RDP 2005-08: Declining Output Volatility: What Role for Structural Change? 3. Methodology and Data

This paper uses a fixed-effects panel data regression with output volatility as the dependent variable and measures of structural change as the independent variables. Data are annual from 1974 to 2003, except for the indicator of product market regulation, which is available only about every five years from 1978 to 1998 (the early 1980s observation is for 1982). Partly for this reason regressions are run with observations over five-year blocks. Output volatility is measured as the standard deviation of the annual growth rate of GDP within each five-year block; this and other key data are summarised in Table 1 (and in Figure B1 in Appendix B, which also includes a description of data sources). Blanchard and Simon (2001) also measure volatility according to the standard deviation of GDP growth rates, though they use quarterly data and a rolling 5-year window (in Section 4.1 we test the sensitivity of our results by adopting a variant of this rolling window approach).^[7]

A key innovation of this paper is to examine the role of direct measures of economic structure in explaining the volatility of output. We define a direct measure as one which is closely tied to the actual regime/structure in place, as opposed to an indirect measure, which is a consequence of that regime/structure. One direct measure we consider is an index of product market regulation produced by the OECD, which provides an internationally comparable measure of the degree to which government policies inhibit competition. This index covers regulations related to barriers to entry (including legal and administrative barriers to entrepreneurship), public ownership, market structure, vertical integration and price controls (for more details see Appendix B; Nicoletti et al 2001; and Nicoletti and Scarpetta 2003). The index ranges from high regulation (6) to limited regulation (0).

The other direct structural measure we examine relates somewhat loosely to the ‘effectiveness’ or ‘strictness’ of the monetary policy regime, which ultimately affects the level and volatility of inflation.^[8] This is measured by a dummy variable, which takes a value of 1 if the regime is deemed to be strict on inflation and 0 otherwise. As a benchmark, Germany, Japan and Switzerland are assumed to have had strict regimes throughout the sample period.^[9] Monetary policy in the US is deemed to have become strict starting from the Volcker chairmanship and continuing through that of Greenspan. For all other countries, policy is deemed to have been strict during periods when they were either tied closely to Germany through membership of the Exchange Rate Mechanism (ERM) and later the euro area, or following the adoption of inflation-targeting regimes. The possibility that the ERM may not have been as effective as other strict monetary policy regimes, such as euro-area membership or inflation targeting, is also explored by including a separate ERM dummy variable.

As shown in Table 2, the crude dummy variable measure of monetary policy regimes appears to be related to both the level and standard deviation of inflation; across all countries, average inflation and average volatility of inflation fell substantially when moving to the stricter regime. This is also true of most countries individually, with the exception of Sweden.

Ideally, we would also include a direct measure of labour market regulations in the regressions; however, a useful measure is not readily available.^[10] Hence, we use a proxy based on the number of days lost in labour disputes. This shows a trend decline across most countries, which appears to be consistent with the variation in the extent of labour market reforms across countries. Further, because the approach to industrial relations reform has been quite different across countries, an outcome-based measure may be better than a direct measure. For example, Wooden and Sloan (1998) show that while Australia and the UK adopted different approaches to labour market reform, they have resulted in very similar labour market outcomes. Nicoletti, Scarpetta and Boylaud (1999) note that for 1998, there is a significant positive cross-country correlation between indices of employment protection legislation and product market regulations, suggesting that the latter might also proxy for labour market regulations in the regression analysis (the correlation between product market reforms and days lost in labour disputes is 0.26; see Table 3).

Other indirect structural measures considered are openness to international trade (proxied by the ratio of exports and imports to GDP) and financial liberalisation (proxied by the ratio of private sector financial assets or liabilities to GDP). Also, inflation volatility can be used as an indirect measure of the effectiveness of monetary policy regimes.^[11] Finally, controlling for any effects due to changes in the behaviour of fiscal policy is achieved by including the volatility of the cyclically-adjusted primary budget balance (as a ratio to GDP). This measure of discretionary policy is preferred over the primary budget balance, which is endogenous with respect to output volatility since it includes the effect of automatic stabilisers.^[12]

The distinction between direct and indirect structural indicators is relevant for the lag structure in the regressions. For direct measures we match the volatility of annual GDP growth over a given five-year period with the value of the structural indicator that applies in the year just prior to this (for example, output volatility over the five years ending 1983 is matched with the level of the product market regulations index in 1978). This captures the likely lagged effect of structural change, as well as having the desirable property of ensuring that the structural indicators are exogenous with respect to output volatility. In contrast, indirect measures of structural indicators are included in the regressions contemporaneously, consistent with other studies of this type. Finally, we control for one type of supply shock by including the volatility of oil prices contemporaneously.

In summary the basic regression takes the following form:

where: is the standard deviation of annual growth of real GDP for country i; X_it is a vector of direct structural indicators; Z_it is a vector of indirect structural indicators; W_t is a vector of other possible explanators, such as oil price volatility and a time trend; and t indicates each five-year block ending in 1983, 1988,…, 2003.

Simple correlations across the panel using data in five-year blocks are generally consistent with the graphical analysis in Figures 2, 3, 5 and 6. Most notably, the lag of product market regulation is positively correlated with output volatility, the lagged monetary policy regime dummy is negatively related to output volatility, and a decline in days lost due to labour disputes is associated with a decline in output volatility. Greater trade openness is associated with a (contemporaneous) rise in output volatility, while financial liberalisation is negatively related to output volatility. Volatility in inflation and oil price growth are positively related to output volatility. Of the cross-correlations among explanatory variables, the largest in absolute terms is the −0.47 correlation between the lagged direct measure of the strictness of monetary policy and the indirect inflation volatility measure. Looking at the correlation between the direct measure of product market regulation (lagged) and the three relevant indirect measures, the largest in absolute terms is with financial liberalisation (−0.45) followed by days lost to labour market disputes (0.26); the correlation with openness (0.14) is relatively low and positive (suggesting that, overall, this measure of openness may not be adequately capturing the trend towards lower trade barriers).

Footnotes

Barrell and Gottschalk (2004) examine a measure of GDP volatility based on the standard deviation of the output gap. Estimates for Model 1a based on a measure of the volatility of the output gap (constructed by applying an HP filter to the log of GDP) produce a similar coefficient for product market regulations (though with a p-value of 0.15) and a larger (absolute) and statistically significant coefficient for the monetary policy regime (−0.88). Other coefficient estimates are qualitatively similar. While the average trend across countries of output volatility based on the output gap is similar to that based on output growth, it generally displays greater short-term volatility within countries. [7]

Bergman, Bordo and Jonung (1998) link different monetary policy regimes to changes in output volatility. They distinguish four regimes: the Gold Standard, the inter-war period, Bretton Woods and post-Bretton Woods. However, they fail to find any significant relationship, possibly because their regimes are too broadly defined; in particular, the post-Bretton Woods period captures an array of quite different policy regimes. [8]

It could be argued that Japanese monetary policy has not been so effective over our full sample. Even if we alter our assumption and deem Japan's monetary policy regime to have been ineffective throughout the period, the results of the paper are essentially unchanged. [9]

The Economic Freedom of the World Index provides an overall measure of labour market regulations. While useful for cross-country comparisons, it tends to understate the degree of reform within countries over time – indeed, for Australia this measure suggests that the labour market was more regulated in recent years compared with the early 1990s notwithstanding significant reform over this period (Dawkins 2000). This may reflect the fact that this measure (and others like it) is only able to capture a limited set of factors that determine how the labour market operates, and it tends to rely heavily on subjective interpretations of the legal framework. [10]

Barrell and Gottschalk (2004) and Blanchard and Simon (2001) find that the level of inflation is insignificant in explaining changes in output volatility. [11]

Commodity price volatility was also examined. The results of Maccini and Pagan (2005) might suggest that the coefficient would be positive, in line with trend declines in the volatility of output and commodity prices. However, the coefficient estimate is negative (and significant). The inclusion of commodity price volatility pushes up the coefficient estimate on product market reform, possibly due to a multicollinearity problem (the correlation between the two variables is 0.65). Hence, commodity prices are ignored in what follows. [12]