This paper presents four studies on selected issues of the Chilean economy. The new framework for fiscal policy represented by the authorities' target of a central government structural surplus of 1 percent of GDP is also discussed. This study examines whether the variance of the cyclical component of output in Chile and 11 other countries increased during the 1990s, a period during which the variance of inflation declined. The alternative measures of potential output for Chile are also estimated.


This paper presents four studies on selected issues of the Chilean economy. The new framework for fiscal policy represented by the authorities' target of a central government structural surplus of 1 percent of GDP is also discussed. This study examines whether the variance of the cyclical component of output in Chile and 11 other countries increased during the 1990s, a period during which the variance of inflation declined. The alternative measures of potential output for Chile are also estimated.

III. Unemployment Persistence in Chile31

A. Introduction

Motivation and stylized facts

79. In 1998–99, the Chilean economy suffered its first recession after an extraordinary period of high and sustained economic growth lasting 15 years. The combination of external shocks, originating from the Asian crisis in 1997, the Russian crisis in October 1998 and the Brazilian devaluation in January 1999, led to adverse terms of trade and diminished supply of external finance. These shocks, together with tight monetary conditions, prompted a sharp fall in domestic demand and economic activity. From the fourth quarter of 1998 to the third quarter of 1999, Chile’s annual GDP growth rate decreased by 2.7 percent while domestic demand fell by 14 percent. Yet, the cyclical downturn proved short lived as economic activity recovered at a strong pace by the fourth quarter of 1999.

80. Following the economic downturn, labor market conditions deteriorated significantly. Total unemployment grew by more than 50 percent on a annual basis throughout 1999 while total employment experienced a prolonged downturn trend. As a result, unemployment shot up to an annual average rate of 9.7 percent in 1999, compared to an annual average rate of 6.9 percent in the decade prior to the economic downturn.32 In 2000, despite the initial strength of the recovery in economic activity, unemployment remained high averaging 9.2 percent.

81. Recent discussion on the response of the labor market to the economic crisis has pointed out to the weak job creation capacity of the economy. A question arises whether the 1998–99 economic shock would have only a temporary effect on labor market dynamics or whether it could have a highly persistent impact on job creation.

82. Large, cyclical shocks could lead to a significant transformation of employment by sector, occupation and skill. Rapid output adjustments imply increases in temporary unemployment as contracting sectors shed workers before expanding sector rehired them. This cyclical unemployment reflects the time required for job search in the context of higher job destruction. However, if skills required for the new jobs in expanding sectors are substantially different from those in the contracting sectors, then a deterioration in employment could become a more serious problem of persistent mismatch. Unemployed workers moving to expanding sectors would have to acquire the necessary skills involving a costlier training process. Correspondingly, employment dynamics resulting from an economic downturn accompanied by significant shifts across sectors and occupations would reflect a more protracted response to economic recovery.

Objective of the chapter

83. The focus of this chapter is to understand the extent to which aggregate shocks leading to significant labor reallocation across economic sectors help explain unemployment persistence despite the resumption in aggregate output growth. Following the business cycle literature, we view aggregate shocks as the primary factors determining unemployment fluctuations. More significant, economic disturbances inducing labor to be reallocated across sectors affects labor market flows due to search and matching costs. An increase in job reallocation across sectors could lead to a sharp decline in job creation, an increase in unemployment duration, and a possible reduction in labor force participation. Consequently, aggregate shocks generating a high output volatility across sectors could help explain continued unemployment persistence long after aggregate activity recovers.

84. Our main argument does not imply that sector reallocations represent the driving factors explaining output volatility at high frequencies. Rather, we claim that an uneven labor demand adjustment across sectors can exacerbate the consequences of “productivity shocks” as discussed by the real business cycle literature. We expect that asymmetric output adjustments across sectors have important consequences to the dynamics of key aggregate variables.

85. Output fluctuations and labor reallocation across sectors could help explain the persistence in unemployment and weakness in job creation in Chile. Much of the decline in employment can be explained by the sharp downturn in two sectors in the economy: construction and manufacturing. Employment in construction and industry fell by more than 20 percent and 10 percent in 1999 respectively. Since both sectors account for 30 percent of nonagricultural total employment, changes in employment flows in these two sectors are largely responsible for the increase in aggregate unemployment and might explain its persistence.

86. The model we present in this chapter helps elucidate the relationship between changes in aggregate employment and asymmetric output response to shocks across sectors. The model integrates a labor market matching model in a two sector economy. The labor market matching approach follows the basic framework developed by Mortensen and Pissarides (1994), which explicitly formulates the searching process that firms and workers undertake and the bargaining process between parties to share the rent from job matching. By introducing a labor matching model in the context of a two-sector economy, we show through a numerical simulation the effects that asymmetric productivity shocks across sectors have on the dynamics of aggregate labor market flows and economic activity.

87. More important, we present empirical evidence relating employment fluctuations across sectors on aggregate unemployment using Chilean data from 1986 to 2000. Rather than attempting to estimate a structural econometric specification, our analysis follows a more parsimonious approach as presented by Lilien (1982) and further refined by Abraham and Katz (1986) and Mills, Peroni and Zevoyianni (1995). We use a baseline unemployment equation and proxies of employment and output dispersion across sectors to test whether the asymmetric response across sectors to unexpected aggregate shocks is significantly related to changes in aggregate unemployment and can explain its persistence.

88. As labor market regulations play a fundamental role in determining labor market outcomes, we summarize some of the main elements of labor market legislation in Chile and discuss how institutional rigidities could affect the response of labor markets to shocks. This discussion aims at providing a background to understanding how labor market regulations affect employment dynamics and to refer to some of the empirical literature that shows the impact of regulations on unemployment in the case of Chile.

Organization of the chapter

89. The chapter is organized as follows. Section B provides a brief description of labor market trends over the past two decades, characterizing its dynamics over the business cycle. It also presents the salient facts showing the increased employment fluctuation across sectors following the 1998–99 recession. In section C, we present the basic theoretical framework and discuss the results from the numerical simulation exercise. Section D shows the results of the empirical analysis testing the relevance of employment fluctuations across sectors after controlling for the impact of aggregate shocks. In section E, we summarize the main elements of labor market legislation shedding light on the impact of institutional rigidities on unemployment persistence. We conclude by summarizing the main findings from analysis and discuss basic policy implications.

B. Main Characteristics of Chilean Labor Market (1980–2000)

90. In this section, we describe the main characteristics of the labor market in Chile. We present a broad view of its basic trends over the past two decades and compare them to other countries’ experiences. Outlining the basic characteristics of labor market dynamics over the business cycle, we note the procyclical nature and persistence of employment measures. Finally, we present evidence on the asymmetric effect of the 1998–99 shock across sectors pointing out how employment variation can largely be explained by changes in construction and industry.

Labor market facts: trends and cross-country comparison

91. Table 1 sets out some information on general labor market characteristics across a sample of economies. We focus on the past two decades to point out some basic differences among countries. Not surprisingly, there is much variation across economies without any one definable pattern. Cross-country comparison using labor market figures are notoriously unreliable given the accounting differences across countries. Nonetheless, some indication of how some measures of labor market for Chile fare against other countries serves as a prelude to explaining the behavior of unemployment in Chile.

Table 1.

Chile: Comparing Labor Statistics Across Countries

(In percent)

article image
Source: Fund staff estimates.

92. Chile’s averaged unemployment in the past two decades was the highest among the countries shown in the table, except for Spain. In the 1990s, however, unemployment in Chile fell considerably below the high level of some European countries, but remained above the level of the United States and that of fast-growing economies such as Korea, Taiwan and Singapore. In general European countries are not noted for the flexibility of their labor markets. Nonetheless, labor market rigidities are not the only reasons for high unemployment as observed in the case of the United Kingdom with its flexible labor markets.

93. Total employment growth rate in Chile during the 1980–2000 period was above those of most countries in Table 1. These differences do not take into account differences in rates of population growth. Nonetheless, employment growth in Chile slowed down significantly in the 1990s.

94. Labor force participation figures show some important differences across countries. While developed economies show high levels of labor force participation (except for Spain), fast-growing emerging economies tend to have relatively low levels. In the case of Chile, labor force participation has shown a sharp increase until the mid 90s largely associated to the incorporation of female workers.

95. Real wage growth in Chile outstripped that of developed economies. Differences in real wage growth generally reflect labor productivity gains. As expected, fast-growing emerging economies tend to outperform developed countries given that labor productivity changes are higher in developing countries adopting new technologies and improving their human capital. Labor productivity in Chile has seen large gains in the past decade before edging down as a result of the crisis in 1998–99.

Labor market dynamics over the business cycle

96. We now outline key stylized facts about the cyclical behavior of the Chilean labor market. We follow the standard approach of quoting standard deviations and correlations of the changes in the cyclical components of employment, unemployment, real wages, and labor productivity with respect to output.33 The main thrust of these statistics is to show the relative volatility of these variables to aggregate activity fluctuations over the business cycle and their relative co-movements.34

97. From table 2, we observe that employment and unemployment tend to lag output. While employment is procyclical, unemployment is countercyclical. Labor productivity is highly correlated to output and is contemporaneous to output. In contrast, real wages show a very low correlation with output, employment or unemployment. Moreover, the cyclical component of employment is much less volatile than output while unemployment shows a much higher degree of volatility. These basic observations of Chilean data behavior are similar to those of the United States and the United Kingdom35 In Chile, labor market fluctuations are fundamentally connected to the business cycle.

Table 2.

Chile: Business Cycle Facts of Chile’s Labor Market

article image
Note: All variables correspond to first difference of the cyclical component of the seasonally adjusted variable using the Hodrick-Prescott filter to detrend the data. The first column shows standard deviations divided by the standard deviation of output.

98. As a corollary, a generalized observation in the business-cycle literature is that total hours worked increase and decrease at the same time as output even though employment fluctuations are much lower. This implies that employers prefer to adjust average hours worked (intensive margin) rather than shedding workers (extensive margin). The existence of hiring and firing costs explain why employers are usually hesitant to layoff workers when they perceive output changes as only temporary.

99. The presence of hiring and firing costs could also provide some insight on why measures of employment dynamics show strong persistence over the business cycle. To measure this persistence in the Chilean labor market, we use the unit root test of Cochrane (1988). This test is based on the simple observation that if a variable is a random walk, then the further ahead it goes in time, the more uncertainty there is about the level of the variable. The two-period variance of a random walk should be twice the one-period variance. Cochrane (1988)’s unit root test examines how the variance of the k-period difference changes relative to k times the variance of one-period difference.36 If this measure converges below unity, then the variable considered behaves in a random walk manner, meaning that shocks only have a temporary impact on the variables behavior. If the test measures goes above one, shocks leave a more persistent effect on the variables behavior suggesting that the effect of the shock is amplified over time.

100. Figure 1 shows the persistence of unemployment, employment and real wages for the case of Chile. While employment and real wages persistence measures edge down below unity, the unemployment measure goes above unity and continues to increase showing that there is high persistence in unemployment.

Figure 1:
Figure 1:

Measures of Persistence

Citation: IMF Staff Country Reports 2001, 120; 10.5089/9781451807585.002.A003

Source: Original data from Banco Central de Chile

Employment fluctuations across sectors in the late 1990s

101. Figure 2 illustrates the relation of the high unemployment persistence to the aggregate economic downturn in 1998–99. It also shows that during the same period the variance of employment across sectors increased significantly. The downturn in employment was largely led by construction and industry. The third panel in figure shows that net employment changes in these two sectors alone accounted for more than 100 percent of net nonagricultural employment. While employment in these sectors was contracting sharply, other sectors, particularly in service related sectors, expanded. This evidence suggests that important job reallocation across sectors occurred in the period following the recession.

Figure 2.
Figure 2.

Chile: Output Growth and Unemployment

Citation: IMF Staff Country Reports 2001, 120; 10.5089/9781451807585.002.A003

102. Table 3 shows that output recovery in 2000 was largely concentrated on more capital-intensive sectors. Electricity, transportation, financial services and mining led the recovery by increasing productivity as seen from the poor net employment creation in these sectors. These sectors accounting for about 35 percent of total output are significantly the most productive sectors in the economy; they employ about 17 percent of total employed workers. The fact that the aggregate shocks affected more on labor-intensive sectors and that the subsequent recovery was led by the capital-intensive sectors sustains the protracted employment response at the aggregate level. In the following section, we develop a simple numerical exercise that simulates the asymmetric productivity shock across sectors leading to slow adjustment in the unemployment rate.

Table 3.

Chile: Employment and Output by Sector 1/

article image
Source: Central Bank of Chile.

Annual average.

C. Theoretical Model: Specification and Simulation

103. The model presented in this chapter provides a simple formal analysis to understand how cyclical, temporary shocks inducing significant job reallocation across sectors can have a more lasting impact on unemployment. We consider a labor market matching model in the context of a two-sector economy. We borrow the labor matching framework from Mortensen and Pissarides (1994) and so introduce a friction on labor flows in terms of an implicit search cost. In this respect, the chapter follows the literature of job search and the business cycle, first formulated in a general equilibrium context by Andolfatto (1995) and Merz (1996). We proceed in a similar fashion to Rogerson (1987) and Phelan and Trejos (2000) that consider a multi-sector economy in the context of labor-market search and business cycle framework.

104. The objective of this section is to illustrate how the labor market response to an aggregate economic downturn varies depending on the nature of the cyclical shock. In the context of the model, we compare two alternative scenarios where different productivity shocks to the two sectors in the economy could have a significantly different effect on the labor market adjustment process, while having a similar impact on the output downfall. First, we show the impact of a temporary negative productivity shock on the labor-intensive sector of the economy coupled with a positive shock to the less labor-intensive sector of the economy. We compare this case to one where the temporary shock affects evenly the two sectors of the economy. We show that the first scenario leads to a much higher unemployment persistence even though the initial impact on aggregate activity is the same in both cases. These simulation experiments seek to provide a rationale for why unemployment persistence could be higher in the context of cyclical aggregate downturn.

105. Notwithstanding the results from the model, the simulation of the model does not follow a rigorous calibration of the Chilean economy given the simplicity of the model. The results are only meant to illustrate a possible explanation for the case of high unemployment persistence in Chile. Appendix 1 has a detailed explanation of the model and the assumptions made for the simulations.

106. The model is founded on two constructs, a matching function that characterizes the search and recruiting process by which new job-worker matches are created and an idiosyncratic termination shock that captures job exit process. The job-worker matching process is similar to a production process in which employment is produced as a intermediary production input. In this context, the output corresponds to the flow of new matches that is produced with search and recruiting efforts supplied by workers and employers respectively. The market matching function is thus a simple construct describing this search and recruiting process.

107. By inserting the labor market matching process into a simple dynamic model, we can solve the dynamic planning problem through a social planner formulation rather than a competitive equilibrium. Because the technological frictions in the job market search do not generate externalities on the labor market, the solution to the social planner problem represents also a solution to the decentralized competitive equilibrium problem. The job matching frictions translate into search costs and do not represent market failures. Resorting to a simpler dynamic planning problem allows us to compute a solution to the resource allocation problem and so simulate the impact of uneven productivity shocks across sectors on the dynamics of aggregate labor market flows and aggregate output.

Simulation results

108. Using the calibration assumptions presented in the appendix section, we study the dynamic response of labor markets and output to temporary shocks to productivity across sectors. We analyze the following two scenarios. We first consider a decrease in productivity in the two sectors of the economy of about 7.5 percent. By assumption, sector 1 is the low labor-demand sector while sector 2 represents the high-labor demand sector. In the second scenario, we suppose a productivity drop of 20 percent to the high labor-demand sector (i.e sector 2) and a 10 percent increase in the low labor-demand sector (i.e. sector 1). In both cases, the magnitude of the initial aggregate output downturn is the same and equivalent to approximately 9 percent.

109. Figure 3 presents graphs illustrating the path of convergence to the steady-state equilibrium of total unemployment, output, total hours supplied and hours spent doing recruitment for the two cases considered. We find that unemployment shows significant persistence relative to the output adjustment for both scenarios. More important, unemployment persistence is significantly higher in the second scenario while the aggregate output adjustment processes are roughly the same in the two cases. Total hours supplied and recruitment efforts also decrease by a larger margin in case 2 than in case 1.

Figure 3a:
Figure 3a:

Unemployment Deviation from SS

Citation: IMF Staff Country Reports 2001, 120; 10.5089/9781451807585.002.A003

Figure 3b:
Figure 3b:

Output Growth Deviation from SS

Citation: IMF Staff Country Reports 2001, 120; 10.5089/9781451807585.002.A003

Figure 3c:
Figure 3c:

Recruitment Effort Deviation from SS

Citation: IMF Staff Country Reports 2001, 120; 10.5089/9781451807585.002.A003

Figure 3d:
Figure 3d:

Hours Worked Deviation from SS

Citation: IMF Staff Country Reports 2001, 120; 10.5089/9781451807585.002.A003

110. The basic intuition behind these results is related to the trade-off between the costs and benefits from recruiting efforts. The gradual adjustment in employment in the two scenarios considered is due to the convexity of new hires to the recruiting efforts. Given that hiring workers in the next period imply the sacrifice in hours per worker that has to be put into the matching process, recruitment efforts entail an explicit cost of forgone production. In the margin, firms have to equate the increased gain from an additional worker next period to the cost of forgone production due to the increased recruitment effort. Given the convexity of the matching function, adding workers following the output adjustment is costlier than the cost of the foregone production so recruitment efforts are less intensive than one would expect. The low intensity of recruitment efforts explains the gradual adjustment in employment.

111. Another important feature of the model is that it allows for a complete divisibility of labor. This suggests that the effect of output downfall is concentrated on the intensive margin (i.e. hours worked) rather than the extensive margin. This explains why changes in total hours are much more pronounced than changes in employment. These two assumptions (i.e. convexity and divisibility) explain the large shift in total hours and the slow adjustment in employment in both the two scenarios considered.

112. Our analysis also shows that while the adjustment process in aggregate output is roughly similar in the two scenarios, unemployment persistence in the second one is higher than in the first one. The main difference between the two cases considered is that the relative costs between sectors in the two experiments are different. In the second scenario, the growing sector (i.e. sector 1) does not increase enough hours worked and the number of hires relative to the shrinking sector (i.e. sector 2). This implies the fall in aggregate output and employment. For the growing sector, the gains from hiring new workers increase, but so do the costs of foregone output in terms of recruitment effort. Because internal costs in the sector 1 increase, the adjustment in employment is slower, and so the aggregate adjustment in unemployment is slower. Correspondingly, the differences in the cost-benefit trade-offs in the two scenarios explain the differences in unemployment persistence.

113. The difference in these examples illustrate how shocks leading to a similar output recovery paths entail different adjustment rates in employment. In particular, uneven shocks across economic sectors result in higher unemployment persistence. In the next section, we provide some empirical evidence relating the uneven response of employment adjustment across sectors to unemployment persistence as illustrated in these simple experiments.

D. Empirical Analysis

114. In this section, we test the basic conjecture of the chapter using Chilean data. We want to know whether aggregate shocks generating significant labor reallocation across economic sectors contribute significantly to unemployment persistence despite the resumption in aggregate output growth. Economic disturbances requiring labor to be allocated across sectors could induce a longer adjustment period in labor markets due to search and matching costs. This section provides the empirical ground to the qualitative findings of the model described in the previous section. Rather than attempting to estimate a structural econometric specification, we follow a more parsimonious approach as presented by Lilien (1982) and further refined by Abraham and Katz (1996) and Mills, Peroni and Zevoyianni (1995).

115. Our approach uses a baseline unemployment equation and tests whether shocks leading to significant labor reallocation across sectors have an impact on changes in aggregate unemployment. The specification of the baseline unemployment equation follows Barro (1997) by modeling changes in the unemployment rate as a function of unanticipated aggregate shocks. To measure labor reallocation across sectors, we adopt Lilien (1982)’s procedure to construct an employment dispersion index. Responding to Abraham and Katz’s criticism of Lilien’s measure, we purge the employment dispersion index of aggregate shocks. This allows us to test whether changes in unemployment are significantly related to net employment across sectors after controlling for the effect of aggregate shocks.

116. The main component of the empirical analysis is to identify a measure of unanticipated shocks. We try two alternatives. First, we consider purely exogenous variables as proxies to unanticipated shocks. In the case of Chile, we use changes in copper prices, changes in the real exchange rate, changes in terms of trade and changes in foreign financing conditions as potential candidates. This approach is similar to an instrumental variable estimation insofar as changes in these variables are expected to be uncorrelated to anticipated aggregate demand changes.

117. In the second alternative, we construct a measure of unanticipated aggregate shocks by using the residuals from an estimated money demand equation. The estimated residuals from a money growth model reflect are meant to capture a variety of possible aggregate shocks unanticipated by economic agents, and so do not reflect only the effect of monetary policy actions. Insofar as monetary aggregates are endogenous processes in the context of an inflation targeting monetary framework such as in Chile, estimating unanticipated money demand changes provides a robust proxy for unanticipated aggregate shocks. To estimate a money growth equation, we use a single-equation error correction model following the methodological lines espoused by Hendry (1993).

Identifying unanticipated aggregate shocks

118. We begin our empirical analysis by estimating measures of unexpected and expected money growth. We use these measures to construct the baseline unemployment model. To develop a forecasting equation for money growth, we search for an equation specification. Our specification considers a single-equation error correction model relating money growth Δmt, output growth Δyt, changes in inflation Δit, interest rates Δrt, and unemployment Δut. The model stipulates a cointegrating relationship among these variables. The rejection of the Augmented-Dickey-Fuller tests for the each of the first differences of the selected variables shows that they are integrated of order 1. More important, Johansen (1988)’s test for existence of a cointegrating relationship shows that we can reject, at a 1 percent confident level, the null of no-cointegrating vectors in favour of the alternative of having at least one.37 We include a time trend to the cointegrating vector in order to capture the effects of financial innovation. Various studies on the gains from financial liberalization in emerging economies show that financial innovation occurs over time in a gradual manner through a learning-by-doing process.38

119. The proposed money growth equation model has the following form:


We first derive the error correction estimator ECt-1 as a residual from a cointegration regression of levels of money demand relating money demand to income, inflation, unemployment and interest rates. The cointegrating regression also includes the leads and lags of the first differences in output, inflation, interest rate and unemployment as suggested by Phillips and Loretan (1991) in order to achieve an efficient estimate of the cointegrating vector.

120. Table 4 presents the results from the money growth regressions using two different specifications. The results are robust to a variety of misspecification and parameter constancy tests.39 From the money growth regression, we use the residuals as a measure of unanticipated money growth resmt.

Table 4.

Chile: Money Growth Regression

article image
Source: Central Bank of Chile.Note:

significant at 1 percent confidence interval

significant at 5 percent confidence interval

significant at 10 percent confidence interval

Employment dispersion

121. We construct an employment dispersion index across sectors as a measure of the magnitude of the disturbances in labor reallocations across sectors. We use Lilien (1982)’s measure of a weighted sample variance of cross-sectoral employment growth rates:


where eit represents the employment in sector i, and Et stands for total employment. Responding to Abraham and Katz’s criticism of Lilien’s measure, we purge the employment dispersion index from aggregate shocks. We regress the relative change employment for each sector on a measure of anticipated aggregate shocks derived from the money growth regressions. We then use the residuals from these equations to construct the purged dispersion as a weighted sample variance of employment residuals for each sector. In our estimations below we use both the original and the purged dispersion indices.

Baseline unemployment equation

122. The two general specifications of the unemployment equation are:




where resmt stands for the residual from the money growth equation and qt represents the instrumental variables to capture the effect of the unanticipated aggregate shocks, which in our specification correspond to changes in the copper price, the real exchange, terms of trade and an index of Chilean corporate bond spreads.

123. Tables 5 and 6 summarize the main results from these regression estimations using the various proxies for unanticipated aggregate shocks and the two measures of dispersion index. The main observation from these regression results is that the dispersion index, using both the purged and nonpurged measures, remains consistently significant for all specifications onsidered.

Table 5.

Chile: Unemployment Equations

article image
Source: Central Bank of Chile.Tab

significant at 1 percent confidence interval

significant at 5 percent confidence interval

significant at 10 percent confidence interval

Table 6.

Chile: Unemployment Equations

article image

significant at 1% percent confidence interval

significant at 5% percent confidence interval

significant at 10% percent confidence interval

124. This is a robust indication that the dispersion of net employment across sectors does have an impact on unemployment changes. In particular, the first lag of the dispersion index, irrespective of being purged, is positively and significantly correlated to changes in unemployment. However the third lag of the index is negatively correlated and significant and the magnitude of the coefficient is similar to the previous one. This suggests that even though the dispersion index is related to changes in unemployment the persistence is relatively transient. A simple impulse response function analysis shows that a temporary shock to the dispersion index does not have a long-lived effect on unemployment changes. Correspondingly, while being significantly correlated to unemployment, the measure of employment dispersion across sectors explains only partially the persistence in unemployment.

125. We also note that the exogenous variables used as instruments to proxy unanticipated aggregate shocks (changes in copper prices, changes in real exchange rate and changes in corporate bond spreads) cannot explain changes in unemployment. In contrast, the estimated unanticipated aggregate shock using the money growth residual proves a more significant proxy for unexpected aggregate shocks insofar as these lead to changes in unemployment. Also the sign of the unanticipated aggregate shock is negative suggesting that an unexpected negative shock leads to a higher unemployment in the next period.

126. Nonetheless, it is worth noting that the parsimonious approach to model specification suggests the potential for misspecification error. Such criticism remains a valid if the dispersion index (though purged from the aggregate shocks captured by the money growth model) were correlated to an omitted variable, also correlated to the dependent variable. In such case, the result would be biased and not significant. We have tried to reduce this type of error by controlling for the various variables presented in the regressions and the robustness of the results indicate that the margin error for misspecification might be small.

127. To conclude, this empirical exercise provides evidence, at least partially, to the fact that economic disturbances leading to significant labor reallocation across sectors help explain the increase in unemployment after controlling for business cycle effects. More specifically, in the case of Chile, shocks leading to higher unemployment in the late 1990s were largely concentrated on labor intensive sectors of the economy such as construction, commerce and industry. However, the evidence found also suggests that the measure of employment dispersion does not have a long-lived effect on unemployment changes and explain only partially the persistence in unemployment. We note that that the evidence presented is limited insofar as we do not consider microeconomic factors such as institutional rigidities in labor markets that could give much hindsight towards understanding unemployment persistence in Chile.

E. Labor Market Legislation

128. Labor market regulations play a fundamental role in determining labor market outcomes by altering the labor matching process, employment costs and incentives to work. Seeking to improve workers’ welfare, these institutions aim at increasing their income security through employment protection legislation, severance payments, minimum wage, and social security benefits in case of sickness, work accidents and old age. They also benefit workers by strengthening their bargaining power through collective bargaining and union support. However, these regulations come at a costs as they introduce some implicit and explicit costs restraining labor market flexibility. Because labor market rigidities limit the labor matching process, they contribute to the slow response of labor markets to business cycle fluctuations.

129. In this section we review of some of the basic characteristics of the two set of labor market reforms in Chile over the past two decades.40 The first reform was implemented in the early 1980s by the military regime that started in 1973, while the second was carried out during the early 1990s by the first democratic government. Given the highly protected and rigid labor market legislation prevailing in the 1960s, which was significantly relaxed during the first years of the military regime, the reforms in the 1980s and 1990s sought to modernize labor relations, reduce labor market distortions and increase labor market flexibility. The reforms covered three basic areas: (i) employment protection, (ii) collective bargaining and wage setting, and (iii) social security.

Employment protection

130. The Labor Plan of 1980 made significant amendments to job security legislation. It established severance payments as part of the overall job contract guaranteeing a minimum severance of a month wages per year of tenure with a five-month ceiling. As an extension to this reform, legislation was enacted in 1984 ruling out economic or financial needs as “just cause” for dismissal, effectively entitling workers with the severance payment benefit.41 The 1990 reform extended the minimum severance payment to eleven months, but also reinstated the notion of dismissals with “economic cause.” Nonetheless, it placed the burden of proof of “economic cause” on employers while imposing a 20 percent surcharge for failure to provide legal proof in court.

131. The two sets of reform have maintained a tenure-based severance scheme that tends to lead to two contrasting effects on labor flows. On the one hand, by increasing firing costs of longer-tenured workers, it reduces aggregate job destruction flows, but shifts the burden to shorter-tenured workers, who generally are entering the labor force, in particular young workers.42 More important, it decreases the incentives of shorter-tenured workers to acquire specific skills as they face a much higher probability of dismissal.43 On the other hand, by making the ex-post cost of employment higher, such policy discourages hiring, hence reducing the gross job creation flows. Having a lower job creation rate implies in general a higher unemployment duration, which imposes an extra costs to the matching process: the longer the worker is unemployed, the harder it is for him/her to maintain his/her skills.44

132. Recent studies on labor market legislation construct various indexes seeking to quantify the overall costs of job security regulations in the case of Chile and compare it to other countries.45 Among the most relevant findings of the this literature, we have that job protection has increased in Chile since the early 1980s and it remains among the highest in from an international comparative perspective. More important, studies show that job security policies have a substantial impact on the level and distribution of employment in the case of Chile as they adversely affect employment of young workers.46

Collective bargaining and wage setting regulations

133. The Labor Reform in the 1980s represented a major push towards labor market liberalization by opening labor negotiations to market forces. It made union affiliation voluntary and decentralized collective bargaining to limit it at the firm level. Moreover, it limited the power of workers to repudiate preexisting contracts by allowing temporary layoffs and replacements in case of strikes. While ruling out economy-wide wage adjustments, these reforms introduced a minimum wage setting. Without departing from the main liberalizing thrust of the previous regulation, the 1990s reforms imposed stricter conditions for worker replacement in case of strike and reinstated severance payments to striking workers that are laid off. In 1998, the government committed to an increase of the minimum wage of 40 percent in nominal terms over a period of three years, resulting in a 25 percent in real terms over that period.47

Social security privatization

134. In 1981, the military regime introduced a major social security reform replacing the inefficient, government-run pay-as-you-go system for a privately managed system based on individual retirement accounts.48 The new system mandates individuals to save for their own retirement through individual contributions managed by competing private administrators (AFP). As AFPs use individual contributions to invest in financial assets rather using them to finance the consumption of current retirees, the new system brought a significant reduction in payroll taxes.

135. The reform has lowered social security contributions to about 20 percent of gross earnings, down from 30 percent in 1979.49 Gruber (1997) estimates that the payroll tax burden fell by 25 percent on Chilean firms. A reduction in payroll taxes entails, in principle, lower labor costs encouraging employment creation and labor force participation. Even though there is mixed evidence on the incidence of payroll taxation on unemployment,50 the reform increased the competitiveness of firms and reduced the proportion of the social contributions perceived as pure tax. More important, Chile has benefited from a more efficient system that has raised savings and capital investment, increasing the economy’s potential output growth.

F. Conclusion

136. This chapter has argued that negative cyclical shocks leading to a significant labor reallocation across economic sectors have an effect on unemployment persistence. It has presented both theoretical arguments and empirical evidence supporting this view, based on the simple observation that the aggregate downturn in 1998–99 was led by labor-intensive sectors, while the subsequent recovery was concentrated on more capital-intensive sectors.

137. The simple model developed illustrates how economic disturbances inducing labor to be reallocated across sectors lead to a sharp decline in job creation and an increase in unemployment duration due to job search costs. Using a simple labor matching approach, the analysis demonstrates how job search and labor recruiting costs could introduce sufficient friction to the labor market reallocation process to allow for a slow adjustment, specially when economic sectors are unevenly exposed to the negative impact of aggregate shocks. The model thus provide a rationale for why unemployment persistence could be higher in the context of a cyclical aggregate downturn.

138. The empirical analysis shows that the employment dispersion across economic sectors, after controlling for the impact of aggregate shocks, has a positive and significant effect on unemployment. Nonetheless, this effect is not a long-lived one and explains only partially the high unemployment persistence. The analysis suggests that the protracted unemployment cycle would run its course as recovering economic activity in labor-intensive sectors translates into higher job creation.

139. Nonetheless, a business-cycle approach to understanding unemployment persistence in Chile remains a limited one insofar as it does not explicitly consider institutional factors introducing rigidities in labor markets. As argued in the text, labor market regulations could adversely affect labor market outcomes and dynamics by reducing incentives to work, increasing employment costs and altering the labor matching process. As suggested by a large literature on labor markets, the minimum wage and job protection policies tend to have a negative impact on the level and distribution of employment as they adversely affect employment of young workers. In particular, high severance costs reduce the job creation response to output recovery as firms want to make sure that growth is sustained before hiring additional workers. Similarly, wage indexation prevents the downward adjustment of real wages despite the pressure from labor markets as the pool of unemployed workers increases following the downward output adjustment. Correspondingly, these institutional factors together with the nature of cyclical adjustments across sectors represent important elements explaining the high unemployment persistence in Chile.

APPENDIX Basic model framework

140. We first describe the main assumptions on the economy underscoring its production technology, its labor market matching function, agents preferences and resource constraints. We explain how we could aggregate across individuals and derive a social planner maximization problem. This allows us to characterize the maximization problem in terms of a Bellman formulation. We then derive the optimization conditions and describe the properties of the solution.

141. Consider a two-sector economy populated by a continuum of identical, long-lived individuals ω∈[0.1]. In each period, these agents in the economy either work for one of the sectors in the economy or are unemployed. Let ni be the fraction of individuals working for each sector i∈{1,2}.

142. Production technology: Assume a very simple production function with constant returns to scale where the only input is labor described by the number of total hours employed in production: yi=fi(nili)=ainili. The parameter αi corresponds to labor productivity. Our measure of labor li measures the total hours employed per worker net of the effort spent by workers in the matching process: li = hi-ei. That is, workers in sector i not only spend time producing good xi but are also responsible for recruiting. So niei represents sector i total recruitment effort.

143. Matching technology: Following Mortensen and Pissarides (1994), we represent a job-search process in terms of a matching function describing the flow at which unemployed workers meet vacancies in a particular sector. The matching process is similar to a production process in which employment is the outcome from the search and recruitment efforts supplied by workers and employers. Let mi = niM(ei,u) be the flow of matched workers in sector i given that the unemployment rate is u= 1 – Σini and the recruiting effort ei. As in Mortensen and Pissarides (1994), we assume an exogenous job termination rate γ and a Cobb-Douglas matching function:M(et,u)=βu0ei1θ. With this specification we can describe the following employment transition function of for each sector:


Correspondingly, next period’s employment in sector i depends the fraction of workers that stay in the market.

144. Preferences: Assume a time separable utility function for the infinitely live individual ω having a constant discount factor δ, constant elasticity of substitution among consumption goods ct =(c1t,c2t) and separable leisure. The net present value of individual ω is equal to:


where utility function follows a simple log specification:


145. In this economy, the agents make decisions about consumption ct and total working hours ht, while firms choose the amount of recruitment effort eit. Given that all individuals are identical ex-ante, their consumption and working hours allocation should be the same. Hence, the competitive equilibrium solution is equivalent to solving the allocation problem of a social planner maximizing aggregate utility wW(ω)dω given total output. Given the recursive structure of this infinite-horizon, deterministic problem, we formulate the economy’s allocation problem in terms of a Bellman function representation:


146. In this programming problem, the social planner, given current labor distribution by sector nt = (n1t,n2t)’, chooses current total labor hours and recruitment effort in each sector (ht,et)=((h1t,h2t)’,(e1t,e2t)’) to maximize total current utility of consumption net of total labor disutility. The first constraint of the maximization program corresponds to the transition equation of the state vector nt determining the labor market matching process, characterized by the matching function and job termination rate. The second constraint represents the resource constraint for each of the sectors.

Lemma 1: There exists a solution to the dynamic programming problem above such that the implied dynamic system converges to a unique steady state solution with the following characteristics:

  • (i) There is positive Unemployment in steady state: u*=1n1*n2*>0;

  • (ii) Average labor hour per is the same in each sector: h*=h1*=h2*;

  • (iii) Total recruitment effort is the same in each sector: e*=e1*=e2*.

Proof: The proof of the existence and uniqueness of the value function is given by the results in Chapter 4 of Lucas and Stokey (1989). To find the characteristics of the steady state solution solve the first-order condition and envelope condition at the steady state:


From the first constraint at the steady state, we have that ei=(γβuθ)11θ=e for any i. Similarly, after some algebraic manipulation of the FOCs and ENV, we have that: hi=(1+βδ(1θ)γ)(ηη1)e for all i.

We have thus from the FOC that u(c)ciai=ληhn1 for all i, implying that relative shadow price of good j with respect to good a corresponds to the relative labor productivity of each sector ajai.

Finally to show that unemployment is positive in the steady state, we do it by contradiction. Suppose u = 0, then e → ∞ implying that u(c)ci which is a contradiction.

147. We make the following assumptions on the parameters of the model in order to do the simulation:

Preference parameters: consumption elasticity of substitution parameter α = 0.4; leisure substitution parameter μ = 8; discount factor δ = 0.95˄(1 / 4).

Technology parameters: productivity in sector 1 a1 = 150; productivity in sector 2 a2 = 75. Labor matching technology parameters: unemployment matching elasticity θ = 0.2; termination rate γ = 0.2˄(1/4), β parameter is chosen so that the unemployment rate in the steady state is 6.5 percent.

148. The consumption elasticity parameter is chosen arbitrarily so that agents prefer to consume more of the good produced in sector 2. Also we assume that sector 1 is twice as productive as sector 2. These two assumptions imply that sector 2 is more labor intensive than sector 1. The steady state employment share in these two sectors is (n1,n2)= (0.36,0.575).


  • Andolfatto, D., 1996, “Business Cycles and Labor-Market Search,” American Economic Review, Vol. 86 (1), pp. 112132.

  • Abraham, K. and L. Katz, 1986, “Cyclical Unemployment: Sectorial Shifts or Aggregate Disturbances?” Journal of Political Economy, Vol. 94, No. 3, pp.507522.

    • Search Google Scholar
    • Export Citation
  • Davis, S.J., J.C. Haltiwanger, and S. Schuh, 1996, Job Creation and Destruction, Cambridge: MIT Press.

  • Diamond, P. and S. Valdés-Prieto, 1994, “Social Security Reform,” in Boswroth, B., R. Dornbusch, and R. Laban eds. The Chilean Economy: Policy Lessons and Challenges, Washington DC: Brookings Institution.

    • Search Google Scholar
    • Export Citation
  • Edwards, S. and A. Cox Edwards, 2000, Economic Reforms and Labor Markets: Policy Issues and Lessons from Chile, NBER Working Paper No. 7646, April 2000.

    • Search Google Scholar
    • Export Citation
  • Engel, E. and A. Hernando, 2000, “Tipo de Cambio Real y Flujos Brutos de Empleo, Informe Final, Banco Central de Chile, November.

  • Greenwood, J. G. MacDonald, G. Zhang, 1996, “The Cyclical Behavior of Job Creation and Job Destruction: A Sectorial Model,” Journal of Economic Theory, Vol. 7, pp. 113124.

    • Search Google Scholar
    • Export Citation
  • Grubel, J., 1997, “The Incidence of Payroll Taxation: Evidence from Chile,” Journal of Labor Economics; 15(3).

  • Hansen, G., 1985, “Indivisible Labor and the Business Cycle,” Journal of Monetary Economics, Vol. 29, pp. 324.

  • Heckman, J. and C. Pagés, 2000, The Cost of Job Security Regulation: Evidence from Latin American Labor Markets, Inter-American Development Bank, Working Paper No. 430.

    • Search Google Scholar
    • Export Citation
  • Hendry, D., 1993, Econometrics: Alchemy or Science? Oxford: Blackwell.

  • Lilien, D., 1982, “Sectorial Shifts and Cyclical Unemployment,” Journal of Political Economy, Vol. 90, No. 2, pp. 777793.

  • Lucas, R, S, and N. Stokey, 1989, Recursive Methods in Economic Dynamics, Cambridge: Harvard University Press.

  • Merz, M., 1995, “Search in the Labor Market and the Real Business Cycle,” Journal of Monetary Economics, Vol. 36, pp. 269300.

  • Millard, S., A. Scott, and M. Sensier, 1997, “The Labour Market over the Business Cycle: Can Theory Fit the Facts?” Oxford Review of Economic Policy, Vol. 13(3), pp. 7092.

    • Search Google Scholar
    • Export Citation
  • Mills, T.C., G. Pelloni, and A. Zervoyianni, 1995, “Unemployment Fluctuations in the United States: Further Tests of the Sectorial-Shifts Hypothesis,” Review of Economics and Statistics, Vol. 77, Issue 2 (May), pp. 294304.

    • Search Google Scholar
    • Export Citation
  • Mizala, A., 1998, “La Regulación del Mercado Laboral en Chile,” Perspectivas en Política, Economía y Gestión, Vol. 1, No. 2.

  • Mizala, A. and P. Romaguera, 1996, “Capacitación y Características Educacionales de la Fuerza de Trabajo,” Colección de Estudios CIEPLAN No. 43, September.

    • Search Google Scholar
    • Export Citation
  • Mortensen, D.T. and C.A. Pissarides, 1999, “New Developments in Models of Search in the Labor Market,” in Handbook of Labor Economics, eds. O. Ashenfelter and C. Card, Amsterdam: North-Holland.

    • Search Google Scholar
    • Export Citation
  • Nickell, S., 1997, “Unemployment and Labor Market Rigidities,” Journal of Economic Perspectives, Vol. 11 (3), pp. 5574.

  • OECD, 1994, The OECD Jobs Study: Evidence and Explanations, Paris: OECD, Vol 1-2.

  • Pagés, C. and C. Montenegro, 1999, Job Security and the Age Composition of Employment: Evidence from Chile, Inter-American Development Bank, Working Paper No. 398.

    • Search Google Scholar
    • Export Citation
  • Prescott, E., 1986, “Theory Ahead of Business-Cycle Measurement,” Carnegie-Rochester Conference on Public Policy, Vol. 24 (11), pp. 4462.

    • Search Google Scholar
    • Export Citation
  • Phelan, C. and A. Trejos, 2000, “The Aggregate Effects of Sectorial Reallocations,” Journal of Monetary Economics, Vol. 45, pp. 249268.

    • Search Google Scholar
    • Export Citation
  • Pissarides, C.A., 2000, Equilibrium Unemployment Theory, Cambridge MIT Press. 2nd Edition.

  • Rogerson, R., 1987, “An Equilibrium Model of Sectorial Reallocation,” Journal of Political Economy, Vol. 95, No. 4.

  • Sapelli, C., 1996, “Modelos para Pensar el Mercado de Trabajo: Una Revisión de la Literatura Chilena,” Cuadernos de Economía, No 99, August.

    • Search Google Scholar
    • Export Citation

Prepared by Rodolfo Luzio. The author would like to thank for the helpful comments provided by Saul Lizondo, Ketil Hviding, Fransisco Nadal-De Simone, Steven Phillips, and participants at Banco Central de Chile seminar in Santiago, in particular, Pablo Garcia.


Excluding government-sponsored temporary employment programs average unemployment went up to 10.5 percent in 1999.


See for instance Hansen (1985).


We follow the standard procedure of detrending data using the Hodrick-Prescott filter. Though the accuracy of this procedure has been subject to criticism, this procedure remains valid for our main purpose of pointing out the salient facts.


The actual statistic is hk=kVar(Δxt)Var(Δkxt)


These results are available from the author upon request.


See Arrau, De Gregorio, Reinhart and Wickham (1995) for a study assessing the effect of financial innovation on the demand for money in developing economies.


Results from tests for misspecification and parameter constancy can be obtained from the author upon request.


For a more detailed review of labor market reforms, see Mizala (1998) and Edwards and Cox Edwards (2000).


The principle of “just cause” for dismissal includes “grave faults” such as criminal behavior and absenteeism.


Using Chilean data from 1960 to 1997, Pages and Montenegro (1999) shows the distortions of tenure-job security on employment composition by reducing youth employment.


Mizala and Romaguera (1996) evaluates job training programs developed by the Chilean government. In particular, they study participation rates of initiatives aiming at young workers such as Chile Joven.


Heckman (1990) shows a positively sloping hazard function for the conditional probability of remaining unemployment.


See, in particular, Edwards and Cox Edwards (2000) and Heckman and Pages (2000).


See Pages and Montenegro (1999) for empirical evidence.


Preliminary results from a recent study by Bravo and Contreras indicate that the increase in the minimum wage since 1997 has had a negative impact on employment among young workers.


Diamond and Valdes-Prieto (1993) provide a detailed analysis of the privatized system and a comparison to the previous system.


Of the 20 percent contribution, 10 percent goes toward retirement, 7 percent towards heath and 3 percent towards disability.


Gruber (1997) finds that the incidence of lower payroll taxation was translated into wages, with no effect on employment.

Chile: Selected Issues
Author: International Monetary Fund