Australia: Selected Issues and Statistical Appendix
Author: International Monetary Fund
Publication Date:
03 Mar 2000
eISBN:
9781451801965
Language:
English
Keywords:
ISCR; CR; saving; household saving; Australia; productivity performance; productivity-cycle peak; growth rate; market-sector TFP growth; productivity leader; log productivity level; market sector
Download PDF (3.9 MB)

The first part of the study investigates Australia’s recent growth and productivity performance and then analyzes the impact of structural reforms on productivity growth. The short- and long-term effects of structural reforms on productivity growth are estimated using pooled and fixed effect distributed lag models. The study also includes the following statistical data: labor market, selected price index, selected fiscal indicators, credit aggregates, money supply, banking soundness statistics, current account, exports and imports, exports by commodity group, direction of trade, capital and financial account, interest rates, and so on.

Abstract

The first part of the study investigates Australia’s recent growth and productivity performance and then analyzes the impact of structural reforms on productivity growth. The short- and long-term effects of structural reforms on productivity growth are estimated using pooled and fixed effect distributed lag models. The study also includes the following statistical data: labor market, selected price index, selected fiscal indicators, credit aggregates, money supply, banking soundness statistics, current account, exports and imports, exports by commodity group, direction of trade, capital and financial account, interest rates, and so on.

I. Australia: Productivity Growth and Structural Reform1

A. Introduction

1. Economic growth in Australia has averaged almost 4½ percent during 1994-98, substantially higher than the average in the previous two decades and approaching average growth rates experienced in the “golden age” of the 1960s TableI.1 and Figure I.1). The recent performance is attributable both to faster growth in hours worked and to a pickup in labor productivity, capital productivity, and total factor productivity (TFP) growth, with most measures of these reaching rates that exceed those in the 1960s.

Table I.1.

Australia. Output, Inputs, and Productivity During Productivity Cycles 1/

(Average annual growth during fiscal years)

article image
Sources: Australian Bureau of Statistics (ABS) and Fund staff estimates.

Data are for fiscal years which end in June, Productivity cycles are dated by the ABS. Cycle peaks are identified by the maximum deviation of TFP from its long-run trend, which is estimated using an 11-period Henderson moving average.

Aggregate hours worked and labor productivity are calculated from first date available (1979).

TFP (or multi-factor productivity, MFP) is estimated by the ABS using a Thornqvist-Theil divisia index with TFP growth as the weighted average of labor and capital productivity growth, where the weights are, respectively, labor and capital income shares averaged over consecutive years.

Figure I.1.
Figure I.1.

Australia: Output, Inputs, and Productivity Growth

(Percent, fiscal year annual averages)

Citation: IMF Staff Country Reports 2000, 024; 10.5089/9781451801965.002.A001

Sources: Australian Bureau of Statistics; and Fund staff estimates.

2. This chapter examines Australia’s recent growth and productivity performance from two standpoints. First, it investigates the contribution of cyclical factors to the recent pickup, and second, it examines to what extent the structural or trend improvement in productivity growth is attributable to structural reforms. Regarding the latter, the paper analyzes the impact of structural reforms on productivity growth in a panel study of 20 OECD countries during 1965-98.

3. The analysis finds that the recent pickup in productivity growth reflects both cyclical and structural factors. The paper then attempts to link—by means of a cross-country study—the structural improvement to microeconomic reforms that have been pursued since the 1980s. The analysis suggests that, in the long run, structural reforms exert a significant positive impact on productivity growth, although the short-run impact may be weak or negative, possibly due to adjustment costs and the need for firms to learn how to operate in a less regulated and more competitive environment. Relative to the rates in the 1980s, the analysis suggests that reforms have lifted trend total factor productivity growth by between 0.5 and 0.9 percentage points. Under reasonable assumptions, this increase in trend TFP growth would imply that potential output growth in Australia over the next four to six years is likely to fall in the range of 3.2 to 4.3 percent. While the range is relatively wide—which underscores the need to continue to set policies with some uncertainty regarding the level of Australia’s potential growth rate—the midpoint is nevertheless higher than previous staff estimates.

B. Productivity Growth in the Market Sector:2 Cyclical and Trend Factors

4. During the most recent productivity cycle (1994-98), annual output growth in the market sector increased to 4.6 percent, compared to a (33-year) long-run average of percent, while annual labor productivity and TFP growth increased to 3.1 percent and percent, compared to long-run averages of 2.3 percent and 1.4 percent, respectively. 3,4 Annual capital productivity growth in the market sector also rose to 0.8 percent during 1994-98, compared to a long-term average annual decline of 1.0 percent.

5. Because productivity growth is positively correlated with output growth, however, the pickup in productivity may partly reflect cyclical factors. For example, market sector labor productivity growth has a 0.69 correlation with market sector output growth, while capital productivity and TFP growth are even more highly correlated with output growth (Table I.2). These positive correlations may be explained by labor hoarding or fluctuating capacity utilization. Alternatively, advocates of real business cycle theory would argue that the causation mainly runs from productivity growth to output growth—that is, positive technology shocks produce economic booms. In any case, an underlying trend productivity growth that excludes cyclical factors can be estimated. This paper employs three methods to estimate trend growth: average growth between productivity-cycle peaks; average growth between business-cycle peaks; and growth estimated by smoothing the underlying series to filter out business cycle fluctuations using the Hodrick-Prescott (1997) filter.

Table I.2.

Australia. Growth Correlations 1/

(Correlation with output growth; 1965-98)

article image
Sources: Australian Bureau of Statistics and Fund staff estimates.

Data are for fiscal years which end in June.

Correlations for aggregate hours worked and labor productivity are calculated from first date available (1979).

6. All three methods indicate that trend productivity growth has risen, but each one provides a different estimate of the trend growth rate. Trend productivity growth estimated by the average growth between productivity-cycle peaks (see Table I.1) or between business-cycle peaks (Table I.3) rose in the most recent economic expansion to levels even higher than those in the so-called “golden age” of the 1960s and early 1970s.5 In the most recent productivity cycle, for example, annual labor productivity growth in the market sector reached 3.1 percent, annual TFP growth 2.4 percent, and annual capital productivity growth 0.8 percent, while in the most recent business cycle, annual labor productivity growth increased to 2.9 percent, annual TFP growth 2.0 percent, and annual capital productivity growth 0.1 percent. These growth rates exceed or match those experienced in the 1960s and are well above the productivity growth in the 1980s. Moreover, labor productivity and TFP growth in the most recent expansion, unlike the previous two expansions, have remained strong even eight years after the previous business cycle peak (Figure I.2).

Table I.3.

Australia. Output, Inputs, and Productivity During Business Cycles 1/

(Average annual growth during fiscal years)

article image
Sources: Australian Bureau of Statistics (ABS) and Fund staff estimates.

Data are for fiscal years which end in June.

Aggregate hours worked and labor productivity are calculated from first date available (1979).

Figure I.2.
Figure I.2.

Australia: Market Sector Productivity Growth During Business Cycles1

(Percent, fiscal year annual averages)

Citation: IMF Staff Country Reports 2000, 024; 10.5089/9781451801965.002.A001

Sources: Australian Bureau of Statistics; and Fund staff estimates.1 Time t=0 is at business cycle peaks(1965,1974,1982,1990)

7. Trend market-sector productivity growth, estimated using the Hodrick-Prescott filter, has also risen, although generally to a lesser extent (Figure I.3). In 1998, trend labor productivity growth was 2.4 percent—only slightly higher than its long-term average and substantially below trend rates in the 1960s, while trend TFP growth was 1.8 percent—still slightly below trend rates in the 1960s, although almost 1 percentage point higher than its low point in the late 1980s. However, trend capital productivity growth increased to 0.5 percent, its highest level. In addition, labor productivity has risen faster than would be predicted by the long-run relationship between it and the capital-labor ratio, also indicating that TFP growth has increased above its long-run trend. 6

Figure I.3.
Figure I.3.

Australia: Market Sector Trends

(Fiscal year averages)

Citation: IMF Staff Country Reports 2000, 024; 10.5089/9781451801965.002.A001

Sources: Australian Bureau of Statistics; and Fund staff estimates.1 Trend productivity is estimated using the Hodrick-Prescott Filter.2 OLS regression of the log level of labor productivity on the log level of the capital-labor ratio and a time trend for the period 1965-98.3 OLS regression of the log level of labor productivity on the log level of the capital-labor ratio and a time trend with a Cochrane-Orcutt correction for autocorrelated residuals.

8. The data presented above, therefore, provides some indication that productivity growth, even controlling for cyclical factors, has risen in recent years. It is difficult, however, to estimate the precise improvement. For trend or structural market-sector TFP growth, the estimates from the three methods range from 1.8 percent to 2.4 percent. In addition, the estimates of trend growth that are based on averages between productivity cycles or business cycles may be biased upwards because the current cycle is not yet complete and productivity growth tends to decrease towards the end of such cycles. The Hodrick-Prescott filter also suffers from end-period problems. Moreover, it remains unclear whether the improvement in productivity growth is sustainable or possibly the result of a somewhat longer than normal, but not pathbreaking, expansion.7

C. Productivity Growth and Structural Reforms: A Cross-country Analysis

9. Over the past two decades, Australia has implemented a wide range of structural reforms, including trade, product market, and labor market reforms. In terms of trade reform, average tariff rates have been reduced from over 12 percent in the mid-1970s to under percent in 1998 (Figure I.4), while rates of effective protection have also declined and are projected to continue to decline (Industry Commission, 1995 and Productivity Commission, 1998). On the domestic side, key sectors—such as financial services, telecommunications, and aviation—have been liberalized, other product markets have been deregulated to enhance domestic competition, and labor markets have been reformed and decentralized. 8

Figure I.4
Figure I.4

Astralia: Structural Indicators1

(Percent)

Citation: IMF Staff Country Reports 2000, 024; 10.5089/9781451801965.002.A001

Sources: OECD; IMF International Financial Statistics; and Fund Staff estimates.1 For explanation of construction of indicators, see data appendix.

10. To examine the potential impact of these structural reforms, this paper analyzes productivity growth and indicators of structural reforms across OECD countries since the 1960s. 9 Only OECD countries are included in the analysis to maintain a set of relatively homogeneous countries. A variety of structural indicators are examined, with the average tariff rate as the primary indicator of trade reform, 10 and the average unemployment benefit replacement rate as the indicator of labor market reform. Several indicators for product market reform are used as proxies for product market competition, namely structural change variables and the price-average-cost markup. 11

11. The choice of indicators or proxies for structural reforms is limited by the availability of data for a number of countries, particularly on an annual time-series basis. This means that the proxy measures used in the empirical analysis in this paper may not give optimal measures of structural reforms undertaken in a particular country. For example, as an indicator or proxy for trade reform, the average tariff rate does not capture the benefits of removing import quotas or other forms of nontariff barriers to trade. Also, because it is (implicitly) trade-weighted (as opposed to production-weighted), the proxy does not fully capture the trade protection offered even by tariffs. 12 For Australia, effective rates of protection, a better proxy for trade protection, have been calculated for agriculture and manufacturing (for example, see Industry Commission, 1995), but these measures are not available for most other countries (particularly, on a time-series basis).

12. There is also only limited panel data on indicators of labor market flexibility and reform. Unemployment benefit replacement rate data are perhaps not optimal because the data indicate that labor market flexibility has declined in most OECD countries, including Australia. While this might indicate less labor market flexibility, other labor market indicators (which are, unfortunately, available only for selected years and therefore, not usable in the empirical analysis below) generally indicate slightly improving or constant labor market flexibility (OECD, 1999).

13. Nevertheless, these data limitations do not necessarily invalidate the results. The proxy measures will capture some of the effects of reform, and the deviations of these proxies from true measures of reform may be random across countries—that is, the deviations may cancel. In addition, the results presented below are tested for robustness using different proxy measures and different specifications.

International comparison of Australia’s performance

14. Estimates of labor productivity, capital productivity, and TFP growth for the business sector can be derived for 20 OECD countries between 1960 and 1998. 13 Because hours worked are available for only a limited number of countries, labor productivity is calculated as output per worker. TFP is calculated using the same methodology as employed by the Australian Bureau of Statistics (ABS), under the assumptions of constant returns to scale and perfect competition. 14 Furthermore, labor and capital are assumed to be homogeneous and fully employed. 15 To the extent that these assumptions are incorrect, inputs and output are mismeasured, and hours per worker change over time, TFP growth will be an inaccurate measure of technological progress and improving economic efficiency.

15. With these caveats in mind, the data show that productivity growth in these OECD countries, including Australia, generally slowed down in the early 1970s. Annual labor productivity growth, which averaged over 4½ percent in the 1960s across the 20 countries, slowed to about half that rate in the 1970s and further during the 1980s and 1990s (Table I.4). Average annual TFP growth followed a similar pattern, with a sharp drop from almost 3 percent in the 1960s to about 1 percent in the 1980s and 1990s (Table I.5). While average annual capital productivity growth also slowed by half a percentage point between the 1960s and the 1970s, it has since picked up in the 1990s (Table I.6). In the most recent decade, labor productivity and TFP growth appear to have partly rebounded in some countries, including Australia, but are still below levels reached in the 1960s. Interestingly, capital productivity in Australia has remain almost unchanged over the long run (and has risen recently), suggesting that investment inefficiencies—possibly arising from credit market or other distortions—are not plaguing Australia’s economic performance. 16

Table I.4.

Selected Industrial Countries: Business Sector Labor Productivity Growth 1/

(Annual averages)

article image
Sources: OECD and Fund staff estimates.

Labor productivity is calculated as output per employee.

Table I.5.

Selected Industrial Countries: Business Sector TFP Growth 1/

(Annual averages)

article image
Sources: OECD and Fund staff estimates.

TFP growth is calculated using a Thornqvist-Theil divisia index as the weighted average of labor and capital productivity growth, where the weights are, respectively, labor and capital income shares averaged over consecutive years.

Table I.6.

Selected Industrial Countries: Business Sector Capital Productivity Growth 1/

(Annual averages)

article image
Sources: OECD and Fund staff estimates.

Capital productivity is calculated as output per capital input.

16. During the last several decades, many OECD countries have instituted structural reforms. On the trade side, average tariff rates across these countries have declined from almost 9 percent in 1960 to under 2 percent by 1995 (Table I.7). Product markets have also become more competitive, as indicated by price-average-cost markups which have declined from a cross-country average of about 20 percent in the 1960s to 11 percent in 1995 ( Table I.8). However, average unemployment benefits replacement rates have increased from 16 percent in 1960 to almost 30 percent in 1995 (Table I.9). For product markets, the structural change variables are relatively volatile ( see Figure I.4).17

Table I.7.

Selected Industrial Countries: Average Tariff Rates 1/

(In percent)

article image
Sources: OECD and IMF, International Financial Statistics.

For explanation of construction of Tariff Rates, see data appendix.

Table I.8

Selected Industrial Countries: Price-Average Cost Markup 1/

(In percent)

article image
Source: OECD and Fund staff estimates.

For explanation of construction of Price-Average Cost Markup, see data appendix.

Table I.9

Selected Industrial Countries: Unemployment Benefit Replacement Rate 1/

(In percent)

article image
Source: Blanchard and Wolfers (1999).

For explanation of construction of Unemployment Benefit Replacement Rate, see data appendix.

Results

17. The short- and long-run effects of structural reforms on productivity growth are estimated using pooled and fixed effect distributed lag models. The explanatory variables are the structural indicators (described above) and a term allowing for convergence of productivity levels. For most regressions, lags range from 1 to 10. 18 As the data are annual, ten-period lags may seem long; however, one objective of this study is to estimate the long-run impact of structural reforms, and indeed, the coefficients for the ten-year lagged variables were often found to be significantly different from zero.

The estimated equations have the following form:

yi,t=ai,t+Σj=nkmkβi,jTxt,t−j+εi,t

where y is the dependent variable, a is the constant term, x is a (kxl)-dimensional vector representing the explanatory variables, β is a (kxl)-dimensional vector representing the coefficients for the explanatory variables (with T representing the transpose of the vector), e is the error term, k is the number of explanatory variables (excluding the constant term), irepresents the cross-sectional units (in this case, countries), j represents the number of lags, rtk and ntk represent the range of the lags, and t represents time periods. The dependent variables are productivity growth or more specifically, first differences of the log productivity levels. The explanatory variables are in log levels or first differences of log levels (as specified in the Tables). For the pooled regressions, αi = α and βi = β for alli, while for the fixed effects regressions αi’s were not constrained to be equal.

18. The results are presented in Tables I.10-I.14. In Tables I.10-I.12, the dependent variable is TFP growth, while in Table I.13, the dependent variable is labor productivity growth, and in Table I.14, capital productivity growth. In Tables I.10,I.13, and I.14, the explanatory variables are the ratio of per capita income to per capita income in the United States, the productivity leader (Gap), the tariff rate (Trade), the price-average-cost markup (Product), and the unemployment benefits replacement rate (Labor). 19 In Table I.12, the price-average-cost markup is replaced as the proxy for product market reform by the (two- digit) structural change variable. 20 In interpreting these results, it should be noted that for the indicators of structural reform, other than the structural change variables, a negative coefficient implies a positive impact of reform on productivity growth. (For the structural change variables, a positive coefficient implies a positive impact). A negative coefficient on the Gap variable implies convergence in productivity.

Table I.10

Industrial Countries: Impact of Structural Reforms on TFP Growth 1/

article image

Explanatory variables for the regressions are: Gap = Per capita income relative to the United States; Trade = Tariff rate; P-AC = P-AC markup; and Labor = Replacement Rate. Lags from 1 to 10 for all variables and regressions. Short-run is coefficient on first lag, while long-run is sum of the coefficient on all lags. P-values for T-statistic on short-run coefficients and F-statistic (for null hypothesis that all coefficients are zero) on long-run coefficients in parentheses. The underlying standard errors are White heteroskedasticity consistent.

Differences for structural indicators only. Gap is always in log levels.

Table I.11.

Industrial Countries: Impact of Structural Reforms on TFP Growth

(Excluding Labor Variable) 1/

article image

Explanatory variables for the regressions are: Gap = Per capita income relative to the United States; Trade = Tariff rate; P-AC = P-AC markup; and Labor = Replacement Rate. Lags from 1 to 10 for all variables and regressions. Short-run is coefficient on first lag, while long-run is sum of the coefficient on all lags. P-values for T-statistic on short-run coefficients and F-statistic (for null hypothesis that all coefficients are zero) on long-run coefficients in parentheses. The underlying standard errors are White heteroskedasticity consistent.

Differences for structural indicators only. Gap is always in log levels.

Table I.12.

Industrial Countries: Impact of Structural Reforms on TFP Growth

(With Structural Change Variable) 1/

article image

Explanatory variables for the regressions are: Gap = Per capita income relative to the United States; Trade = Tariff rate; P-AC = P-AC markup; and Labor = Replacement Rate. Lags from 1 to 10 for all variables and regressions. Short-run is coefficient on first lag, while long-run is sum of the coefficient on all lags. P-values for T-statistic on short-run coefficients and F-statistic (for null hypothesis that all coefficients are zero) on long-run coefficients in parentheses. The underlying standard errors are White heteroskedasticity consistent.

Differences for structural indicators only. Gap is always in log levels.

Table I.13.

Industrial Countries: Impact of Structural Reforms on Labor Productivity Growth1/

article image

Explanatory variables for the regressions are: Gap = Per capita income relative to the United States; Trade = Tariff rate; P-AC = P-AC markup; and Labor = Replacement Rate. Lags from 1 to 10 for all variables and regressions. Short-run is coefficient on first lag, while long-run is sum of the coefficient on all lags. P-values for T-statistic on short-run coefficients and F-statistic (for null hypothesis that all coefficients are zero) on long-run coefficients in parentheses. The underlying standard errors are White heteroskedasticity consistent.

Differences for structural indicators only. Gap is always in log levels.

Table I.14

Industrial Countries: Impact of Structural Reforms on Capital Productivity Growth 1/

article image

Explanatory variables for regressions are: Gap = Per capita income relative to the United States; Trade = Tariff rate; Product = P-AC markup; and Labor = Replacement Rate. Lags from 1 to 10 for all variables and regressions. Short-run is coefficient on first lag, while long-run is sum of the coefficient on all lags. P-values for T-statistic on short-run coefficients and F-statistic (for null hypothesis that all coefficients are zero) on long-run coefficients in parentheses. The underlying standard errors are White heteroskedasticity consistent.

Differences for structural indicators only. Gap is always in log levels.

19. The estimation results generally indicate that fixed effects matter—that is, there are differences in performance across the countries even after controlling for the effects of the convergence term and the structural indicators. 21 In addition, the labor variable is rarely ever significant—never in the short run and only a few times in the long run. In Table I.11, the labor variable is dropped (but the other explanatory variables are the same as in Table I.10). As a proxy for product market reform, the price-average-cost markup is significant more often in these regressions than the structural change variable. The relative income per capita term indicates convergence in the short and long run in almost all the regressions. 22

20. Over the short term, structural reforms appear to have a weak or even negative impact on productivity growth. In general, the short-run coefficient (or the coefficient on the first lag of the explanatory variable) is insignificant or does not have the expected sign. This result holds in almost every regression and may be explained by the short-run costs of adjusting to reform and the need for firms to learn how to operate in a deregulated environment. While this result provides some support for a strategy of phasing or sequencing the introduction of structural reforms, the analysis indicates that the negative effects are generally reversed within two years.

21. In the longer run, specifically after ten years, the results indicate that trade and product market reforms have a positive impact on productivity growth, as the long-run coefficient (or the sum of the coefficients on all lags) typically has the expected sign and the F-statistic that the coefficients on the lags are different from zero is usually significant.23

22. An estimate of the long-run impact of the structural reforms for particular countries can be made by multiplying the long-run regression coefficient (for regressions with log levels) or the long-run coefficient divided by the lag length (for regressions with first differences of log levels) by the change in the structural indicators. Such an estimate is inherently rough because, as discussed above, while the deviations of the proxy measures from the true measures of structural reforms may be washed out across countries, these deviations may not be random for individual countries. With this caveat in mind, this methodology implies that structural reforms that have been implemented in Australia during the last decade could lift TFP growth between 0.5 and 0.9 percentage points over the long run.24 As a comparison, the estimates of the long-run impact of structural reform range from 0.3 to 0.4 percentage points for New Zealand and from 0.1 to 0.2 percentage points for the United States.

23. While some of this improvement in Australia’s TFP growth is already apparent in the data, productivity growth should continue to strengthen above what would have occurred without these reforms, as the full impact of recent structural reforms—including, for example, recent or planned reductions in tariffs and trade protection and the National Competition Policy which was adopted in 1995—may not be felt for as long as a decade. In addition, the result in this study that labor market reforms, in general, do not lead to improvements in productivity growth might be because the unemployment benefit replacement rate, which has increased in most OECD countries including Australia, is a poor indicator of labor market flexibility and reform. Because labor market reforms have often been implemented along with and as a complement to other structural reforms in many of these countries, and because the unemployment benefit replacement rate may be a poor indicator of the labor market reforms, the positive impact of the labor market reforms may be included in the estimated positive effects of the trade and product market reforms. Indeed, recent labor market reforms in Australia (for example, the Workplace Relations Act of 1996), which have improved labor market flexibility, could also lead to improved productivity growth.

24. Positive effects of trade and product market reforms have also been found in a study of the impact of these reforms on productivity growth across 14 OECD countries during 1970-90 (Economic Planning Advisory Commission, 1995).25 In addition, Chand et al. (1998) find that declining assistance to manufacturing in Australia is positively related to manufacturing industry productivity growth.

Australia’s potential growth rate

25. Rough estimates can also be made of the impact on potential growth over the next four to six years based on the estimated improvement in TFP growth owing to structural reforms. These calculations assume that the underlying production function is Cobb-Douglas, TFP growth and employment growth are exogenous, and Australia is on its steady-state (or balanced) growth path. Under these assumptions, potential output growth is equal to employment growth added to TFP growth divided by the labor income share. Over the medium term (or through 2005), employment is assumed to grow 1.8 percent per annum. 26 Market-sector TFP growth without structural reforms (ranging from about 0.8 percent to 0.9 percent) is assumed equal to trend growth in the 1980s estimated using the three methods, as discussed in Section II. The estimated improvement in TFP growth because of structural reforms ranges from 0.5 to 0.9 percentage points, as discussed above. Labor income share (equal to about 67 percent of GDP) is estimated from OECD data as the average share during the past decade. Furthermore, because the market sector excludes some private service sectors, government administration and defense, ownership of dwellings, and indirect taxes and subsidies, an additional adjustment is made to translate market sector productivity growth to aggregate productivity growth. These calculations indicate that potential growth would range from 2.4 to 2.9 percent without the structural reforms but range from 3.2 to 4.3 percent with the reforms.

D. Conclusion

26. In recent years, productivity growth in Australia has increased to rates not seen since the golden age of the 1960s. This chapter has examined the contribution of both cyclical and structural factors to the performance, and has attempted to link the structural improvement to a variety of microeconomic reforms implemented since the 1980s.

27. The analysis found that, while cyclical factors explain part of the improved productivity performance, even controlling for these, there has been an improvement. It is difficult, however, to quantify the structural or trend improvement precisely, which underscores the need to continue to attach sizeable uncertainty bands around a point estimate of Australia’s potential growth rate in setting policies.

28. The analysis further suggests that structural reforms, particularly trade and product market reforms, are important in explaining improvements in trend productivity growth, even though the impact of such reforms on productivity may be weak or negative in the short run, possibly due to adjustment and learning costs. The results suggest that structural reforms have lifted Australia’s trend TFP growth rate by between 0.5 and 0.9 percentage points since the 1980s. Under some reasonable assumptions, this increase in trend TFP growth would imply that potential output growth in Australia over the next four to six years is likely to fall in the range of 3.2 to 4.3 percent. The midpoint of this range is significantly higher than previous staff estimates of Australia’s potential growth rate.

APPENDIX I: Data Appendix

Data sources: Cross-country analysis

The main sources for the data in the cross-country analysis are OECD databases, including the Analytical Database (AD), the Economic Outlook Database (EOD), and the Structural Analysis (STAN) industrial database. In addition, tariff rates are calculated based on tariff revenues mainly from the OECD Revenue Statistics and imports from the IMF, International Financial Statistics (IFS). Unemployment benefit replacement rates are from Blanchard and Wolfers (1999), who derived the rates from the OECD’s Database on Unemployment Benefit Entitlements and Replacement Rates.

Data construction: Cross-country analysis

  • Labor productivity is calculated as output per employee and provided in the EOD. The series (PDTY) is indexed to 1991. An unindexed series can be constructed as well by calculating labor productivity in 1991 as the ratio of real GDP (GDPV) to total employment (ET).

  • Capital productivity is calculated as output per capital input using data from the EOD. Specifically, capital productivity is the ratio of business sector real GDP at factor cost (GDPBV) to business sector capital stock (KBV).

  • TFP is calculated using a Thornqvist-Theil divisia index with TFP growth as the weighted average of labor and capital productivity growth, where the weights are, respectively, labor and capital income shares averaged over consecutive years.

  • Labor income share is calculated using data from the EOD as the ratio of the product of compensation of employees (WSSS) and ET to the product of dependent employees (EE) and GDP at market prices (GDP) excluding net indirect taxes (TIND-TSUB). Capital income share is 1 minus labor income share.

  • Relative per capita income is calculated using data from the EOD as the ratio of U.S. dollar valued per capita real GDP using 1991 purchasing power parity (PPP) exchange rates (GDPVD/POP) to GDPVD/POP in the United States. Relative TFP is also calculated using data from the EOD and using PPP exchange rates,

  • Tariff rate is calculated as the ratio of Customs and Import Duties (from OECD Revenue statistics supplemented with data from IMF, Government Finance Statistics) to imports (from the IFS).

  • Openness is calculated as the ratio of the sum of exports and imports (from the IFS) to GDP (from the EOD).

  • Import penetration is calculated as the ratio of imports to apparent domestic consumption, which is the sum of domestic production and imports less exports, and export intensity is the ratio of exports to domestic production. At the aggregate level, the sources for the data are the IFS and the EOD, while for the manufacturing sector, the source is the STAN database.

  • Price-average-cost markup is calculated using data from the EOD and AD as the ratio of GDP at market prices less net indirect taxes to the sum of labor income (WSSS*ET/EE) and capital income, where capital income is constructed as the product of the real capital stock, capital price deflator (PIT), and the real rental rate for capital, which (following Hall and Jorgenson, 1967 and Martins and Scarpetta, 1999) is the real interest rate plus depreciation (respectively, IRLRE and RSCRB in the AD).

  • Structural change is calculated using data in the STAN database as half the sum of the absolute value of annual changes in share of GDP and is calculated at two- and three- digit industry levels.

References

  • Blanchard, O. and J. Wolfers, 1999, “The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence,” (unpublished). Available via the Internet: http://web.mit.edu/blanchar/www/articles.html (March).

    • Search Google Scholar
    • Export Citation

  • Chand, S., P. McCalman, and P. Gretton, 1998, “Trade Liberalisation and Manufacturing Industry Productivity Growth,” Chapter 8 in Productivity Commission and Australian National University, Microeconomic Reform and Productivity Growth, Workshop Proceedings (Auslnfo: Canberra) (July).

    • Search Google Scholar
    • Export Citation

  • Domowitz, I, R.G. Hubbard, B.C. Peterson, 1986, “Business Cycles and the Relationship Between Concentration and Price-cost margins,” Rand Journal of Economics, Vol. 17, pp. 1-17.

    • Search Google Scholar
    • Export Citation

  • Domowitz, I., R.G. Hubbard, B.C. Peterson, 1988, “Market Structure and Cyclical Fluctuations in U.S. Manufacturing,” The Review of Economics and Statistics, Vol. 70, pp. 55-66.

    • Search Google Scholar
    • Export Citation

  • Economic Planning Advisory Commission, 1995, Tariff Reform and Economic Growth (Australia).

  • Hall, R.E., 1988, The Relation between Price and Marginal Cost in U.S. Industry,” Journal of Political Economy, Vol. 96, pp. 921-47.

  • Hall, R.E. and D.W. Jorgenson, 1967, “Tax Policy and Investment Behavior,” The American Economic Review, Vol. 57, pp. 391-414.

  • Hodrick, R.J. and E.C. Prescott, 1997, “Postwar U.S. Business Cycles: An Empirical Investigation” Journal of Money, Credit, and Banking, Vol. 29 pp. 1-16.

    • Search Google Scholar
    • Export Citation

  • Industry Commission, 1995, Assistance to Agricultural and Manufacturing Industries (Australia Government Publishing Service: Canberra).

  • Martins, J.O. and S. Scarpetta, 1999, “The Levels and Cyclical Behaviour of Mark-ups across Countries and Market Structures,” OECD Economics Department Working Paper No. 213.

    • Search Google Scholar
    • Export Citation

  • Morrison, C, 1990, “Market Power, Economic Profitability, and Productivity Growth Measurement: An Integrated Structural Approach,” NBER Working Paper No. 3355.

    • Search Google Scholar
    • Export Citation

  • OECD, 1996, OECD Economies at a Glance: Structural Indicators (Paris).

  • OECD, 1999, OECD Employment Outlook (Paris).

  • Parham, D., 1999, The New Economy? A New Look at Australia’s Productivity Performance, Productivity Commission Staff Research Paper, (Auslnfo: Canberra).

    • Search Google Scholar
    • Export Citation

  • Productivity Commission, 1998, Trade and Assistance Review 1997-98, Annual Report Series 1997-98, (Auslnfo: Canberra).

  • Roeger, W., 1995, “Can Imperfect Competition Explain the Difference between Primal and Dual Productivity Measures? Estimates for U. S. Manufacturing,” Journal of Political Economy, Vol. 103, pp. 316-30.

    • Search Google Scholar
    • Export Citation

  • Schweitzer, M.E., 1998, “Productivity Gains During Business Cycles: What’s Normal?” Economic Commentary (Federal Reserve Bank of Cleveland).

    • Search Google Scholar
    • Export Citation

  • Singh, A., J. Felman, R. Brooks, T. Callen, and C. Thimann, 1998, Australia: Benefiting from Economic Reform, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
1

Prepared by Ranil Salgado. Since this chapter was finalized, the ABS has published a revised experimental series—going back to the 1960s—on capital service flows, which would affect the productivity estimates in this chapter. The upshot of the revisions would be somewhat lower market-sector total factor productivity growth during 1965-98, but an increase since the 1980s of approximately the same magnitude as identified below.

2

The market sector, as defined by the Australian Bureau of Statistics (ABS), excludes five industry sectors—government administration and defense, property and business services, education, health and community services, and personal and other services (and also excludes ownership of dwellings). The ABS provides estimates of total (or multi-) factor productivity and capital productivity only for the market sector. The data from the ABS refer to fiscal years ending in June.

3

It should be noted that strong output and labor productivity growth continued through June 1999. The paper includes ABS data only through June 1998 because capital stock data—and therefore, estimates of capital and total factor productivity—were not available for 1999 at the time of writing.

4

TFP (or multi-factor productivity) is estimated by the ABS using a Thornqvist-Theil divisia index with TFP growth as the weighted average of labor and capital productivity growth, where the weights are, respectively, labor and capital income shares averaged over consecutive years.

5

The ABS estimates productivity growth during productivity cycles in order to remove cyclical factors. Cycle peaks are identified by the maximum deviation of TFP from its long- run trend, which is estimated using an 11-period Henderson moving average. In this paper, the standard international methodology that estimates productivity growth during business cycles, which are identified by cyclical peaks in output, is also used. See Parham (1999) for more discussion about recent productivity trends in Australia.

6

Assuming a Cobb-Douglas production function (i.e., constant returns to scale and a constant and unit elasticity of substitution between capital and labor), labor productivity is a log-linear function of TFP and the capital-labor ratio.

7

See, for example, Schweitzer (1998) which examines the current U.S. expansion and concludes that the recent increase in U.S. productivity growth is remarkable, but not unusual, in that the recent performance is not outside standard error bands when compared to previous expansions. One difficulty with providing a similar analysis for Australia is that historical data is more limited and in particular, covers only four business cycles.

8

For more details, see Australia: Benefiting from Economic Reform (IMF, 1998).

9

See Appendix for data sources and details of variable construction.

10

Other indicators of trade reform are also examined, including openness, import penetration, and export intensity—the latter two, both for manufacturing industries only and at the aggregate level.

11

Structural change is defined as half the sum of the absolute value of annual changes in the sectoral share of GDP, and is calculated at two- and three-digit industry levels—see OECD (1996). Price-average-cost markups, which are allowed to vary on an annual basis, are defined as the ratio of nominal GDP (excluding net indirect taxes) to total factor cost–see Domowitz et al. (1986) for a similar approach. Alternative approaches, which estimate price- marginal-cost margins but assume constant margins over time, include Hall (1998), Domowitz et al. (1988), and Roeger (1995). Morrison (1990) proposes a methodology to estimate time-varying markups using a structural model; however, this procedure is beyond the scope of this paper.

12

As an example, prohibitive tariff rates afford full trade protection, but the implied protection would not be included in the average tariff rate because sectors with prohibitive tariffs would have no imports.

13

These countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the United Kingdom, and the United States. While estimates for labor productivity growth can be derived for other OECD countries, the lack of capital stock data for these countries precludes estimating capital productivity or TFP growth. The business sector includes both market and nonmarket sectors, but excludes producers of government services.

14

At the aggregate level, most economic studies have found that constant returns to scale cannot be rejected. For tests of these assumptions with Australian aggregate data, see Chapter 2 in Australia: Benefiting from Economic Reform (IMF, 1998).

15

For example, labor is not differentiated by level of skill or education while capital is not differentiated by vintage.

16

In general, if investment is inefficient, then capital productivity (or also, the marginal product of capital) would be expected to fall.

17

Note that unlike the other indicators, an increase in the structural change variable indicates increased market competition or flexibility.

18

For the structural change explanatory variables, the lag length is 6, as the coefficients beyond the sixth lag were insignificant. This may reflect the shorter time series (from 1971) available for these variables. Also, regressions (not shown) without the Gap variable (see below for the description of this variable) or with the Gap variable lagged only one period or only ten periods also confirm the main regression results, particularly on the impact of structural reforms.

19

Similar results (not shown) are found when the relative per capita income term is replaced by the ratio of the TFP level to the TFP level in the United States, although convergence in the long run is generally rejected.

20

Regressions (not shown) substituting the three-digit structural change variable instead of the two-digit one produce similar results, except the three-digit structural change variable is almost always insignificant in both the short and long run.

21

The fixed effects for Australia are usually small and positive and often insignificant.

22

The regressions (not shown) with the Gap variable lagged only one period or only ten periods also generally indicate convergence.

23

For the regressions with log levels, the long-run coefficients can be interpreted as the long- run impact of a one-unit change in the log level of the explanatory variable. For the regressions with first differences of log levels, the long-run coefficients can be interpreted similarly, except that because the coefficient is calculated as the sum of the coefficients on all of the lags and the explanatory variables are first differences of log levels, the coefficient must be divided by the lag length in order to estimate the impact of a 1 unit change in the log level of the explanatory variable.

24

These calculations are based on the fixed effects regressions with first differences in Tables 1.10-1.12, using the long-run coefficients significant at the 10 percent level. The range for the impact on TFP growth is because of the differences in the coefficients. Trade reforms generally account for about 80 percent of the impact because the coefficient on trade reforms is higher.

25

The Economic Planning Advisory Commission study differs from this paper in a number of respects, including: (1) not modeling the dynamic effects of structural reform so as to differentiate between short-run and long-run effects; (2) using only pooled estimation (although a few country dummies are included in some regressions); (3) having a smaller sample (both in cross section and time series); and (4) using only the structural change variable as an indicator of product market reform.

26

Employment growth is calculated from estimates of annual labor force growth (1.6 percent through 2005) along with a decline in the unemployment rate to 6.3 percent.

APPENDIX II: Data Appendix

Household saving: saving by households and unincorporated enterprises to GDP. Ratios of gross and net savings to GDP were used in the text table, and the ratio of household net saving to household disposable income was used in the time-series regression. ABS 5206-52 (original) and ABS 5206-23 (seasonally adjusted).

Household disposable income. ABS 5206-52 and ABS 5206-23.

Public saving: general government saving. ABS 5206-55 (original) and ABS 5206-27 (seasonally adjusted).

National saving. ABS 5206-49.

Corporate saving: enterprise saving (national saving minus household and general government saving).

FinDereg: the ratio of household debt to household disposable income, normalized so that the starting value equals 0 and the end-value equals 1. Household debt was based on the Reserve Bank Bulletin table D02 (Lending and Credit Aggregates).

Inflation rate: underlying inflation rate.

Real interest rate: the difference between 10-year government bond yield and the inflation rate.

Social assistance: the ratio of social assistance benefits to household disposable income. ABS 5206-23.

Unemployment rate. ABS 6202-5.

Wealth: private wealth. The variable used in the paper is the ratio of private wealth to household disposable income. ABS TRYM Table 33.

STATISTICAL APPENDIX

Table 1.

Australia: Selected National Accounts Aggregates at 1997/98 Prices, 1994-99 1/

article image
Source: Australian Bureau of Statistics, National Accounts.

Quarterly data are seasonally adjusted.

Includes real estate transfer expenses.

Includes livestock and intangible fixed assets.

Contribution to GDP growth, at annual rates.

Table 2.

Australia: Sectoral Components of Gross Domestic Product at 1997/98 Prices, 1994-99 1/

article image
Source: Australian Bureau of Statistics, National Accounts.

Quarterly data are seasonally adjusted.

Includes defense.