Journal Issue
Share
Article

Portugal—Selected Issues

Author(s):
International Monetary Fund
Published Date:
January 1998
Share
  • ShareShare
Show Summary Details

I. An Empirical Investigation of The Business Cycle2

A. Introduction and Summary

6. The characteristics of Portugal’s business cycle are perceived to have become increasingly similar to those of cycles in other advanced economies, while cyclical output fluctuations are believed to have remained comparatively more pronounced. This chapter investigates whether these commonly held views are supported by an empirical analysis using the Bank of Portugal’s newly released long-term economic time series. Specifically, it seeks to compare the amplitude and stability of output fluctuations in Portugal with those in larger European economies. It also examines the cyclical behavior of an array of macroeconomic data measuring aggregate demand, supply, and labor market conditions in Portugal and compares the results with those found in studies of other industrial countries. Finally, the chapter tests for changes in the nature of Portugal’s business cycle, and assesses the likelihood of continued relatively higher output variability in the period ahead.

7. An improved empirical understanding of Portugal’s business cycle is helpful both in projecting macroeconomic performance and the formulation of policy. More specifically, the finding that Portugal’s business cycle is relatively more volatile, and likely to remain so in comparison to its larger EU partners, in conjunction with the loss of the monetary policy lever under EMU, suggest the need for a relatively stronger fiscal position. The degree to which fiscal policy will need to assume a heightened countercyclical role has direct implications in determining the appropriate fiscal balance to be pursued in the medium term.

8. The evidence confirms that Portugal’s catch-up process to European real income levels has, as could be expected, been associated with output fluctuations that are considerably more marked than in larger European economies. The stylized facts of business cycles in Portugal nonetheless conform closely with those of other industrial countries. Specifically, all aggregate demand variables except government consumption move tightly with the output cycle and, as is the case for overall output, are highly persistent. Consumption and investment are procyclical, and net exports are countercyclical. Furthermore, investment and net exports are substantially more volatile than GDP. Employment is also procyclical but less volatile than real output.

9. The past four decades have seen significant shifts in the nature of Portugal’s business cycle, with the major driving force being the series of structural and policy changes that have occurred since the 1974 revolution, including notably the opening up and liberalization of the economy in the wake of EU accession in 1986. Output fluctuations exhibit high persistence, and the co-movement of output with other macroeconomic variables such as investment, net exports, and employment, now more closely resembles the co-movements found in other advanced economies. At the same time, the differences in Portugal’s recent economic and political history that set it apart from its EU partners have also made for a comparatively stronger rise in output volatility, not only with respect to the experience of large European economies but also to that of other southern European countries. The evidence presented here indicates that large European economies have not experienced a significant change in the amplitude of cyclical fluctuations during the 40 years studied. Other studies on southern European economies report either no significant change (Italy after 1973) or a decline in output volatility (Spain in the 1980s compared with the 1970s).

10. This study conducts a preliminary investigation of the factors underlying the observed rise in output volatility. There has, first and foremost, been a marked increase in exposure to foreign trade, together with much higher volatility of such trade. Furthermore, Portugal experienced a pronounced rise in price volatility, against the background of successive shifts in the stance of financial policies. The data also suggest that structural factors played a role in increasing output volatility. In a similar vein, previous work by the staff on the growth payoff of policy reform found that international trade was an important factor that had boosted trend growth in Portugal after EU accession.3 Taken together with the findings of the present study, the results would imply that trade integration has not only increased long-term growth—as would be predicted by theories of comparative advantages—but that it has also been associated with much larger fluctuations of output around its trend.

B. Methodology

Trend-cycle decomposition

11. The first problem that typically arises in the analysis of the business cycle is its measurement, which requires that the economic series be decomposed into trend and cycle. This mere distinction is controversial because the two components may be closely related and determined by the same underlying factors. Furthermore, there are many methods for decomposing time series into trend and cyclical components. Bearing in mind these qualifications, this chapter removes the potential nonstationarity in aggregate time series by means of low-frequency filtering, a procedure adopted by Hodrick and Prescott (HP, 1980), who propose a filter whose main attractiveness lies in its flexibility, simplicity, and reproducibility.4 Another advantage of the HP filter is that it has been used in several studies of business cycles in industrial economies, and has also been applied to the Portuguese economy by Correia, Neves, and Rebelo (1993).

12. The HP filter defines a trend τ for a series y as the solution to the problem:

Fluctuations are defined as deviations from trend, y − τ In equation (1) the parameter λ represents the choice between perfect smoothness of the trend (λ=∞), that is, a linear trend, versus perfect fit of the trend (λ=0), that is, the trend replicates the series. The benchmark value chosen in this study is λ=100 for all series, which is the most common value assigned when filtering annual data.

Statistical characterization of business cycles

13. A common theme of the line of research followed in this chapter is that business cycles consist not simply of fluctuations in aggregate output but also of common patterns of correlations between aggregate time series. To characterize economic fluctuations and patterns among variables, this analysis first obtains the cyclical components of the series, and then computes the following statistics:

(1) Volatility: The standard deviation of the filtered series is used as a measure of the amplitude of business cycle fluctuations. To compare any series’ volatility with that of real GDP, this study also presents the ratio of the standard deviation of the filtered series to the standard deviation of filtered real GDP.

(2) Persistence: The autocorrelation of the filtered series is used as a measure of persistence of cyclical deviations from trend.

(3) Co-movement: The correlation of a macroeconomic variable with real output (for lags up to three periods) is used as a measure of the variable’s co-movement with the output cycle. It indicates whether the macroeconomic variable is strongly or weakly cyclical, whether it moves pro- or countercyclically, and whether it leads, is coincidental, or lags real output.5 A variable is procyclical if the cross correlation is positive and countercyclical if it is negative. Finally, a variable leads the cycle if the maximum absolute value of the cross correlation occurs in (t−i), it is coincidental if it occurs in t, and it lags the cycle if the absolute correlation peaks in (t+i).

14. Data for this analysis are taken from the newly released long-term economic time series constructed by the Bank of Portugal. The data are annual and span the years 1953 to 1993. All variables are expressed in natural logarithms except inventories, which are expressed as a ratio to GDP, and net exports, the natural log of which is approximated by the following transformation: net exports / |average net exports| − 1.

C. Growth and Output Fluctuations in a European Context

15. To place the long-term properties of Portugal’s output growth in perspective, Table 1 (top section) summarizes means, standard deviations, and autocorrelation coefficients for output growth in Portugal and the combined output of three of the largest European economies: Germany, France, and the United Kingdom.6 In the four decades to 1993, Portugal’s economy grew by 4.6 percent annually on average, a full percentage point above the three large EU countries. Growth was also 40 percent more volatile than in these countries.

Table 1.Portugal: Properties of Growth and Cyclical Fluctuations
Output growth rates
Mean(Portugal)/StandardSD(post-74)/SD(Portugal)/
Mean 1/Mean(EU3) 2/Deviation 1/SD(pre-74)SD(EU3) 2/Autocorrelation
Portugal
GDP (1953-1993)4.631.302.971.410.37
GDP (1954-1971)4.850.963.312.09−0.03
GDP (1977-1993)4.271.992.580.781.940.58
EU3 2/
GDP (1954-93)3.562.100.57
GDP (1954-1971)5.061.590.16
GDP (1977-1993)2.151.330.840.59
Cyclical fluctuations
StandardSD(post-75)/SD(Portugal)/
Deviation 1/SD(pre-75)SD(EU3) 2/Autocorrelation
Portugal
GDP (1953-93)3.172.110.62
GDP (1953-1971)1.903/1.450.27
GDP (1977-1993)3.163/1.661.990.69
EU3 2/
GDP (1953-93)1.500.51
GDP (1953-1971)1.324/0.39
GDP (1977-1993)1.594/1.210.68
Cross correlation between output fluctuations in Portugal and in Germany, France and the UK.
T−3T−2T−1TT+1T+2T+3
GDP (1953-93)−0.48−0.190.160.570.670.440.04
GDP (1953-1971)−0.27−0.24−0.360.040.490.670.15
GDP (1977-1993)−0.66−0.190.370.760.870.700.08

In percent.

EU3 stands for the combined output of Germany, France and the United Kingdom.

The hypothesis that these two variances belong to the same distribution is rejected at the 95 percent confidence level. The confidence interval for the variance, assuming that the mean of the distribution in unknown, is distributed as Chi-square with (n-1) degrees of freedom.

The hypothesis that these two variances belong to the same distribution cannot be rejected at the 95 percent confidence level. The confidence interval for the variance, assuming that the mean of the distribution in unknown, is distributed as Chi-square with (n-1) degrees of freedom.

In percent.

EU3 stands for the combined output of Germany, France and the United Kingdom.

The hypothesis that these two variances belong to the same distribution is rejected at the 95 percent confidence level. The confidence interval for the variance, assuming that the mean of the distribution in unknown, is distributed as Chi-square with (n-1) degrees of freedom.

The hypothesis that these two variances belong to the same distribution cannot be rejected at the 95 percent confidence level. The confidence interval for the variance, assuming that the mean of the distribution in unknown, is distributed as Chi-square with (n-1) degrees of freedom.

16. In the 40 years covered by the sample, Portugal underwent many structural changes that clearly had a strong impact on its trend growth and cyclical fluctuations. A major political change with profound economic implications was the 1974 revolution, which ended an authoritarian regime that had been in place for almost half a century. In order to take the significant change in the political and economic environment duly into account, the sample is split in two subperiods. Furthermore, to abstract from the considerable turmoil surrounding the revolution, the sample excludes the two years before, and the two years following, 1974.

17. There was a marked difference in the characteristics of output growth after the revolution. First, although output growth slowed throughout Europe, the slowdown was much less pronounced in Portugal. After 1974, Portugal grew at twice the speed of the three large EU countries. Second, growth in Portugal became markedly persistent, indicating a lengthening of growth cycles.7 Indeed, after 1974, growth in Portugal became as persistent as growth in the three large EU economies.

18. As with growth, output fluctuations, namely, deviations from long-term trends, clearly differ in Portugal and the three large EU countries (Table 1, middle section). Output volatility in Portugal is twice that in the these countries. Furthermore, it increased by 66 percent after 1974 and deviations from trend became highly persistent. It would appear that the cumulative effect of the many structural and policy changes that occurred after the 1974 revolution changed the nature of economic fluctuations in Portugal. Specifically, the amplitude and length of cycles increased substantially (Figure 1, top panel) and the volatility of output cycles in Portugal relative to the three large European countries changed permanently, as shown by the plot of the nine-year rolling standard deviation of output fluctuations (Figure 1, lower panel).8 The switch from an inward-oriented economy and an inert political situation to one marked by increased openness, and a series of changes in political and economic approaches in the course of the democratization process, clearly induced a strong and lingering rise in output volatility. Output fluctuations remained comparatively greater even after the immediate effects of the 1974 shock had faded, while output volatility in the three large European countries did not rise significantly.

Figure 1.Portugal: International Comparison of Output Fluctuations

Sources: Bank of Portugal; IMF, World Economic Outlook; and staff calculations.

1/ Cyclical component of combined output series.

19. The relationship between Portugal’s output fluctuations and those of large European economies has also changed. Real output movements became more tightly associated with those in Europe and more closely synchronized (Table 1, lower section). In 1953-71, Portugal’s output fluctuations lagged Europe’s by two years. In 1977-93, correlation coefficients increased and output fluctuations lagged those in Europe by only one year.9

20. The experience of the last 40 years reviewed above leads to three main conclusions. First, faster growth in Portugal than in the large economies in Europe has been associated with larger cyclical fluctuations. Second, there has been a marked change in cyclical fluctuations in Portugal after 1974, with output fluctuations in the later years being significantly more volatile and persistent than in the earlier years. Third, output fluctuations in Portugal are now more strongly correlated and closely synchronized with the European cycle.

D. Characteristics of the Portuguese Business Cycle

21. To obtain a fuller understanding of the characteristics of the Portuguese business cycle, this section investigates the behavior of aggregate demand, sectoral composition of output, and the labor market during 1953-93. It also presents a qualitative comparison with other studies to assess the extent to which the empirical regularities found in Portugal match those found in other industrial countries.10

22. The cyclical behavior of aggregate demand variables, productive sectors, and the labor market is studied with reference to the output cycle. Figure 2 depicts actual and trend output, and cyclical fluctuations. The troughs of the last three cycles have occurred every nine years: 1975, 1984, and 1993. Furthermore, visual inspection reveals the very similar, and large, amplitude of recent cycles.

Figure 2.Portugal: Output, Trend, and Cyclical Fluctuations

Sources: Bank of Portugal; and staff calculations.

1/ Series expressed in natural logs at 1953 prices.

Aggregate demand and prices

23. The main empirical regularity that emerges from an analysis of the characteristics of de-trended aggregate demand variables is that they are highly persistent and co-move with the reference output cycle (Table 2 and Figure 3). Private consumption is strongly procyclical, coincidental, and as variable as output. These characteristics accord with the evidence presented by Backus and Kehoe (1992) for ten industrial countries (including six in Europe) covering a century of data. Christodoulakis and others (1994) extend the work to include twelve EU countries (including Portugal), using annual data for 1960-90, and find that in two-thirds of the countries the variability of consumption exceeds that of output. In their study on Portugal, Correia and others (1993) find that consumption is less volatile than output, using annual data for 1958-91. When consumption is separated into consumption of durable goods, nondurable goods, and services, the patterns of co-movement with GDP remain the same but those of volatility differ. Consumption of durable goods exhibits the greatest volatility, and services the least. However, contrary to the permanent income and life-cycle hypotheses, consumption of nondurable goods is not smoother than income. Dolado and others (1993) find a similar behavior in Spain, using quarterly data for 1970-91, and favor the explanation that liquidity constraints may have prevented a stronger smoothing behavior in the consumption of nondurable goods. Moreover, frequent tax and transfer policy changes are likely to have affected disposable income.

Table 2.Portugal: Moments of De-Trended Series
StandardRelative

standard
Auto-Cross Correlation with GDP 1/
deviationdeviationcorrelationT-3T-2T-1TT+1T+2T+3
Gross domestic product3.171.000.62
Private consumption3.241.020.53−0.41−0.050.420.860.710.25−0.31
Non-durables4.121.300.50−0.250.070.380.770.600.16−0.35
Durables8.332.630.56−0.230.050.340.640.600.27−0.21
Services3.020.950.39−0.53−0.340.230.550.500.24−0.09
Public consumption3.851.210.44−0.10−0.050.160.280.290.210.00
Investment9.943.140.56−0.230.270.680.760.35−0.11−0.42
Gross fixed investment9.432.980.51−0.40−0.040.440.720.500.08−0.32
Change in inventories1.380.440.140.170.540.510.13−0.25−0.37−0.22
Net exports27.368.640.570.11−0.21−0.40−0.50−0.37−0.150.12
GDP deflator4.451.410.810.09−0.25−0.55−0.65−0.50−0.240.01
Employment1.480.470.63−0.380.020.490.770.640.14−0.48
Productivity2.240.710.50−0.410.040.550.910.490.02−0.36
Primary sector
Output6.141.940.30−0.33−0.170.180.490.300.19−0.25
Employment1.870.590.340.380.20−0.20−0.43−0.48−0.270.00
Productivity7.012.210.37−0.40−0.200.210.540.390.24−0.22
Manufacturing
Output5.621.770.58−0.410.090.570.900.570.03−0.42
Employment2.840.900.62−0.55−0.130.450.760.630.10−0.49
Productivity3.991.260.48−0.180.210.490.730.37−0.03−0.24
Services
Output2.010.640.61−0.36−0.010.460.740.540.06−0.40
Employment2.040.640.74−0.110.150.440.640.550.18−0.29
Productivity1.860.590.62−0.26−0.160.020.11−0.02−0.13−0.11

Correlation coefficients below 0.32 (in absolute terms) are not statistically significant from zero at the 95 percent confidence level in a two-sided test for bi-variate random variables.

Correlation coefficients below 0.32 (in absolute terms) are not statistically significant from zero at the 95 percent confidence level in a two-sided test for bi-variate random variables.

Figure 3.Portugal: Cyclical Fluctuations of Aggregate Demand Components

Sources: Bank of Portugal; and staff calculations.

24. Investment is over three times more volatile than output in Portugal, a result which falls in the middle of the range of two to five that is found in the 10 countries studied by Backus and Kehoe (1992). Investment fluctuations are strongly procyclical, coincidental, and the most persistent among aggregate demand components. The behavior of investment over the business cycle is mostly driven by gross fixed investment, while inventories show smaller volatility and less persistence, and lead output by two periods. The evidence on inventories contrasts with that of Fiorito and Kollintzas (1994), whose study of G7 countries using quarterly data for 1960-89 shows that inventories are by far the most volatile component of aggregate demand. In contrast, Dolado and others (1993) find inventories to be acyclical in their study on Spain.

25. Government consumption is 20 percent more volatile than output, significantly persistent, and acyclical, which conforms with the evidence of other countries where government spending exhibits little regularity (Backus and Kehoe, 1992; Christodoulakis and others, 1994; and Fiorito and Kollintzas, 1994). Although this result may seem surprising since discretionary fiscal policy as a whole is traditionally viewed as countercyclical, it may be explained by the exclusion of other elements of fiscal policy (for example, tax revenues and transfers) which can be expected to move strongly against the cycle, but that are not captured in the definition of public consumption.

26. Net exports of goods and services are over eight times more variable than GDP, and the most volatile component of output. They are significantly persistent, countercyclical, and coincidental. The evidence on net exports in other industrial countries also points to strong volatility (although not the highest among demand components of GDP) and countercyclical behavior. Previous studies on Portugal also report net exports as the most volatile component of GDP.11

27. Prices are very persistent, which may reveal informal indexation mechanisms that are characteristic of relatively high inflation countries, as was the case of Portugal particularly after the 1974 revolution. Prices are also more volatile than output, strongly countercyclical, and coincidental. The countercyclical behavior of price level fluctuations is found in most studies using post-World War II data.12

Employment and production

28. Employment in Portugal shows the most persistent behavior of the real variables. However, it is much less volatile than output, strongly procyclical, and coincidental. Evidence on the volatility and co-movement of employment and output accords with that found in the other studies reviewed. It is also consistent with theories of labor hoarding: firms find it relatively more costly to adjust employment than hours per worker, so that they have an incentive to smooth employment over the business cycle and use labor more intensely in expansionary phases and less intensely in contractionary phases. In contrast to volatility and co-movement, the international evidence on the phase shift of employment (that is, on whether it leads or lags the cycle) does not show as clear a pattern as the components of aggregate demand because institutional environments affecting hiring, firing, job search costs, and unemployment compensation differ across countries. Fluctuations in productivity during the cycle closely resemble the behavior of employment fluctuations; as in G7 countries, productivity is procyclical and coincidental.

29. The data reveal sharp disparities in the cyclical behavior of production and employment by sector (Figure 4). Output in both industry and services is strongly procyclical and coincidental, while the primary sector is less correlated with output, but much more volatile. The same patterns are found in a study by Schlitzer (1995a) on Italy using quarterly data for 1959-92. Persistence also differs across sectors. Output fluctuations in the primary sector are not significantly persistent, in contrast to fluctuations in manufacturing and services. Similarly, employment in manufacturing and services is less volatile than output, significantly persistent, strongly procyclical, and coincidental. As with output, the cyclical characteristics of employment in the primary sector differ from other sectors. It is weakly countercyclical and it lags the cycle by one year.

Figure 4.Portugal: Cyclical Fluctuations of Employment and Productive Sectors

Sources: Bank of Portugal; and staff calculations.

30. In sum, the main characteristics of the business cycle in Portugal conform with the stylized facts of business cycles in industrial countries, as follows:

  • (i) output deviations from trend are persistent;
  • (ii) consumption and investment are strongly procyclical;
  • (iii) net exports are strongly countercyclical;
  • (iv) investment and, particularly, net exports are significantly more volatile than GDP;
  • (v) there is no systematic tendency on government consumption;
  • (vi) price level fluctuations are countercyclical; and
  • (vii) employment is much less volatile than output, procyclical, and persistent.

E. The Stability of the Cycle: A Comparison Before and After 1974

31. As documented above, Portugal experienced an increase in output fluctuations after the 1974 revolution. The question addressed in this section is whether the change in output fluctuations has been associated with significant changes in other characteristics of the Portuguese business cycle.

32. The evidence shows an increase in the volatility of output and of all of its components, except public consumption (Table 3). Output volatility is 66 percent higher after 1974 than before, while the volatility of investment more than doubles and that of net exports posts an almost eightfold increase. To test the significance of such changes, the confidence interval is determined for the variance of output and aggregate demand fluctuations (Table 4).13 In all cases the variance of output and its demand components during the first subperiod falls outside the 95 percent confidence interval for the variance of the second subperiod. The null hypothesis that both variances come from the same distribution is thus rejected, confirming that the increase in volatility of output since 1974 extends to its demand components.

Table 3.Portugal: Moments of De-Trended Series Before and After 1974
StandardRelative

standard
SD(post74/Auto-Cross correlation with GDP 1/
deviationdeviationSD(pre74)correlationT-3T-2T-1TT+1T+2T+3
First Sub-Period: 1953-1971
Gross domestic product1.900.27
Private consumption2.131.120.110.020.110.230.420.380.150.02
Public consumption4.792.520.420.200.190.580.06−0.21−0.040.09
Investment4.942.600.15−0.39−0.590.04−0.09−0.26−0.200.13
Net exports5.382.83−0.010.220.23−0.100.680.200.09−0.28
GDP deflator2.251.180.58−0.22−0.55−0.440.020.000.500.38
Employment0.840.440.60−0.59−0.56−0.470.080.350.320.30
Productivity2.051.080.390.27−0.170.340.830.01−0.17−0.35
Primary sector
Output5.863.080.310.530.370.320.24−0.34−0.25−0.38
Employment1.200.630.820.020.03−0.30−0.29−0.13−0.04−0.10
Manufacturing
Output4.732.480.30−0.38−0.36−0.250.690.460.10−0.09
Employment2.031.070.49−0.61−0.45−0.110.400.500.430.26
Services
Output1.030.540.360.300.330.40−0.08−0.010.200.18
Employment1.790.940.86−0.33−0.40−0.44−0.060.170.110.26
Second Sub-Period: 1977-1993
Gross domestic product3.161.660.69
Private consumption3.781.201.780.69−0.340.330.780.960.740.22−0.43
Public consumption3.090.980.650.62−0.290.170.650.890.900.51−0.15
Investment11.843.752.400.65−0.020.470.780.850.500.04−0.56
Net exports41.9613.297.800.600.07−0.46−0.71−0.80−0.60−0.190.35
GDP deflator5.241.662.330.800.07−0.34−0.75−0.94−0.80−0.48−0.16
Employment1.740.552.080.55−0.390.200.740.880.750.38−0.44
Productivity1.770.560.860.49−0.170.280.650.910.660.24−0.21
Primary sector
Output6.642.101.130.35−0.51−0.590.030.370.230.28−0.15
Employment2.350.751.970.060.750.570.11−0.32−0.46−0.38−0.23
Manufacturing
Output4.901.551.040.68−0.090.480.850.940.590.06−0.56
Employment3.161.001.550.57−0.590.020.670.920.700.23−0.64
Services
Output2.350.742.270.75−0.350.100.500.870.950.680.05
Employment1.880.601.050.64−0.490.070.650.840.830.57−0.10

Correlation coefficients below 0.32 (in absolute terms) are not statistically significant from zero at the 95 percent confidence level in a two-sided test for bi-variate random variables.

Correlation coefficients below 0.32 (in absolute terms) are not statistically significant from zero at the 95 percent confidence level in a two-sided test for bi-variate random variables.

Table 4.Portugal: Stability Tests for Volatility
VarianceConfidence interval for

post-1974 variance 1/
Pre-1974Post-1974MinimumMaximum
Gross domestic product3.629.975.8724.53
Private consumption4.5214.308.4335.19
Public consumption22.929.575.6423.53
Investment24.42140.2982.66345.13
Net exports28.951,760.951,037.654,332.30
GDP deflator5.0627.4816.1967.61
Employment0.703.041.797.47
Productivity4.193.121.847.67
Primary sector output34.3344.1426.01108.59
Primary sector employment1.435.543.2713.64
Manufacturing output22.3624.0314.1659.11
Manufacturing employment4.129.965.8724.49
Services output1.075.513.2513.57
Services employment3.213.552.098.72

The confidence interval for the variance, assuming that the mean of the distribution in unknown, is distributed as a Chi-square with (n-1) degrees of freedom.

The confidence interval for the variance, assuming that the mean of the distribution in unknown, is distributed as a Chi-square with (n-1) degrees of freedom.

33. Price volatility also increased significantly after 1974. This confirms the empirical regularity that higher inflation, as was experienced after the 1974 revolution, is associated with higher volatility (Lucas, 1973; Cukierman, 1979; Hercowitz, 1981; and Blejer, 1982). The average rate of inflation (measured by the GDP deflator) in the second subperiod increased sixfold, to close to 16 percent, from the first subperiod average. As would be expected in conjunction with the rise in output volatility, employment fluctuations also became wider after 1974.

34. The persistence of deviations from trend continues to be high. The hypothesis that persistence is the same across periods cannot be rejected at the 90 percent significance level. Table 5 presents the F-statistics and probability values for the autocorrelation coefficient.14

Table 5.Portugal: Stability Test for Autocorrelation and Cross Correlations
Aurocorrelation 1/Cross Correlation with GDP 2/
F-statisticProbabilityF-statisticProbability
Gross domestic product0.810.46
Private consumption1.350.271.340.28
Public consumption0.170.840.120.89
Investment0.710.502.960.07
Net exports0.260.618.340.01
GDP deflator0.150.860.910.41
Employment0.090.927.810.00
Productivity0.100.917.810.00
Primary sector Output0.020.980.270.77
Primary sector Employment1.980.160.050.95
Manufacturing Output0.680.510.450.64
Manufacturing Employment0.090.911.650.21
Services Output0.990.385.310.01
Services Employment0.860.443.020.06

This is a Chow-test of the stability of the autocorrelation coefficient across periods.

This is a Chow-test of the stability of the contemporanous cross correlation coefficient across periods.

This is a Chow-test of the stability of the autocorrelation coefficient across periods.

This is a Chow-test of the stability of the contemporanous cross correlation coefficient across periods.

35. The co-movement between output and its aggregate demand components has strengthened after 1974 and conforms more closely with the stylized facts of the business cycle in industrial countries. Moreover, many of the pre-1974 results appear atypical (including a countercyclical leading behavior of investment, procyclical behavior of net exports, and low persistence of output fluctuations). Stability tests on the contemporaneous correlation of each variable with GDP yield a significant increase in the co-movement between output and investment, net exports, employment, and productivity (Table 5).15 The correlation between private consumption and GDP also increased after 1974, but the increase is not statistically significant. After 1974, and except for public consumption, all output components are coincidental. Government consumption has continued to be highly procyclical but has changed the phase-shift with respect to the output cycle: it led output fluctuations before 1974 and it has lagged them since then.

36. The evidence presented in this section indicates that Portugal’s increase in output volatility after 1974 has been associated with higher volatility in all aggregate demand components, except public consumption, and continued high persistence. The data also suggest that, after 1974, Portugal’s business cycle resembles more closely the stylized facts of business cycles in industrial countries, with stronger co-movements of output fluctuations with investment, net exports, and employment.

37. Stability tests were also performed for other subperiods, specifically those between 1974 and EU accession in 1986, and from EU accession onward, to test whether the change in policy regime brought about by EU integration has had a significant impact on output volatility. The null hypothesis that output volatility was the same between 1974 and EU accession, and after EU accession, could not be rejected at standard significance levels. It should be noted, however, that splitting the sample to perform stability tests drastically reduces the number of observations and the power of the tests. In any event, the evidence seems to differ from that on Spain, reported by Dolado and others (1993), who find a substantial reduction in output volatility in the 1980s compared with the 1970s, and on Italy, reported by Schlitzer (1995a), who finds no significant change in output volatility after 1973.

F. Another Look at the Increase in Output Volatility

38. This section examines some of the structural and policy-induced changes that might have increased the amplitude of economic cycles in Portugal after 1974. The first hypothesis examined is whether output volatility might have responded to structural changes that increased the relative importance of highly volatile sectors. Figure 5 illustrates the change in the structure of productive sectors and their employment, showing that in the 40 years to 1993 there was a strong recomposition of output away from the most volatile sector of the economy, the primary sector. Its share in GDP fell from 30 percent in 1953 to 8 percent in 1993. Conversely, services—the most stable sector of the economy—grew from almost 40 percent of GDP in 1953 to 46 percent in 1993. Although the services sector remained the most stable in relative terms, its absolute volatility has increased significantly (in a statistical sense) after 1974, and its co-movement with output has also risen significantly. Thus, the increase in output volatility in Portugal cannot be traced simply to a change in the relative importance of the most volatile sector. Instead, the evidence is more subtle and harder to interpret: volatility has increased in all sectors, and especially in the comparatively more stable sector, services.

Figure 5.Portugal: The Changing Structure of the Economy

Sources: Bank of Portugal; and staff calculations.

1/ Series expressed in natural logs at 1953 prices.

2/ Series expressed in natural logs.

39. The second hypothesis investigated is whether higher output volatility was policy-induced. Among the variables more closely linked to policies, net exports exhibited the sharpest change. Volatility increased almost eightfold after 1974. This increase occurred together with a significant rise in the co-movement of trade with output, and a sharp rise in openness to international markets. Exports and imports of goods and services rose from 34 percent of GDP in 1953 to 127 percent of GDP in 1993. Thus, Portugal’s sustained and deep integration in foreign markets since the mid-1970s—boosted by EU entry in 1986-and the associated fluctuations of foreign trade, are likely to have had a notable impact on real output fluctuations.

40. In assessing the causes of the rise in output volatility, due consideration must also be given to the prolonged period of political instability that followed the 1974 revolution, accompanied by successive shifts in economic policy approaches and a number of external shocks, both of a worldwide (oil prices) and country-specific nature (the loss of privileged markets in Africa, the massive return of settlers, etc.) These developments were reflected in balance of payments crises that led to two stand-by arrangements with the Fund (in 1978 and 1983); double-digit inflation in all but the last two years of the period under review; and a marked fiscal deterioration (from an average surplus of 1¼ percent of GDP in the pre-revolution period to a deficit of 6½ percent of GDP thereafter). Earlier work by the staff on the construction of a financial conditions index for Portugal showed that there were frequent shifts in the stance of monetary and fiscal policies during most of the 1980s, and that financial policies tended to stabilize only in the early 1990s.16 It would not be surprising that the higher volatility of financial policies that typified most of the post-1974 period played a role in inducing a rise in output deviations from trend and in price volatility.

41. In sum, it would appear that the sharp rise in the deviations of output around its long-term trend that has characterized Portugal’s business cycle after 1974 has been partly the cumulative result of policy induced-factors, notably the far-reaching foreign trade integration and lesser stability of financial policies (compared with the pre-1974 period), and partly the result of structural factors. It should be noted, however, that the evidence presented above is not one of causality but of association, and does not distinguish the relative importance of the factors under review; it therefore needs to be interpreted with caution.

G. Concluding Remarks

42. The overall evidence presented in this chapter shows that economic fluctuations in Portugal conform with the general characteristics of the business cycles in European and other industrial countries. The similarities between Portugal and these countries, which have become even more marked since 1974, extend to the persistence of output fluctuations, and the co-movement and volatility of a wide array of variables relative to the output cycle. Furthermore, output movements in Portugal are now more closely synchronized with those in larger European partners. However, a principal characteristic of Portugal’s business cycles, that is, the volatility of output deviations from trend, has distanced itself further from developments in European partners, with such volatility increasing markedly since 1974.

43. Predictions of the nature and amplitude of future economic cycles in Portugal are subject to contrasting considerations. On the one hand, Portugal has recently made great strides toward nominal convergence, and price stability has been virtually achieved. Thus, lower price volatility may exert a dampening influence on output fluctuations. On the other hand, as Portugal joins EMU, it will experience further trade and financial market integration and thus be exposed to a set of external shocks that may magnify output fluctuations. Hence, when assessing the likely future nature of Portugal’s business cycles to determine the appropriate stance of policies in the medium term, it would be prudent to assume that (i) Portugal will continue to experience stronger output fluctuations than the large industrial countries in Europe, (ii) the cycles between Portugal and large EU partners will be more closely synchronized, and (iii) the co-movements and relations between output and aggregate demand variables in Portugal will increasingly conform to the stylized facts of the business cycle of other advanced economies, as recalled in this chapter.

44. The main policy implication of continued sharper output fluctuations in comparison to the large EU countries, whose situation will presumably determine the overall stance of monetary policy within EMU, is that Portugal’s fiscal policy position will need to be comparatively strong. The underlying fiscal balance in the medium term will, in other words, need to be sufficiently robust to allow the automatic stabilizers to work in the face of economic downturns, while still meeting the requirements of the Stability and Growth Pact.

References

    BackusD.K. and P.J.Kehoe1992“International Evidence on the Historical Properties of Business Cycles” American Economic ReviewNo. 82.

    • Search Google Scholar
    • Export Citation

    BlejerM. and L.Leiderman1982“Inflation and Relative Price Variability in the Open Economy” European Economic Review.

    ChristodoulakisN.S.P.Dimelis and T.Kollintzas1994“Comparisons of Business Cycles in the EC: Idiosyncracies and Regularities”Economica.

    • Search Google Scholar
    • Export Citation

    CorreiaI.H.J.L.Neves and S.T.Rebelo1992“Business Cycle in Portugal: Theory and Evidence” in The Portuguese Economy Toward 1992ed. byJ.AmaralD.Lucena and A.Mello (Norwel, MA: Kluwer).

    • Search Google Scholar
    • Export Citation

    CorreiaI.H.J.L.Neves and S.T.Rebelo1992“Business Cycles from 1850 to 1950: New Facts about Old Data”European Economic Review No. 36.

    • Search Google Scholar
    • Export Citation

    CorreiaI.H.J.L.Neves and S.T.Rebelo1993“Business Cycles in a Small Open Economy,”Working Paper No. 16/93 Bank of Portugal.

    CukiermanA.1979“The Relationship between Relative Prices and the General Price Level: A Suggested Interpretation”American Economic ReviewJune.

    • Search Google Scholar
    • Export Citation

    DoladoJ.J.M.Sebastian and J.Vallés1993“Cyclical Pattern of the Spanish Economy”Bank of Spain.

    FioritoR. and T.Kollintzas1994“Stylized Facts of Business Cycles in the G7 from a Real Business Cycle Perspective”European Economic ReviewNo. 38.

    • Search Google Scholar
    • Export Citation

    HercowitzZ.1981“Money and the Dispersion of Relative Prices”Journal of Political Economy.

    HodrickR.J. and E.C.Prescott1980“Post-war U.S. Business Cycles: An Empirical Investigation” (Carnegie-Mellon UniversityPittsburgh, PA).

    • Search Google Scholar
    • Export Citation

    KingR.G. and S.T.Rebelo1993“Low Frequency Filtering and Real Business Cycles” Journal of Economic Dynamics and ControlNo. 17.

    • Search Google Scholar
    • Export Citation

    KydlandF.E. and E.C.Prescott1990“Business Cycles: Real Facts and a Monetary Myth”Federal Reserve Bank of Minneapolis Quarterly Review No. 2.

    • Search Google Scholar
    • Export Citation

    KydlandF.E. and E.C.Prescott1977“Understanding Business Cycles”in Stabilization of the Domestic and International Economyed. byK.Brunner and A.H.Meitzer (North-Holland, Amsterdam).

    • Search Google Scholar
    • Export Citation

    SchlitzerG.1995a“Business Cycles in Italy: A Statistical Investigation,”European Journal of Political EconomyVol. 11.

    SchlitzerG.1995b“Have Economic Fluctuations Been Dampened? New Empirical Evidence from Italy,”Open Economies ReviewNo. 6.

    SheffrinS.M.1988“Have Economic Fluctuations Been Dampened? A Look at Evidence Outside the United States,”Journal of Monetary EconomicsNo. 21.

    • Search Google Scholar
    • Export Citation
1

The availability of this series, it should be noted, has been a major contribution to the advancement of research on the Portuguese economy, and was used extensively throughout the present paper.

2

Prepared by Maria Carkovic.

3

”PortugaPs Performance After Accession to the European Union: The Growth Payoff of Policy Reform,” (SM/96/253, 10/4/96).

4

According to Kydland and Prescott (1990), the HP filter satisfies the following criteria:”The trend component of real output should be approximately the curve that students of business cycles and growth should draw through a time plot of the time series. The trend of a given time series should be a linear transformation of that time series, and that transformation should be the same for all series. Lengthening the sample period should not significantly alter the value of the deviations at a given date, except possibly near the end of the original sample. The scheme should be well defined, judgement free, and cheaply reproducible.”

5

Specifically, following Fiorito and Kollintzas (1994), a variable is strongly cyclical if the absolute value of the maximum cross correlation is between 0.5 and 1. If the value is between 0.32 and 0.5, the variable is weakly cyclical, and if it is below 0.32 the variable is acyclical. The cutoff point of 0.32 corresponds to the value required to reject the null hypothesis that the correlation coefficient is zero at the 5 percent significance level in a two-sided test for bi-variate random variables.

6

The results presented in this paper also hold if Portugal is compared with Germany and France alone.

7

For Portugal, Table 1 reports a reversal in the sign of the autocorrelation coefficient in the two subperiods. Backus and Kehoe (1992) report a similar change in their examination of pre-and post-war growth rates in 10 industrial countries. They point out that this reversal may reflect a change in the nature of economic fluctuations, but measurement error is also a possibility, since white-noise errors in the output series would have precisely this effect on the autocorrelation of growth rates.

8

Each observation is computed using a rolling window of nine years. Thus, the standard deviation dated t is computed using the cyclical component of real GDP from t−4 to t+4.

9

The gradual synchronization of output cycles seems to have continued beyond 1993. It is widely acknowledged that the last European recession, which bottomed in 1993, and the subsequent recovery, have been felt contemporaneously in Portugal.

10

A quantitative comparison of the statistical moments of the de-trended series of Portugal and those reported for other countries is not possible due to differences in time periods, periodicity, and the choice of de-trending parameters to analyze business cycles.

13

The confidence interval for the variance, which assumes that the mean of the distribution is unknown, is distributed as a Chi-square with (n-1) degrees of freedom.

14

The statistics are derived from performing Chow-tests on the parameters of regressions of each detrended variable on its lagged value.

15

The statistics are derived from performing Chow tests on the parameters of regressions of each detrended variable on detrended output.

16

“An Indicator of Monetary and Financial Conditions in Portugal,” (SM/96/253, 10/4/96).

Other Resources Citing This Publication