generate a nonneutrality of money as proposed by the Keynesian model. Instead, results support the view that contracts cannot propagate nominal shocks.
The Link Between Output and Contracts Under Hyperinflation CAROL A. DOLE,* DAVID A. DENSLOW,** AND MARK RUSH**
In the conventional Keynesian model, nominal wage contracts (acting as a friction) transmit monetary shocks to real variables. In contrast, the new classical or real business cycle theory claims that firms and workers ignore the behavior of the actual real wage and instead generate an efficient level of employment (hence, output) based on a shadow real wage. Using Brazilian data covering a period during which the economy suffered hyperinflation and wage contracts were indexed by the government, results show that these fixed nominal wage contracts did not generate a nonneutrality of money as proposed by the Keynesian model. Instead, results support the view that contracts cannot propagate nominal shocks. (JEL E3)
Introduction
In the Keynesian [1964] framework, nominal wage contracts transmit monetary shocks to real variables. This non_neutrality occurs because over the period of a fixed wage contract, an increasing price level erodes the real wage, thereby inducing firms to increase their labor input and, likewise, output. That is, increases in the money supply raise the price level. Because the nominal wage remains unchanged, firms move along their demand curves and, thus, output responds positively to the increase in the money supply. Fischer [1977] presents a typical model that assumes nominat wages are rigid to some degree and that the level of employment is adjusted in accord with the firm's labor demand schedule. ~Firms can raise their labor input in a variety of ways, including hiring more workers, eliciting greater work intensity, using overtime hours, or a combination of these choices. 2 To represent the persistence empirically, however, Fischer's model must be altered. Taylor [1979] shows that real effects exhibit persistence beyond price and wage adjustments so long as the contracts are staggered and terms are adjusted periodically. Moreover, Taylor [1980] obtains results that mimic the persistence in unemployment and output similar to the U.S. postwar experience (with respect to nominal disturbances) using a model with staggered wage contracts. Gali [1992] suggests that the IS-LM model fits the postwar U.S. experience fairly well. Providing additional support, Card's [1990] empirical analysis finds that wages negotiated at the start of a contract period are positively correlated to terms made at the end of the previous period. This linkage creates persistence in the costs of labor and, therefore, employment.
*Universityof North Carolinaat Charlotteand **Universityof Florida--U.S.A. 140
DOLE ET AL.: HYPER1NFLATION
141
In contrast to the Keynesian model, an alternative view asserts that monetary shocks are neutral. Neutrality occurs as a result of firms and labor ignoring the movement of the actual real wage and, instead, basing their employment decisions on an equilibrium or shadow real wage) Analyzing this efficient behavior, Barro [1977] has pointed out that even though a labor contract fixes the nominal wage, firms and workers can equate the marginal product of labor to the marginal value of workers' time over all states of nature and, thus, jointly determine the efficient level of employment. Hall [1980], in extending Barro's work, notes that the relationship between labor and management is long-term in nature. Therefore, either party is willing to occasionally lose as a result of unanticipated price level changes, knowing that they may well win with the next unexpected change in the price level. In essence, neither party reacts to monetary shocks because, on average, an efficient level of employment is obtained where the marginal product of labor equals the equilibrium real wage, which equals the marginal value of worker's time. 4 Hence, the actual real wage, affected by nominal disturbances, does not equilibrate the labor market because it does not produce an efficient level of employment. Instead, both labor supply and demand are taken into account and jointly determine the level of employment at the shadow equilibrium real wage. Therefore, when the money supply changes, shadow real wages and, hence, employment are unchanged as money wages and prices change proportionately. In this framework, wage contracts are no longer a source of nonneutrality. It is apparent though that because of long-term relationships, real disturbances can generate changes in the marginal product of labor, employment, and output. Allowing both real and nominal disturbances to generate the same output response is, of course, the major difference between the Keynesian and new classical or real business cycle approaches to labor contracts. On a related point, Hall discusses the fact that to meet an increase in demand, firms can expand their labor input in a variety of ways: hire more labor, ask current workers to work overtime, persuade them to work more intensely, or a combination of these options2 This present study uses Brazilian data to examine whether macroeconomic shocks can be linked to output behavior via nominal wage contracts.6 Brazil's wage indexation policy allows for examining and testing the opposing theories. Brazilian firms generally fixed their employees' nominal wages but were required by law to adjust workers' salaries a t specified intervals to keep pace with inflation and productivity. From 1965 to 1979, the adjustments were required annually. Because of the lagged indexing formula and high Brazilian inflation, real wages fell until readjustment occurred, when they were raised to maintain purchasing power. According to Fischer's model, as real wages fell prior to the indexing date, firms would add workers and output would respond positively. Conversely, the real business cycle approach suggests that firms and workers would not change the level of employment in the face of monetary shock, so there would be no tendency for contracts to influence output. In the empirical work, in addition to considering the relationship between the real wage and employment, this paper also investigates the correlation between the real wage and output, providing more insight into the effects of nominal s h o c k s . 7
142
AEJ: JUNE 2000, VOL. 28, NO. 2
The second section discusses the Brazilian economic environment, and the third section examines past empirical work. The fourth section describes the data, the fifth section presents the estimation technique and regression results, and the sixth section summarizes the results. The Brazilian Economy
Between 1945 and t964, a democratic government managed the Brazilian economy. During this period, economic growth varied between 10 percent annually in 1954 to t .5 percent annually in 1963. This downturn in the economy was accompanied with hyperinflation, sometimes as much as 80 percent annually. Tired of the faltering economy, a military regime overthrew the government in March 1964. Targeting unions and their demands for higher wages as a primary cause of the uncontrollable inflation, the military government legislated the first wage indexation formula in June 1964. Under the law, wages were adjusted annually based on past inflation, expected inflation, and productivity improvements. At the same time, the government passed laws that, for the most part, banned strikes. In 1968, weaknesses in the original wage indexation formula were corrected. Specifically, an average real wage was included in the new formula to correct wage deterioration. Between I968 and 1974, the economy enjoyed economic growth, averaging over 10 percent annually. In 1974, the government tried to improve the wage adjustment process again. As part of this change, a refined forecasting procedure was developed and the weighting of productivity improvements was adjusted. These changes did not keep inflation rates at bay however. Additional refinements were made in 1976 to address worker complaints about eroding real wages. As the inflation rate continued to soar, the government made a pivotal change in 1979: wages were to be adjusted semiannually. Previous Analyses
To test the role of wage contracts under the opposing theories, empirical analysis centers on three primary approaches. The first examines the pure time series relationship between real wages and employment activity. A countercyclical real wage, which the Keynesian approach suggests, would occur as the price level rises and labor demand (along with output) increases. Early work [Dunlop, 1938; Tarshis, 1939], though, found a procyclical real wage. More recently, mixed results have been obtained. For instance, Geary and Kennan [1982], studying Canadian data, find no statistical relationship between employment and the real wage. However, using disaggregated data, Bils [1985] finds a positive relationship between real wages and output. In addition to his use of panel data, Bils accounts for these differences due to the inclusion of overtime earnings and using a later sampling period than earlier studies. It is not surprising that these results lend uncertainty to the worth of nominal wage rigidity models, given the somewhat simplistic time series models that are necessary to assume.
DOLE ET AL.: ffYPERINFLATION
143
In a related method, Dole et al. [1999] used nationwide Brazilian data to examine whether a relationship exists between inflation and output. As in this paper, they exploit the legally enforced nominal contracts that existed in Brazil to examine whether changes in the inflation rate played a role in determining changes in industry output. They fail to find a relationship between inflation and industry output, which they interpret as supporting the new classical or real business cycle approach. Finally, Ahmed [1987], Card [1990], and Keane [1993] approach the debate using a different strategy. Ahmed investigates whether employment and output are sensitive to the degree of wage indexation. Essentially, if nominal wage contracting is supported, a monetary shock will have a greater impact on employment and output, the less indexed (that is, less flexible) the wage. Using Canadian manufacturing industries, which employ a variety of indexing schemes, Ahmed tests this implication of contracting models. He concludes that nominal wage rigidity cannot account for business cycles since he failed to find a relationship between the degree of contracting and the extent to which a monetary shock affects real variables. On the other hand, Card claims that it is the unexpected components of the real wage (namely, unanticipated price changes) along with the degree of indexation that should generate a countercyclical real wage. Therefore, he (also using Canadian data) constructs measures of unexpected inflation and real wage changes for use in a labor demand model. These results support a role for surprise inflation in impacting employment levels via end-of-contract real wage levels. Hence, Card concludes that labor and output are materially affected by nominal wage contracts. Using data from the National Longitudinal Study of Young Men, Keane studies the response of real wages to unanticipated inflation and monetary growth. He finds no evidence that unexpectedly higher inflation lowers the real wage. Given these conflicting results, this strand of literature fails to produce definitive conclusions. The Data
Specification In comparison to Ahmed, Card, and Keane, the data in this paper has a major advantage. Because inflation in Canada was relatively low and the unexpected inflation Card tries to measure is even smaller, both Ahmed and Card must search for relatively small effects. 8 Inflation in Brazil, though, was significantly more pronounced. Thus, if there are any effects from nominal contracts, the data here may be better able to disclose them. 9 The data start in 1965 with the implememation of the indexing policy and continue through 1976.1° The basic data consist of monthly output series from Anudirio Estatistico do Brazil [Instituto Brasileiro de Geografla e Estatistica (IBGE), various] for seven industries: machinery, tobacco, plastics, chemicals, beverages, transportation, and rubber. 11 Using the wholesale price index (WPI) from Conjuntura Econ6mica, the nominal wage is deflated. It is important to account for the fact that the real wage becomes endogenous for part of the sample period. As discussed more thoroughly below, governmem spending and the money supply were for instruments in the regression analysis. Conjuntura EconOmica
144
AEJ: JUNE 2000, VOL. 28, NO. 2
[IBGE, various] and Indtistria de Transformaggto [IBGE, various] are sources for both series, and they provide employment and value of production figures at the statewide level as well. The Estimation Technique and Results
A number of specifications allowing for different functional firms were estimated for the seven industry groups: EMPtj
= c + o T +
~. RWAGE t J
,J
(1)
+ ~j M O N T H L Y DUMMIES + )~j CARNIVAL + ~tt where EMPt, j reflects the number of employees at time t for industry j, T is a time trend, RWAGEt. j represents the real wage for each group, M O N T H L Y DUMMIES are dummy variables for January through November, and CARNIVAL is a dummy variable for the annual month of Carnival. In Brazil, Carnival is a pre-Lenten festival during which there is an obvious decrease in the amount of work performed. The error term, gt' is assumed to be normally distributed. Obviously, (1) addresses the Keynesian suggestion that firms choose to hire more labor as the real wage changes. However, another possibility exists. For example, when there is more work to do, firms choose to hire new employees, ask current ones to work harder, or both. For example, during the 1970s in the U.S., Hall [1980] finds that when output contracted, decreased work effort (output per worker and hours per worker) accounted for half of output decline, while a decrease in actual employment accounted for the other half. He surmises that the same would occur during booms. Thus, if firms opt to work employees more intensely or increase their hours, there may only be a change in output and the real wage. Equation (2) allows for this behavior: VALUE,./
= c +
o T + ¢x. RWAGEt, (2)
+ [Sj M O N T H L Y DUMMIES + ~,j CARNIVAL + ~tt where VALUEr, j is the value of production in real terms for each industry. In addition to estimating the equations in levels, also estimate them in differences: DEMPt, j = c + ~ DRWAGEt, j
(3) + ~j MONTHL Y DUMMIES + )~.jDCARNIVAL + gt where DEMPt, j and DRWAGEt, j are first differences of employment and the real wage at time t for industry j, respectively.
DOLE ET AL.: HYPERINFLATION
145
Corresponding to (2), a regression in differences is also estimated: DVALUE, t, j = c + c~.jD R W A G E t' J
(4) + ~j M O N T H L Y D U M M I E S + ~"s D C A R N I V A L + ~tt
where VALUEr, j and D R W A G E t , j are the first differences of the value of output and the real wage at time t for industry j, respectively. Recall that the real wage is calculated by deflating the nominal wage by the national WPI. Ordinary least squares would be appropriate if the real wage was always determined by a fixed nominal wage and the (nationwide) WPI because then the (industry-specific) error term would be uncorrelated with the real wage. However, an issue arises because the real wage is endogenous during some months. Thus, estimating these equations requires two-stage least squares where the observations on nominal wages are exogenous for all months except when they are renegotiated.12 That is, during the industry's base month, the wage becomes endogenous in light of the fact that the government mandates only a minimum wage adjustment. Workers may (and quite frequently do) negotiate for additional increases. Failing to account for this fact (that the real wage becomes endogenous when readjusted), the estimated coefficient on the real wage would suffer from simultaneous equation bias. Thus, for those months when the real wage was endogenous--and only for those months--an instrument is formed using price level, government spending, money supply, and oil prices as instrumental variables. 13 The second stage takes this now-exogenous real wage series (the endogenous observation having been estimated unbiasedly) and performs regressions (1) through (4). In accordance with the Keynesian mode, a countercyclical real wage is expected. That is, in response to changes in the real wage, firms should move along their labor demand curves and hire additional labor. Conversely, the approach emphasizing efficiency suggests the absence of a relationship between the real wage and labor demand. Firms and workers ignore the behavior of the actual real wage, basing their joint decisions on a shadow real wage. It is expected that a will not differ significantly from 0 in this case. Results
The results from all four regressions provide no support for the Keynesian model. Tables 1 and 2 provide estimates of 0~, the coefficient on the real wage. A majority of the estimates are significantly positive--a perverse result from the Keynesian perspective, suggesting that as the real wage falls, firms hire less labor and employment (as well as output) falls. 14For the specifications in differences, the results are similar. These reports provide overwhelming support for the efficient model. Most telling, however, is the complete lack of any negative a's. Recalling that the Keynesian model stresses labor contracts as the key to allowing nominal shocks to affect real variables, the failure to find any negative a's attaining conventional levels of significance is a striking result. It strongly suggests that even in a situation in which labor contracts are institutionalized by law, firms do not slide up or down their labor demand curve to unilaterally determine the level of employment or
146
AEJ: JUNE 2000, VOL. 28, NO. 2
output. Thus, the results here support the theory that nominal wage contracts are unimportant in transmitting changes in the money supply to output. TABLE 1 Estimates of t~ in (1) and (2)
(1) Industry
(2)
Coefficient
P-Value
Coefficient
P-Value
.518
.008*
.309
.554
Tobacco
1.315
.000"
.091
.685
Plastics
.711
.000"
1.253
.000"
Chemicals
-.432
.716
-3.691
.575
Beverages
.611
.000"
2.433
.000"
Transportation
.601
.000"
.807
.000"
Rubber
.750
.000'
.258
.376
Machinery
Notes: * denotes significanceof the coefficient c~ at the 1 percentlevel. TABLE2 E~imatesof~in ~)and(~
(3) Industry
(4)
Coefficient
P-Value
Coefficient
P-Value
Machinery
.289
.000"
.699
.009"
Tobacco
.089
.001"
.046
.632
Plastics
.116
.001"
.382
.000"
Chemicals
.089
.038**
.762
.000"
Beverages
.295
.000"
.562
.003"
Transportation
.179
.000"
.788
.000"
-. 198
.484
Rubber
4.731
.042
Notes: * and **denote significanceof the coefficientc~at the 1 and 5 percentlevels, respectively.
DOLE ET AL.: HYPERINFLATION
147
Summary and Conclusions In a neoclassical model, the effects of monetary shocks are transmitted throughout the economy by introducing frictions. In this paper, the frictions analyzed are nominal wage contracts. From a conventional or Keynesian standpoint, these contracts prohibit the nominal wage from changing while allowing the real wage to be determined by movements in the price level. Firms, basing their employment on the real wage, respond to monetary shocks, for instance, by increasing employment when the price level rises. Ultimately, these frictions propagate nominal shocks, and real variables are affected. Conversely, new classical and real business cycle models point out that firms and workers, both part of a long-term relationship, may behave efficiently, jointly determining employment based on an equilibrium (shadow) real wage. This wage equates the marginal product of labor as well as the marginal value of workers' time. Both parties ignore the movement of the actual real wage as determined by wage contracts and the price level, effectively rendering monetary shocks neutral. Brazil's unique wage indexation system allows these opposing theories to be examined and compared. Using statewide data, results provide little support for the Keynesian view of a countercyctical relationship between the real wage and employment or output.
Footnotes 1. 2. 3. 4.
5.
6.
7.
8.
Phelps and Taylor [1977] present a model similar to Fischer's. Hall [1980] elaborates on these options in a new classical framework. Indeed, this observation can justify the neglect of nominal wages and the money supply within a "pure" real business cycle framework. Recognizing the Barro and Hall analysis may be one reason new Keynesians have made the distinct departure from the sticky wage framework. Specifically, Ball et al. [1988] point out that firms and their customers are not part of a long-term relationship---no loyalty exists to prevent buyers from moving up their demand curves as price-setting firms maintain high real product prices in the face of tight monetary policy. This lack of long-term arrangements, then, can produce business cycles, given sticky prices in the goods market. Actual data tend to show that work intensity changes proportionately with output as opposed to employment varying with output. Okun's Law expresses this fact, claiming that a 1 percent change in unemployment rates produces a 3 percent corresponding change in output. In empirical work, both labor and total output are used in separate regressions to allow for both options. Dole et al. [1999] used nationwide data, but nationwide data for industry do not exist. Hence, proxies were required for the real wage. This paper, however, takes advantage of the fact that real wages do exist at a statewide level by using real wage data from S~o Paulo. Data on employment was found only in terms of the number of workers, not hours. If firms do not change the number of employed workers but instead adjust the work intensity or hours, then the effects of a change in the real wage may only be seen in output, not employment. Indeed, this could be why Ahmed failed to fmd any effect from monetary disturbances. Card's results, meanwhile, may be the result of mismeasuring the unexpected component of inflation.
148
AEJ: JUNE 2000, VOL. 28, NO. 2
9.
However, this advantage may also conceal a related disadvantage. One might worry that in high-inflation environments, workers and firms would make efficiency-enhancingadjustments (as Barro suggested) that they might forego in less inflationary times. That is, it might be the case that in low-inflation countries, nominal wage contracts function as postulated by Keynesian analysis, while in high-inflation countries, the contracts are altered to be in line with Barro's approach. Of course, the observation that individuals in high-inflation countries may alter their behavior when compared to residents in low-inflation countries is hardly unique to this study. Also, given the prominence with which high-inflation countries have served as laboratories for testing monetary phenomena, most economists apparently think this potential drawback to be generally minor. 10. It appears that statewide data collection for wage, employment, and output ended in 1976. 11. The wage adjustment month for workers within these industries varied by industry. The authors were unable to uncover a written record of the precise month when the wages were indexed for some of the industries. Thus, to determine the indexing month, wage data from Indtistria de Transforma¢ao [IBGE, various] were examined for these industries from Brazil's two largest states, S~o Paulo and Minas Gerais, because output in these states effectivelydominated the rest of the country. The nominal wage was deflated using the wholesale price index from Conjuntura Econgmica. These seven industries then demonstrated a consistent pattern. In particular, during the period of annual indexing, all industries had one month where the real wage jumped higher and thereafter tended to decline. Thus, the month of the jump was deemed the month that indexing occurred. 12. This estimation technique is described by Kohli [1989]. 13. The source for the price level, money supply figures, and government spending was Conjuntura Econ~mica. International Financial Statistics [International Monetary Fund, various] provided additional data for spending. 14. If anything, these results tend to provide support for the stream of real business cycle literature that predicts procyclical wages.
References Ahmed, Shagil. "Wage Stickiness and the Nonneutrality of Money: A Cross-Industry Analysis," Journal of Monetary Economics, 20, July 1987, pp. 25-50. Ball, Laurence N.; Mankiw, Gregory; Romer, David. "The New Keynesian Economics and the Output-Inflation Trade-Off," Brookings Papers on Economic Activity, 1988, pp. 1-65. Barro, Robert J. "Long Term Contracting, Sticky Prices and Monetary Policy, "Journal of Monetary Economics, 3, July 1977, pp. 305-16. Bils, Mark J. "Real Wages Over the Business Cycle: Evidence from Panel Data," Journal of Political Economy, 93, August 1985, pp. 666-89. Card, David. "Unexpected Inflation, Real Wages and Employment Determination in Union Contracts," American Economic Review, 80, September 1990, pp. 669-88. Dole, Carol A.; Denslow, David A.; Rush, Mark. "The Impact of Labor Contracts: Evidence from Brazil," Economic Inquiry, 37, January 1999. Dunlop, John. "The Movement in Real and Money Wage Rates, "Economic Journal, 48, September 1938, pp. 349-66. Fischer, Stanley. "Long Term Contracts, Rational Expectations and the Optimal Money Supply Rule," Journal of Political Economy, 85, February 1977, pp. 191-206.
DOLE ET AL.: HYPERINFLATION
149
Gali, Jordi. "How Well Does the IS-LM Model Fit Postwar U.S. Data?," Quarterly Journal of Economics, 108, May 1992, pp. 709-38. Geary, Patrick; Kennan, John. "The Employment-Real Wage Relationship: An International Study," Journal of Political Economy, 1982, pp. 854-71. Geary, Patrick; Kennan, John. "Comment: The Employment-Real Wage Relationship, An International Study," Journal of Political Economy, 1982, pp. 854-71. Hall, Robert E. "The Rigidity of Wages and the Persistence of Unemployment,"Brookings Papers on Economic Activity, 2, 1980, pp. 301-35. Instituto Brasileiro de Geografia e Estatistica. Annudrio Estatistico do Brazil, various. m . Conjuntura EconOmica, various. m . Ind~stria de Transforma~ao, various. International Monetary Fund. International Financial Statistics, various. Keane, Michael. "Nominal-Contracting Theories of Unemployment: Evidence from Panel Data," American Economic Review, 83, September 1993, pp. 932-52. Keynes, John. The General Theory of Employment, Interest, and Money, New York, NY: Harcourt, Brace and World, 1964. Kohli, U. "Consistent Estimation When the Left-Hand Variable is Exogenous Over Part of the Sample Period," Journal of Applied Econometrics, July-September 1989, pp. 283-93. Phelps, Edmund S.; Taylor, John B. "Stabilizing Powers of Monetary Power Under Rational Expectations," Journal of Political Economy, 85, February 1977, pp. 163-90. Tarshis, Lori. "Changes in Real and Money Wage Rates," Economic Journal, 49, March 1939, pp. 150-4. Taylor, John B. "Staggered Wage Setting in a Macro Model,"American Economic Review, 69, May 1979, pp. 108-13. . "Aggregate Dynamics and Staggered Contracts," Journal of Political Economy, 89, February 1980, pp. 1-23.
