0.01; comparable estimates for other advanced economies range between 0.1 and 0.2).
Figure 2.1.Potential Output and NAIRU
Source: IMF staff estimates.
Figure 2.2.Inflation, Unemployment Gap, and Output Gap
Source: IMF staff estimates.
Figure 2.3.Potential Output and its Components—Baseline
Source: IMF staff estimates.
Potential output growth is estimated to have increased steadily since 2001, reaching about 1.6 percent in 2005, from around 1 percent in 2001. Nonetheless, potential output growth remains well below estimated levels attained during the 1980s, where it was close to 4 percent.
The pick-up in potential output is mainly attributable to a rise in total factor productivity growth—the outcome of an improved use of resources and increased competition. TFP growth is estimated to have accelerated to 1¼ percent in 2005, from less than ¼ percent in 1998 (Figure 2.4). The positive contribution of total factor productivity to potential output growth has helped offset a declining contribution of capital inputs and a negative contribution of labor inputs.
The contribution of the capital stock has been declining since the collapse of the investment bubble in the early 1990s. Growth in the capital stock now contributes just over ½ percentage point to potential output growth compared with more than 2 percentage points in the early 1990s. This decline reflects in part corporate sector restructuring which has involved delaying new investment and disposing of old or inefficient capital stock.
Labor inputs continue to contribute negatively to potential output growth. The contribution of employment, which has been declining since 1990, has been negative since the mid 1990s, reflecting a shrinking working-age population since 1999, a plateau in trend labor force participation rate since 2000, and a secular rise in structural unemployment. The negative contribution of employment has been partly offset by a positive contribution of the number of hours worked, as a result of a recent pick-up in growth of full-time jobs.
These results are somewhat sensitive to the measure of the capital stock (Figure 2.5). The capital stock series used in the estimation above is from the Japan Industry Productivity (JIP) database and differs from the official System of National Accounts series calculated by the Cabinet Office. It is based on a perpetual inventory method and corrects for the depreciation in the economic value of the capital stock.6 This depreciation can be relatively rapid in the information technology (IT)-related sectors, which have grown in importance during the 1990s: not accounting for it could lead to an overstatement of the level of—and return on—the capital stock. Indeed, using estimates of the Cabinet Office capital stock in lieu of those from the JIP database yields slightly higher potential output growth (by 0.1-0.2 percentage points) and smaller contributions from total factor productivity growth.
Figure 2.4.Contributions to Annual Potential Output Growth
Source: IMF staff estimates.
Figure 2.5.Potential Output and its Components—Sensitivity Analysis
Source: IMF staff estimates.
What are the determinants of productivity?
Sectoral data suggest that the recent pick-up in total factor productivity growth reflects improvements across most sectors of the Japanese economy, particularly manufacturing. Part of the improvements in sectoral TFP are likely to reflect Japan’s cyclical recovery since the five-year averages used in this analysis have not been detrended. Nevertheless, some key stylized facts emerge:
Productivity growth in the manufacturing sector averaged 3¾ percent between 2000 and 2004, up from almost zero percent on average between 1995 and 1999. Within the manufacturing sector, there have been large improvements in TFP growth in IT-related sectors such as “electrical machinery, equipment and supplies,” “precision instruments,” and “machinery” (Figure 2.6). These developments are consistent with the findings by Jorgenson and Motohashi (2005) that the IT sector’s contribution to aggregate productivity growth has increased since the mid 1990s.
The real estate industry, which represents over 10 percent of GDP, has also contributed significantly to the rise in productivity growth. Total factor productivity growth in the real estate industry recovered to an average of ½ percent during 2000-04, compared with minus 3¾ percent during 1995-99. The finance and insurance industry, which accounts for over 6¼ percent of GDP, added significantly to the momentum in aggregate TFP.
However, gains in aggregate TFP growth have been somewhat held back by developments in the wholesale and retail and “other services” sectors, which now account for just over a third of total output and slightly under 50 percent of total employment. This negative impact is reinforced by developments in the construction industry, which continues to contribute negatively to aggregate productivity growth, albeit less than during 1995-99.
Figure 2.6.Contribution TFP Growth to Sectoral Real GDP Growth
Source: IMF staff estimates.
In general, improvements in TFP growth have translated into labor productivity gains in most industries, despite less capital deepening (Figure 2.7). Capital deepening (defined as growth in capital input per hours worked) slowed, reflecting structural imbalances in the financial and corporate sectors.
Figure 2.7.Labor Productivity Growth by Industry
Source: IMF staff estimates.
What is the likely impact on potential output of reforms undertaken in recent years?
Empirical analysis suggests that the recent improvement in total factor productivity stems in part from greater product market competition, higher openness, and increases in R&D spending in relation to GDP. Regression results in Table 2.1 above suggest that reducing mark-ups by 1 percentage point stimulates TFP growth by about the same amount; raising import penetration by 10 percentage points increases TFP growth by about ¼ percentage point; and increasing R&D intensity by 1 percentage point raises TFP by broadly the same amount.
With reforms, potential output growth could be raised further. The results suggest that, going forward, the removal of lingering product market distortions—e.g. cutting excessive domestic regulation, strengthening the anti-trust framework, and further liberalizing trade (specifically, agricultural)—together with R&D investment could boost total factor productivity and potential output growth.
How such reforms above might spur potential growth can be illustrated by contrasting two scenarios over 2006-11 (Figures 2.8):
Figure 2.8.Potential Output Growth Over the Medium Term
Source: IMF staff estimates.
In a baseline scenario (Table 2.2), it is assumed that total factor productivity growth returns to its 2000-05 average of around 1 percentage point. With no excess capacity utilization, the capital stock is assumed to rise in relation to GDP at around its trend rate of about 2 percentage points a year in net terms, contributing positively to potential output growth. The contribution of labor inputs to potential output growth remains negative, reflecting demographic trends. Indeed, the assumed rise in the labor force participation rate (from about 78 percent to around 80 percent by 2011) is not enough to compensate for the adverse impact of the decline in the working-age population. Moreover, reflecting the rising share of the elderly in the labor force, the “natural” unemployment rate rises, reducing over time the contribution of labor to potential output growth. As a result, potential growth declines to 1.7 percent by 2011, after peaking at 1.9 percent in 2008-09.
In the alternative scenario (Table 2.3), while the baseline assumption on the evolution of the capital stock is maintained, it is assumed that product market reforms lead to greater competition, higher openness, and greater R&D intensity.
|Potential Output Growth||1.5||0.6||0.5||1.1||1.5||1.6||1.7||1.9||1.9||1.9||1.8||1.7|
|Potential Output Growth||1.5||0.6||0.5||1.1||1.5||1.6||2.1||2.3||2.4||2.4||2.4||2.3|
— Competition improves, with mark-ups declining at the same pace as the average of the past five years. Trade intensity rises, with import penetration assumed to increase broadly in line with trend, helping to close part of the gap vis-à-vis other OECD countries.7 In contrast, because Japan has one of the highest level R&D expenditure in relation to GDP (currently around 3.5 percent), ranking third after Finland and Sweden, it is assumed that R&D intensity remains at current levels.8
— Under such assumptions, total factor productivity growth is lifted by about ¼ percentage points relative to the baseline scenario.
— In addition, the scenario assumes a moderate increase in female participation rate (currently around 61 percent) toward that of the average of the United States and the United Kingdom (around 69 percent). The female participation rate increases relative to the baseline scenario by 2¾ percentage points to 64 percent over the projection period, adding around ¼ percentage more to potential output growth.
— Overall, potential output growth rises by about ½ percentage point relative to the baseline scenario, reaching 2.3 percent by 2011.
The estimated magnitude of the impact of structural reforms on longer-term growth is very much in line with other recent estimates. Indeed, one official report (METI, 2006) considers that greater competition, efficiency gains in public services, and further IT diffusion and R&D could boost longer-term growth to 2¼ percent (from its current level of 1½ to 2 percent). This order of magnitude for the gains from reforms is, however, at the lower end of estimates of the dynamic medium-term payoffs in various studies, ranging from 0.3 to 2.4 percentage points.9 Nevertheless, even under conservative assumptions on the size of the payoffs, it is clear that further structural reforms could go a long way to help Japan respond to the challenges posed by its aging population and support strong growth and higher living standards in the years ahead.
One methodology for estimating potential output is the production function approach, which links output to inputs of labor and capital and total factor productivity. Under this approach, the current level of potential output is thought to be that which would emerge from the aggregate production function, given the current levels of fixed inputs and sustainable levels of variable ones. Many researchers have used this methodology to estimate potential output and its determinants for Japan, with results depending on model specification and the degree of data disaggregation.
Different applications of the production function approach have, in the event, not added much to the precision of the estimates as the uncertainty in pinning down potential output turns out to be a tradeoff for uncertainty about total factor productivity. In essence, this uncertainty arises from how the sustainable levels of the factor inputs are derived as well as measurement issues, related in part to the aggregation of data across industries. For example, while both the Cabinet Office (CAO) and the Bank of Japan (BoJ) use the traditional production function approach to estimate potential output, the two institutions have differed in how they define factor inputs, resulting in markedly different potential output estimates. The CAO defines potential output as the level that would emerge from the production function, given the current levels of fixed inputs and sustainable levels of variable inputs; while the BoJ defined (until recently) potential GDP as the level of output that would result from variable inputs at full employment.
Another method for estimating potential output involves filtering. In this approach, time-series techniques are used to fit trend lines to the data, and these trends provide the sought measures of underlying “equilibrium” values for the variables of interest. The trend values are used to define “gaps”—deviations of observed values from their trends—that are, in turn, used to describe the dynamics of the system. The trend lines are estimated, at least in part, by their ability to represent closely the observed time series.
The methodology used in this chapter combines the two approaches described above—the production function and a filtering technique. It uses information from both the supply side and the demand side to condition the estimates of potential output. The essential idea behind this approach is that information can be gained by considering more than just the output data. In particular, since there is a link between labor input and output, it may be useful to exploit information about the degree of excess demand in the labor market. Similarly, the evolution of inflation carries information about the likely existence of excess demand/supply in the product market.
Our methodology treats the filtering problem as a small system, where the estimates of potential output, trend labor participation, hours worked, capacity utilization, the NAIRU, and some of the parameters of the dynamic model are determined simultaneously. Thus, the interactions between unemployment, output, variable inputs, and inflation are taken into account. The resulting trend estimates of output, variable inputs, and the unemployment rate are interpreted as the levels consistent with a stable inflation rate.
The system consists of four structural equations (see Appendix 2.2)—a production function, a Phillips curve, a NAIRU, an Okun’s law—and several identities.
The production function, Equation (4), links output to hours worked, capital, and total factor productivity, with the share of hours worked and capital fixed at their 1995-2002 average of about two-thirds and one-third, respectively. At potential, hours worked is the product of the working-age population (as a share in total), a trend participation rate, the trend average of hours worked, and one minus the NAIRU. The trend participation and average hours worked, as well as the NAIRU, are determined simultaneously to be consistent with stable inflation. The capital stock series is the product of the estimated capital stock and the trend capacity utilization rate. The capital stock series is from the JIP database. Total factor productivity depends on R&D intensity, the degree of competition, the degree of openness, and past realizations of total factor productivity.
The Phillips curve, Equation (2), relates inflation to expected inflation, terms of trade shocks (changes in import prices and oil prices), and past values of the output gap. The influence of excess demand is captured through the output gap. This specification is a backward-looking, autoregressive model that has been employed extensively to estimate the parameters of reduced-form expectations-augmented Phillips curves. Inflation expectations are modeled as a pure distributed lag of past inflation, with a restriction that the coefficients sum to one. Needeless to say, an alternative specification for the inflation expectations process could alter our gap estimates. The influence of import prices and oil prices pass-through are also added to the inflation process.
The NAIRU equation, Equation (7), relates the unemployment rate to the share of the old in the labor force, the replacement ratio, and past unemployment rate. The first variable aims to capture the impact of demographic changes on structural unemployment, while the second variable aims to capture that of the generosity of the unemployment insurance system. High replacement ratios can raise the structural unemployment rate by lowering the gap between the income from work and the out-of-work support. The replacement ratio does not, however, fully represent the generosity of the unemployment system as it does not account for conditions on benefits eligibility and other requirements. Hysteresis in the labor market is captured through the inclusion of the lagged unemployment rate which introduces persistence in the dynamics of unemployment (de Serres, 2003).
Okun’s Law, Equation (8), links changes in unemployment to those in the output gap. Some degree of persistence in the dynamics of the unemployment gap is captured by the presence of lagged values. By the same token, the resource utilization equation (Equation (11)) links the capacity utilization rate to the output gap, with excess demand translating into tighter capacity.
(1) Output decomposition
(2) Phillips curve equation
(3) Unemployment rate ;
(4) Stochastic process for potential output
(5) Potential capital stock
(6) Stochastic process for the output gap
(7) Stochastic process for the NAIRU
(8) Stochastic process for the unemployment gap
(9) Capacity utilization
(10) Stochastic process for trend capacity utilization
(11) Stochastic process for the capacity utilization gap
(12) Potential labor input
(13) Hours worked.
(14) Stochastic process for trend hours worked
(15) Stochastic process for the hours worked gap
(16) Participation rate
(17) Stochastic process for trend participation rate
(18) Stochastic process for the participation rate gap
Definition of variables
y is the (100 times) the logarithm of real GDP (2000 base), spliced using 1993 SNA data before 1994
ygap is the output gap
π is (400 times) the quarterly percent change in the CPI index (using the difference in logarithms as an approximation)
πimp (400 times) the quarterly percent change in the implicit import deflator (using the difference in logarithms as an approximation)
πoil is (400 times) the quarterly percent change in the World Economic Outlook crude oil price index defined as a simple average of three spot prices: Dated Brent, West Texas Intermediate, and the Dubai Fateh in US dollars (using the difference in logarithms as an approximation)
k is the capital stock (alternatively from the JIP database and the Nomura database)
h is the average weekly hours worked.
part is the participation rate
u is the unemployment rate
Δrrep is the change in the replacement rate defined as the ratio of unemployment insurance benefits to salaries and wages
old is the share of old
pop is the working-age population
cu is the capacity utilization rate
cugap is the capacity utilization gap
BayoumiTamim and CharlesCollyns1999 “Where Are We Going? The Output Gap and Potential Growth” inT.Bayoumi and C.Collyns eds. Post-Bubble Blues-How Japan Responded to Asset Price CollapseInternational Monetary Fund.
BenesJaromir and PapaN’Diaye2004 “A Multivariate Filter for Measuring Potential Output and the NAIRU: Application to the Czech Republic” International Monetary FundWorking Paper series No. 04/45.
BradfordScott2003 “Paying the Price: Final Goods Protection in OECD Countries” Review of Economics and StatisticsFebruary2003.
CitrinDaniel A.1991 “Potential Output and the Natural Rate of Unemployment:Recent History and Medium-term Prospects” unpublished manuscript. International Monetary Fund.
De SerresAlain2003 “Structural Policies and Growth—A Non Technical Overview” Organization for Economic Cooperation and Development Working Paper series No. 355.
FukaoKyojiTomohikoInuiHirokiKawai and TsutomuMiyagawa2003 “Sectoral Productivity and Economic Growth in Japan: 1970-98: An Empirical Analysis Based on the JIP Database” ESRI Cabinet Office Discussion Paper Series No. 67.
HaraNaokoN.HirakataY.InomataS.ItoT.KawamotoT.KurozumiM.Minegishi and I.Takagawa2006 “The New Estimates of Output Gap and Potential Growth Rate” Bank of Japan Review No. 2006-E-3.
HayashiFumio and EdwardPrescott2002 “The 1990s in Japan: A Lost Decade” Review of Economic Dynamics Vol. 5 No. 1 pp. 206–35.
Japan Center for Economic Research2003 “Regulatory Reform and Demand Creation” December2003.
JorgensonDale and KazuyukiMotohashi2005 “Information Technology and the Japanese Economy” Journal of the Japanese and International Economies Vol. 19 pp. 460–81.
LaxtonDouglas and PapaN’Diaye2002 “Monetary Policy Credibility and the Unemployment Inflation Tradeoff: Evidence from 17 Industrial Countries” International Monetary FundWorking Paper series No. 02/220.
Ministry of Economy Trade and Industry (METI)2006 “New Economic Growth Strategy Interim Report” March.
Organization for Economic Cooperation and Development (OECD)2004OECD Economic Surveys: Japan Vol. 2003/18.
Organization for Economic Cooperation and Development (OECD)2006OECD Economic Surveys: Japan.
ShimpoSeiji and FumiharaNishizaki1997 “Measuring the Effects of Regulatory Reform in Japan: A Review” Economic Planning Agency.
A 2006 government report, “Japan’s 21st Century Vision,” sets out the importance of raising productivity and reaping the benefits of globalization to avoid deteriorating living standards.
The system of equations (1)-(18) in Appendix 2.2 has been estimated with the constrained maximum likelihood procedure applying the methodology described in Laxton and N’Diaye (2002) and Benes and N’Diaye (2004) over the period 1964:Q1-2005:Q4.
Innovation is often perceived as the main channel through which competition and openness affect on productivity growth.
The data were provided by the Bank of Japan. The official capital stock series assumes that additional investment is fully productive for a fixed number of years and is then eliminated. See Hara and others (2006) for details.
Japan has the lowest level of import penetration amongst OECD countries notwithstanding growing trade linkage with China and other Asian economies. See OECD (2006).
The OECD average R&D intensity is 2.2 percent based on 2003 data.