Business Cycle Accounting Analysis of the Japanese Economy

A new method for analyzing economic fluctuations called "business cycle accounting" (BCA) has been developed in recent years. Using this method, I analyzed data on the Japanese economy from the 1980s through 1990s to look into the true cause of the prolonged recession of the 1990s. Some results of the analysis follow (Kobayashi and Inaba, 2005).

As a major tool of macroeconomic analysis of economic fluctuations, real business cycle (RBC) models have been developed based on the groundbreaking research of Kydland and Prescott (1982). BCA, as explained hereunder, is a methodology that is in a sort of dual relation with the RBC method.

The key characteristic of the RBC method is in the explicit modeling of the rational optimizing behavior of consumers and corporations. A nation's economy is represented by a dynamic general equilibrium model in which consumers seek to maximize their utility and corporations compete to maximize their profits. Macro variables such as aggregate consumption, investment and price changes of the whole economy are calculated as consequences resulting from the optimizing behavior of consumers and corporations. In this regard, RBC models are quite different from conventional Keynesian models where it is presumed that macroeconomic factors such as consumption and investment are determined by ad-hoc functions of interest rates and income.

In the real economy, however, consumers and corporations behave in anticipation of government policies and future economic conditions. Thus, in the event of major changes in government policies or in the economic conditions, the configuration of a consumption function estimated from a Keynesian model would change (the Lucas critique). The conventional Keynesian models were unable to take such major changes into account. But the RBC method, whereby the optimizing behavior of consumers and so forth has been put into a model, has overcome this problem.

The standard methodology of RBC is as follows. First, a general equilibrium model will be developed by theoretically assuming factors supposedly causing economic fluctuations (time required for the completion of the project after capital investment, information asymmetry between borrowers and lenders, and so forth). Then, realistic parameter values will be calibrated for a computer analysis by numerical calculation. Finally, the resulting figures -- which supposedly show changes in production, consumption, investment, etc. -- will be compared against the actual fluctuations in the economy to examine the degree of approximation.

Because RBC is merely a computer simulation model, its calculation results would not perfectly match actual economic data. The degree of similarities between the model-produced and actual data -- for instance in the covariance between variables -- determines the validity of the model. In short, the RBC methodology is to set up a theoretical model and examine the closeness of the model-produced results to real data.

In contrast, BCA seeks to explore "what characteristics (or 'wedges' as described later) must be added to a theoretical model in order to fully explain real data." That is, the reversal of the RBC approach is the BCA approach.

This idea was announced almost simultaneously by Chari, Kehoe and McGrattan (2002) and by Mulligan (2002). What follows will explain BCA mainly along the lines set by Chari, Kehoe and McGrattan.

Like RBC, BCA assumes that a nation's economy can be approximated by a neoclassical dynamic general equilibrium model. Consumers and corporations are to make decisions on consumption, investment, labor input and so forth as a result of their optimizing behavior. A function that is supposed to work if consumers and corporations are facing no wedges in selecting each of these variables (i.e. making decisions) can be theoretically derived from the model. For instance, provided that labor input remains efficient, the marginal rate of substitution between consumption and leisure (MRS, i.e., the increase in consumption for which a consumer is willing to work by sacrificing one extra unit of leisure) should theoretically be equal to the marginal productivity of labor (MPL, i.e., the increase in production from each additional unit of labor). However, when labor input becomes inefficient for whatever reason, MRS and MPL do not match. The degree of discrepancy between MRS and MPL shows the size of the wedge for labor input.

Likewise, wedges for capital investment and productivity can be measured by comparing data calculated by the function that assumes no wedge and the real data on capital investment and productivity.

Conversely, when a simulation is performed on the general equilibrium model by assuming the thus-calculated wedges, the resulting data will be identical to the real economic data (because the value of each wedge included in the function has been determined in such a way that the function with the real data on labor input or so forth would be proven true). In other words, BCA is for decomposing changes in the real economy into several kinds of wedge effects and thus the real data can be reproduced by aggregating all such wedge effects. This is why BCA is referred to as an accounting approach to explaining an actual business cycle.

With respect to each of the measured wedges for each sector of the economy, model simulations can be performed to show how economic performance would change with or without the wedge. If there is little difference in the results with or without the investment wedge, for example, it can be judged that distortions in investment had little impact on the economic performance. As such, by performing simulations, the wedges that were crucial contributing factors to the prolonged recession in Japan can be inferred.

However, it should be noted that even if each of the wedges has been measured by applying the BCA method, this does not necessarily provide a clear picture of the theoretical mechanism that has caused these wedges. Such a mechanism must be inferred or identified by some method. What follows will present the results of applying the BCA method to data on the Japanese economy and inferring the cause of major wedges.

Figure 1 shows the simulation results for the Japanese economy. First, the value for each of the following wedges -- which are plausible factors contributing to the recession -- was calculated:

1. labor wedge
2. investment wedge
3. efficiency wedge
4. government wedge

Each line graph in Figure 1 represents the degree of impact that either the presence of or changes in a wedge has had on the gross national product (GNP).

Figure 1: Four wedges in the Japanese economy and GNP

The bold line 1 shows real economic output as measured by GNP. The straight benchmark line 2 represents the hypothetical GNP based on the assumption that the labor, investment, efficiency and government wedges are maintained at a level of their respective average for the period from 1984-1989. This serves as a benchmark for measuring changes in economic performance from the 1990s onward and is provided in the horizontal straight line so as to serve as a standard for the others. Thus, lines 1 and 3-6 indicate relative values against the benchmark GNP of 100.

First, the impact of the government wedge on GNP, that is, the impact of changes in the government expenditures on the economy, is shown by line 3. This line stays above the benchmark line throughout the 1990s, which means that fiscal expenditures continued to have a stimulative effect on the economy in the 1990s. It should be noted that this effect was not lost even in 1997 when the consumption tax rate was raised to 5%. Thus, the conventional view that the consumption tax hike triggered the deterioration of the economy does not match the BCA results.

Second, the impact of the efficiency wedge on GNP, which is the change in output attributable to changes in productivity, is shown by line 4. The shape of this line resembles that of line 1, which represents actual output, but line 4 remained above the benchmark until the late 1990s, the time when Japan was hit by a financial crisis. This suggests that the depressive effect of the decline in productivity was not so strong during the early and mid-1990s. One widely held theory attributes the prolonged recession to an unexplained decline in productivity. But the BCA results point to the possibility that the decline in productivity might have not been a major contributing factor.

Third, the impact of the investment wedge on GNP-- the change in the economy attributable to distortions in capital investment -- is shown by line 5. This line remained close to the benchmark in the 1990s, suggesting that a decline in private sector capital investment was not a major cause of the deterioration of the economy. This also contradicts somewhat the generally accepted notion. It is generally perceived that underinvestment by the private sector, which arose as many companies suffered from debt overhangs, was a major cause of the recession, yet the BCA results do not necessarily support this. At least, there is a possibility that the problem of the financial sector including nonperforming loans did not have an adverse effect on corporate behavior toward capital investment. At the same time, however, it is a plain fact that private-sector capital investment dropped sharply in the 1990s. But then, what caused this drop? It is conceivable, though not conclusive, that the deterioration of distortion in labor input was the cause. As discussed later, collateral constraints and falling asset prices are perceived to have caused the deterioration of the labor wedge, hence the depression of capital investment.

Finally, line 6 shows the depressive effect of labor input on the economy. It is clear that the labor wedge has been continuously deteriorating since the early 1990s. Based in comparison with the other wedges, it is conceivable that distortions in labor input were the biggest contributor to the prolonged recession.

From the results of the BCA analysis it has been understood that the increased distortion in labor input (i.e. the increasing discrepancy between MRS and MPL) was a major contributor to the prolonged recession. When it comes to the cause of the deterioration in labor input, however, there are several possible explanations.

One is that distortions in labor input occurred as a result of long-term structural changes in the Japanese economy. If so, changes in the labor wedge must have been occurring long before the recession. One illustration of this is when employees' bargaining power in wage negotiations is increasing as a result of a rise in their social status. This may be reflected in the data in the form of increasing distortion in labor input.

Figure 2: Changes in labor wedges in 1981-2002

Figure 2 shows the change in the labor wedge over the last two decades. As shown by the light line, the labor wedge began to deteriorate in 1984. However, this may be the result of measurement error. Indeed, recalculation using a modified production function^* gives the bold line, which shows that the deterioration of this modified labor wedge began only when the recession began. It is therefore possible that the original finding -- that the deterioration of the labor wedge began prior to the onset of the recession -- is the result of measurement error.

Even if we presume the deterioration of the labor wedge and the recession began at the same time, this would leave several possible mechanisms for causing the inefficiency of the labor input.

A standard explanation would point to sticky wages (or downward rigidity). This wage rigidity, combined with a deflationary shock such as a contractionary monetary policy, pushes up real wages, resulting in inefficient labor input. This explanation fits with what actually occurred in the early 1990s when the Bank of Japan tightened monetary policy sharply in the final phase of the bubble economy. But the explanation appears to contradict the data from the latter half of the decade. Although the labor wedge continued to deteriorate through the 1990s and beyond, real wages began to fall in the mid-1990s. Had the rise in real wages contributed to the deterioration of the labor wedge, the rise in real wages should have continued in the second half of the 1990s.

What then can account for the continued deterioration of the labor wedge in the latter half of the 1990s? The hypothesis that I am studying is that declines in land and share prices may have aggravated the labor wedge, with this negative impact coming through the channel of collateral constraints. A standard neoclassical model assumes that consumers decide on both consumption and investment (whereas corporations simply borrow capital from consumers and buy labor). Here, let us also assume collateral constraints, i.e., consumers need to borrow in order to finance the total cost of consumption and investment and they are required to pledge collateral in order to borrow. The model with this assumption of such collateral constraints shows that a decrease in collateral values depresses consumption, causing a decrease in labor input. That is, when asset values continue to fall in an economy where collateral constraints exist, it appears that the labor wedge deteriorates.

By comparison, investment would fall more sharply than consumption on a decrease in collateral values because consumers try to intertemporally smooth their consumption.

The above explanation seems convincing given that land-collateral loans are the primary means of financing in Japan and that land prices continued to fall even after the 1990s.

BCA analysis of the Great Depression in the U.S. also shows that the labor wedge sharply deteriorated from the 1930s onward. The Great Depression, considering the fact that it started with a sharp fall in asset prices, seems to have certain commonalities with Japan's recession in the past decade. In any event, it is highly conceivable that deterioration in labor input has some sort of significant relation with the occurrence of a Great Depression-type recession. Clarifying the mechanism linking these two phenomena is a major challenge in establishing a viable economic theory for prolonged recessions.

>> Original text in Japanese

Usually, in the Cobb-Douglas function (Y = A K^α L^1-α), the parameters for capital and labor are assumed to be constant over time. In this modification, it is assumed that these parameters are variable, and that they coincide with the shares of capital and labor for each year.