The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
FEDERAL RESERVE BANK OF CLEVELAND pd papers POLICY DISCUSSION PAPERS Does Wage Inflation Cause Price Inflation? by Gregory D. Hess and Mark E. Schweitzer FEDERAL RESERVE BANK OF CLEVELAND Does Wage Inflation Cause Price Inflation? by Gregory D. Hess and Mark. E. Schweitzer Recent attention has turned from unemployment levels to wage growth as an indicator of imminent inflation. But is there any evidence to support the assumption that increased wages cause inflation? This study updates and expands earlier research into this question and finds little support for the view that higher wages cause higher prices. On the contrary, the authors find more Gregory D. Hess is the Danforth–Lewis Professor of Economics at Oberlin College and an academic consultant to the Federal Reserve Bank of Cleveland. Portions of this text were prepared for his fall 1999 statement to the Shadow Open Market Committee. Mark E. Schweitzer is an economist at the Federal Reserve Bank of Cleveland. The authors would like to thank Allan Meltzer and Charles Plosser for very helpful comments. evidence that higher prices lead to wage growth. Materials may be reprinted if the source is credited. Please send copies of reprinted materials to the editor. We invite questions, comments, and suggestions. E-mail us at email@example.com. POLICY D I S C U S S I O N PAPERS Policy Discussion Papers are published by the Research Department of the Federal Reserve Bank of Cleveland. To receive copies or to be placed on the mailing list, e-mail your request to firstname.lastname@example.org or fax it to 216 -579 - 3050. Policy Discussion Papers are available electronically through the Cleveland Fed’s site on the World Wide Web: www.clev.frb.org/Research. ISSN 1528-4344 FEDERAL RESERVE BANK OF CLEVELAND “Unions Seek Big Pay Gains, Sparking Inflation Worries” Headline, Wall Street Journal, September 3, 1999. Introduction Wall Street economists and journalists have frequently focused on labor market activities to help foretell inflationary price pressures. Earlier in the current expansion, some considered a low unemployment rate (below 6.5 percent) a harbinger of rising inflation. The com monly held view was that if the aggregate demand for goods and services caused unem ployment to fall below some “natural” rate, inflation would accelerate. However, although the unemployment rate continued to fall throughout the 1990s, inflation never rose. This led economists to reconsider whether the “threshold” unemployment rate, termed the natural or nonaccelerating rate of inflation (NAIRU), had fallen from approximately 6.5 percent to 4.5 percent. Recent work by Staiger, Stock, and Watson (1997), however, suggests that even a time-varying natural rate of unemployment is not a very useful tool for predicting inflation. 1 With uncertainty about the unemployment rate’s reliability as an early-warning device for rising inflation, recent attention has turned to wage and compensation growth for a labor market indicator of inflation. The standard argument for how increased labor costs tend to push up prices, leading to the wage-price spiral, is as follows: “… when buoyant demand reduces unemployment (at least relative to recent experi- 1. Stock and Watson (1999) find that unemployment does not help in forecasting U.S. inflation, but that an indicator of aggregate demand does. However, in contrast to the research we present in this paper, they do not explore issues of simultaneity, measures of productivity, or unit labor costs in their forecasts. enced levels), inflationary pressure develops. Firms start bidding against each other for labour, and workers feel more confident in pressing wage claims. If the inflationary pressure is too great, inflation starts spiraling upwards: higher wages lead to higher price rises, leading to still higher wage rises, and so on. This is the wage-price spiral.”2 2. Layard, Nickell, and Jackman (1994), pp. 11. Surprisingly, the recent shift in attention to higher wages as the cause of higher prices leaves unexplained the problem of how wages get high in the first place. The intuition behind this view is that since labor costs are a large fraction of a firm’s total costs of production, rising wages and compensation should put pressure on firms to pass these higher costs on as higher prices. We have several reasons to doubt the accuracy of this view. First, if a wage increase is brought about by increased labor productivity, it will not create inflationary pressure.3 Second, a wage increase will not create inflationary pressure if it leads to a squeeze in profits because a firm cannot pass along cost increases. No firm inherits the right to simply “mark-up” the prices of its output as a constant proportion above its costs; competitive market pressures strongly influence the pricing deci- 3. Indeed, wage increases supported by productivity growth should not be considered “wage inflation” at all, but we will occasionally use this phrase as it is in the existing literature. sions of firms. Finally, causation could work in the opposite direction: An increase in aggregate demand may permit firms to raise the price of their products, and the resulting increase in profits would lead workers to demand higher wages in future negotiations. 1 P O L I C Y D I S C U S S I O N PA P E R S NUMBER 1, APRIL 2000 It turns out that the vast majority of the published evidence suggests that there is little reason to believe that wage inflation causes price inflation. In fact, it is more often found that price inflation causes wage inflation. Our recent research, which updates and expands on the current literature, also provides little support for the view that wage gains cause inflation. Moreover, wage inflation does a very poor job of predicting price inflation throughout the 1990s, while money growth and productivity growth sometimes do a better job. The policy conclusion to be drawn is that wage inflation, whether measured using labor compensation, wages, or unit-labor-costs growth, is not a reliable predictor of inflationary pressures. Inflation can strike unexpectedly without any evidence from the labor market. Data Before exploring the econometric evidence on this debate, we provide a plot of price inflation (as measured by the growth rate of the personal consumption expenditure deflator) and wage inflation (as measured by the growth of nonfarm business compensation) in figure 1.4 The series are graphed over the time period 1960:IQ to 1999:IIIQ. Although the 4. The appendix provides a full . description of each series. growth in wages fluctuates more wildly than does inflation, both series generally share the same pattern (they have a correlation coefficient of 0.408): Both trended upwards throughout the 1960s and 1970s, they both peaked and tended to decline throughout the 1980s, and then they stabilized during the 1990s. Obviously, wages and price changes are related, but the direction of causation isn’t readily apparent. As mentioned above, however, the growth of nominal wages may be a poor measure of cost pressures faced by firms, since if wage growth is driven by productivity growth, then firms will not have to pass higher wages on as higher prices. This argument also would require that employers compensate workers for productivity gains; thus, productivity should be accounted for when asking if wages are driven by inflation. Fortunately, the Bureau of FIGURE 1 G R O W T H O F U N I T L A B O R C O S T S A N D P R I C E I N F L AT I O N Annualized growth rates 16 12 Price inflation 8 4 0 –4 Growth in ULC –8 1959 1963 1967 1971 1975 1979 1983 1987 1991 1995 1999 2 FEDERAL RESERVE BANK OF CLEVELAND Labor Statistics also reports unit labor costs, that is, the labor cost to the firm of producing one unit of output, which adjusts wages for labor productivity. Figure 2 charts price inflation and the growth in unit labor costs for 1960:IQ to 1999:IIIQ. One can see that while unit labor costs are quite volatile, they tend to track inflation closely (the two series have a correlation coefficient of 0.642), but causation is no clearer. Econometric Literature While the raw data show that measures of wages and prices move strongly together, the academic literature is divided as to whether there is empirical evidence that wages cause prices. To be clear, academic economists use the term causality as in “Granger-causality.” The test for Granger-causality involves examining whether lagged values of one series (say wages) have significant in-sample explanatory power for another variable (say prices). Of course, both variables may “Granger-cause” one another, in which case one can conclude only that both economic series are determined simultaneously; hence, a researcher cannot conclude that one series has an independent causal effect on the other. The matter becomes even more complicated if the series in question are “cointegrated,” which is the case if the levels of the series move together over the long run, even though the individual series are best modeled in growth rates. In this case, the researcher must be careful to include “error correction terms” in the Granger-causality tests to allow the series to catch up with one another.5 The significance of the error correction terms in the Granger-causality test simply reflects the fact that the series in question are driven to return to a long-run equilibrium relationship that is noncausal. Furthermore, researchers’ conclusions about the causal effects between wages and 5. The omission of these error correction terms in a Grangercausality test specified in growth rates would lead to a standard omitted variable bias in the test for Granger-causality. prices often depend on the sample length, the number of explanatory variables used (including the number of lags of each variable), and the particular measure of prices used. Two recent papers typify the disagreement. Mehra (1993) examines a system of variables that includes inflation, the output gap, and unit labor costs as a measure of wages, as well FIGURE 2 WA G E A N D P R I C E I N F L AT I O N Annualized growth rates 14 12 Wage inflation 10 8 6 4 Price inflation 2 0 –2 1959 1963 1967 1971 1975 1979 1983 1987 1991 1995 1999 3 POLICY DISCUSSION PAPERS NUMBER 1, APRIL 2000 as dummy variables for wage and price controls and a measure of energy prices, for the period 1956:IQ to 1992:IVQ in the United States. He reports that one can conclude that wages Granger-cause inflation only when one uses the CPI to measure prices, and that one obtains much stronger evidence that prices Granger-cause wages when one uses the more general GDP price deflator. These results are consistent with earlier work by Mehra (1991), Gordon (1988), and Huh and Trehan (1995). More recently, Ghali (1999) re-explores the econometric issues in Mehra (1993) and modifies the system of variables to include the relative price of imported goods, although he shortens the sample period to 1959:IQ–1989:IIIQ and only considers prices as measured by the GDP price deflator. In contrast to the findings of Mehra (1993), Ghali (1999) reports that there is strong evidence that wages Granger-cause prices and advocates that the Federal Reserve should monitor unit labor costs as a predictor of future inflation. However, the findings by Ghali (1999) are atypical and not confirmed by the results reported below. Three further studies find the systematic evidence that wages cause prices woefully insufficient. First, exploring whether sectoral wage growth causes inflation, Rissman (1995) concludes, “ ... In most of the industries examined, the direction of causality runs from prices to wages rather than wages to prices. Only in manufacturing and retail trade does proTABLE 1 GRANGER-CAUSALITY TESTSa Panel A. Wages Measured by Compensation per Hour with Productivity Included Separately Are prices caused by Time period 1960:IQ –1999:IIIQ 1980:IQ –1999:IIIQ Wages 0.330 0.295 Productivity 0.339 0.163 Money 0.296 0.270 Error correction 0.054 0.742 Are wages caused by Time period 1960:IQ –1999:IIIQ 1980:IQ –1999:IIIQ Prices 0.155 0.778 Productivity 0.622 0.862 Money 0.368 0.422 Error correction 0.447 0.317 Panel B. Wages Measured by Unit Labor Cost Are prices caused by Time period 1960:IQ–1999:IIIQ 1980:IQ–1999:IIIQ Unit labor costs 0.146 0.451 Money 0.215 0.687 Error correction 0.233 0.264 Are unit labor costs caused by Time period 1960:IQ–1999:IIIQ 1980:IQ–1999:IIIQ Price 0.000 0.000 Money 0.000 0.102 Error correction 0.000 0.150 Panel C. Wages Measured by Average Hourly Earnings with Productivity Included Separately Are prices caused by Time period 1960:IQ –1999:IIIQ 1980:IQ –1999:IIIQ Wages 0.222 0.269 Productivity 0.330 0.217 Money 0.296 0.492 Error correction 0.012 0.185 Are wages caused by? Time period 1960:IQ –1999:IIIQ 1980:IQ –1999:IIIQ Prices 0.631 0.250 Productivity 0.403 0.143 Money 0.361 0.245 Error correction 0.000 0.002 a. Each column reports the p-values (that is, the level of statistical significance) for the test that the column variable does not Granger-cause either wage inflation or price inflation. 4 FEDERAL RESERVE BANK OF CLEVELAND ductivity-adjusted wage growth appear to help forecast inflation.’’ Second, Clark (1998) has recently explored whether producer prices help predict consumer prices on both an in-sam ple (Granger-causality) and an out-of-sample basis. He finds that while there is evidence that producer prices Granger-cause consumer prices, the nature of these relationships is so fragile that they simply cannot be used on an out-of-sample basis that would be useful for policy purposes. Finally, recent work by Hogan (1998) finds that even in a price-equation, Phillips-curve specification, unit labor costs are not helpful for predicting U.S. inflation.6 Updated Granger-Causality Tests Table 1 presents our own tests for Granger-causality between wages and prices, using the most recent quarterly data available for the United States. 7 Panel A of the table reports evidence on which variables cause prices and wages. The regression includes lagged inflation, 6. Alternatively, Gali and Gertler (1999) find that a measure of the marginal cost of labor is a relevant determinant of inflation. However, their findings do not consider the simultaneity or causality issues that are the focus of our work. 7. All data were current as of December 7, 1999. lagged money growth, lagged wage growth, lagged productivity, the lagged relative growth in energy and food prices, dummy variables for the Nixon wage and price controls, and a constant (see the appendix for a complete description of the data).8 The lag length for each variable is set to four. The regression also includes an error correction term that has been estimated to find the long-run relationship between the log level of wages, prices, money, and productivity.9 Each panel first reports the evidence on whether the column variable Granger-causes prices, followed by the evidence for whether the column variable Grangercauses wages. The null hypothesis is that there is no Granger-causation, and each column reports the level of statistical significance that one can reject this null hypothesis —the socalled “p-values.” In general, a p-value of less than 0.1 is evidence of Granger-causality. The results in the top panel suggest that for the full sample, 1960:IQ to 1999:IIIQ, the evidence does not show that wages Granger-cause inflation, but rather that prices are driven by their long-run relationship with other variables in the model, as identified by the error correction term. Nor do price inflation and productivity growth Granger-cause wage inflation. Price and wage changes are best predicted by their own lags in these estimates.1 0 These results should be viewed as reasonably consistent with those in the literature in that little or no Granger-causality has been found for wages. The inability of these variables to forecast one another “in-sample” continues for the more recent period, 1980:IQ to 1999:IIIQ. Panel B of table 1 reanalyzes whether wages cause prices, except wages are now measured by unit labor costs. The findings suggest that for both samples, prices Granger-cause unit labor costs, while unit labor costs do not Granger-cause prices. In addition, money and the long-run error correction term Granger-cause prices in the full sample. These findings further reinforce the view that wages (even corrected for productivity growth) have 8. These Granger-causality tests do not include a measure of the output gap, as recent evidence suggests that such measures are highly unreliable. Orphanides and van Norden (1999) demonstrate that in some cases the revision in the output gap is as large as the total variability of the output gap. We also do not report results for similar tests, including an import price index, as recommended by Ghali (1999), because including the index restricted the period of analysis but did not materially alter the results from those reported for 1980:IQ to 1999:IIIQ. 9. As pointed out by Ghali (1999), these cointegrating regressions do not include a constant. They were estimated using the dynamic OLS method of Stock and Watson (1993). 10. The weak performance of the money measure might be associated in part with sharply larger rates of money growth in 1999, which were associated with Y2K liquidity concerns. We examined the same relationships excluding the last year and found essentially equivalent results. no independent causal effect on prices. As is found often in the literature, there is more evidence that prices Granger-cause wages than that wages Granger-cause prices. Panel C replaces compensation and the related unit-labor-cost series with a more conventionally measured wage. Compensation includes the value of benefits, bonuses, realized gains in option-based compensation, and the earnings of the self-employed. While it 5 POLICY DISCUSSION PAPERS NUMBER 1, APRIL 2000 is important not to ignore other forms of compensation, these measures necessarily place more weight on the occupations that may be removed from the production process. Perhaps there is a stronger causal relationship between conventional wages and prices than between compensation and wages as reported in panels A and B? The Granger-causality results do not support this view. Only the long-run error correction terms are identified as significant in panel C, when conventional wage data are used. Lengthening the Forecast Horizon Granger-causality tests are based on the one-period-ahead (in this case, one quarter) predictive capacity of a data series. Policymakers typically look further forward when setting monetary policy. While a one-quarter-out estimate of inflation could be used to generate a “dynamic” forecast by plugging in predictions of the explanatory variables to yield a longerhorizon forecast, this process adds no additional information. Instead, we use statistical tests analogous to those yielding the table 1 results to test the information content of wage changes and inflation at a longer horizon. Table 2 repeats the table 1 tests at the more policy-relevant lead time of one year. TABLE 2 ONE-YEAR FORWARD PREDICTION TESTSa Panel A. Wages Measured by Compensation per Hour with Productivity Included Separately Are prices predicted by Time period 1960:IQ – 1999:IIIQ 1980:IQ – 1999:IIIQ Wages 0.298 0.029 Productivity 0.604 0.010 Money 0.006 0.000 Error correction 0.022 0.081 Are wages predicted by Time period 1960:IQ – 1999:IIIQ 1980:IQ – 1999:IIIQ Prices 0.879 0.076 Productivity 0.340 0.557 Money 0.071 0.149 Error correction 0.257 0.926 Panel B. Wages Measured by Unit Labor Cost Are prices predicted by Time period 1960:IQ – 1999:IIIQ 1980:IQ – 1999:IIIQ Unit labor costs 0.176 0.009 Money 0.084 0.001 Error correction 0.814 0.002 Are unit labor costs predicted by Time period 1960:IQ – 1999:IIIQ 1980:IQ – 1999:IIIQ Price 0.000 0.208 Money 0.460 0.042 Error correction 0.328 0.130 Panel C. Wages Measured by Average Hourly Earnings with Productivity Included Separately Are prices predicted by Time period 1960:IQ – 1999:IIIQ 1980:IQ – 1999:IIIQ Wages 0.822 0.174 Productivity 0.469 0.715 Money 0.008 0.231 Error correction 0.002 0.046 Are wages predicted by Time period 1960:IQ – 1999:IIIQ 1980:IQ – 1999:IIIQ Prices 0.611 0.189 Productivity 0.241 0.165 Money 0.054 0.706 Error correction 0.007 0.090 a. Each column reports the p-values (that is, the level of statistical significance) for the test that lags 5 to 8 of the column variables do not aid in predicting either wage inflation or price inflation. 6 FEDERAL RESERVE BANK OF CLEVELAND Throughout table 2, more variables offer a boost to the model’s predictive power at the longer horizon. In panel A, wage changes still do not predict inflation in the full sample. However, prices and wages help predict each other in the second half of the sample. Focusing on the later period, all variables contribute to better estimates of inflation: wages, productivity, money, and the error correction term. Interestingly, money performs far better at one year into the future. Except for predicting wages since 1980, money is consistently significant. Panel B shows similar results when unit labor costs are used as the wage measure. Only in the later period are unit labor costs a significant factor in inflation estimates, but money and the error correction term are frequently significant. In panel C, wages never matter and money works only over the full time span. Overall, there is some support for looking at wages as an indicator of impending inflation, although there is just as much evidence that wages are caused by inflation movements. Money is surprisingly important in this analysis; in fact, it is the most reliable indicator for both wage and price changes. Judging the Predictive Performance A shortcoming of Granger-causality tests is that they are based on in-sample estimates of the data. However, financial markets attempt to use this information to predict future inflation, so it is also important to analyze whether empirical models of inflation estimated over available data are useful for forecasting inflation for a later period (such as the 1990s). Financial forecasts are likely to use all of the data available at a given date to predict inflation a num ber of periods out. Results from such an experiment are presented in table 3a. A number of inflation equations were forecasted over 1990:IQ to 1999:IIIQ using data from 1960:IQ to the assumed forecasting date, which depends on how far out the forecasting exercise is being conducted.1 1 The models are labeled in the first column. The univariate model contains a time trend, the lagged price level (in logs), four lags of inflation, four lags of relative growth in energy and food prices, dummy variables for the Nixon wage and price controls, and a constant (see the appendix for a complete description of the data). The model labeled money also includes the lagged level of money (in logs) and four lags of money growth. The wage, productivity, and unit-labor-cost models are similarly defined.1 2 The second column in table 3a reports the root mean squared errors (RMSE) of one- 11. These static forecasts examine how an equation, estimated over one time period, fits over a later sample with quarterly reestimation of the empirical relationships. Forecasts beyond a quarter again are based on a static forecasting procedure. A dynamic forecast would involve estimating an inflation equation and then forecasting each subsequent period based on earlier forecasts. Static forecasts typically do a much better job of predicting than do dynamic forecasts. See also Clark (1997). 12. In contrast to the in-sample Granger-causality tests, where model specification is essential for making correct statistical inference for hypothesis tests, we take a less structural approach for our out-ofsample comparison of RMSE. quarter-ahead forecasts. Forecasts with lower RMSE are better than those with higher RMSE. The values in parentheses are the probability that the money, wage, productivity, unitlabor-cost, and hourly-earnings models of inflation have lower RMSE than does the simple univariate model. The null hypothesis is that these models do not have lower RMSE than 13. The degrees of freedom are adjusted to account for the lack of independence in one-year-forward forecasts. the univariate model. P-values less than 0.1 generally suggest that the models outforecast the simple univariate inflation model. The third column repeats the analysis with a forecast of inflation one year in advance.1 3 Table 3b performs a similar exercise, except that we now use these key variables to predict wage growth on an out-of-sample basis. The structure of the table is the same as table 3a. Two key results are revealed in these wage growth forecasts. First, neither inflation nor 7 POLICY DISCUSSION PAPERS NUMBER 1, APRIL 2000 productivity has been very helpful in forecasting wage growth out-of-sample. In fact, they have only a marginal effect in lowering the RMSE; and in one instance, they actually raise it. Second, money growth performs well in predicting wage growth at both horizons. The out-of-sample forecasts confirm the absence of convincing evidence that wages are good at forecasting prices on an out-of-sample basis. In particular, while the RMSE for the wage model is lower than that of the univariate model at both horizons, the estimated standard error of the improvement in the forecast is so large that we cannot reject the null hypothesis. 1 4 This outcome is not unusual. The two other wage measures and productivity also offer some mean improvement in the forecast, but they are not consistent enough to be statistically significant. Productivity stands out because it offers the lowest RMSE, but the TABLE 3A OUT-OF-SAMPLE FORECASTING OF PRICE INFLATION FORECASTS OVER 1990:IQ–1999:IIIQ One Quarter Ahead Root Mean Squared Error (p-value) One Year Ahead Root Mean Squared Error (p-value) 0.821 1.240 Money 0.758 (0.311) 0.961 (0.046) Wages 0.711 (0.236) 0.893 (0.335) Productivity 0.698 (0.118) 0.808 (0.208) Unit labor cost 0.768 (0.251) 1.090 (0.106) Average hourly earnings 0.793 (0.673) 14. The comparison of forecast errors is robust to heteroskedasticity and serial correlation of unknown form using a standard Newey-West correction. This approach is similar to that in Blomberg and Hess (1996). See Diebold and Mariano (1995) for general comparisons of forecast accuracy tests. 1.143 (0.677) Model Univariate The univariate model contains four lags of inflation, four lags of the relative growth in energy and food prices, dummy variables for the Nixon wage and price controls, the long-run error correction term, and a constant (see the appendix for a complete description of the data). The model labeled money also includes the lagged level of money (in logs) and four lags of money growth. The wage, productivity, and unit-labor-cost models are similarly defined. The p-value reports the level of statistical significance for the test of the null hypothesis that these models do not have lower RMSE than the univariate model. TABLE 3B OUT-OF-SAMPLE FORECASTING OF WAGE GROWTH FORECASTS OVER 1990:IQ–1999:IIIQ One Quarter Ahead Root Mean Squared Error (p-value) One Year Ahead Root Mean Squared Error (p-value) 1.780 1.904 Money 1.528 (0.126) 1.602 (0.046) Price inflation 1.747 (0.850) 1.765 (0.497) Productivity 1.719 (0.717) 1.940 (NA) Model Univariate The univariate model contains four lags of wage growth, four lags of the relative growth in energy and food prices, dummy variables for the Nixon wage and price controls, the long-run error correction term, and a constant (see the appendix for a complete description of the data). The model labeled money also includes the lagged level of money (in logs) and four lags of money growth. The inflation and productivity models are similarly defined. The p-value reports the level of statistical significance for the test of the null hypothesis that these models do not have lower RMSE than the univariate model. 8 FEDERAL RESERVE BANK OF CLEVELAND forecast improvement is unfortunately quite variable. Importantly, however, money does contain essential information for forecasting prices over a one-year horizon during this time period. Money also offered reliable improvements in the forecast in the 1990s, suggesting that those who wish to gauge where future inflation is headed should consider moneysupply developments. Conclusion There is little systematic evidence that wages (either conventionally measured by compensation or adjusted through productivity and converted to unit labor costs) are helpful for predicting inflation. In fact, there is more evidence that inflation helps predict wages. The current emphasis on using changes in wage rates to forecast short-term inflation pressure would therefore appear to be unwarranted. The policy conclusion to be drawn is that inflation can appear regardless of recent wage trends. 9 FEDERAL RESERVE BANK OF CLEVELAND Data Appendix Average hourly earnings: Average hourly earnings of production workers. The data are seasonally adjusted, monthly, available from January 1964 to September 1999. Quarterly average used in analysis. Money: The adjusted St. Louis Monetary Base. The data are available from 1950:IIQ to 1999:IIIQ. Obtained from FRED, Federal Reserve Bank of St. Louis. Prices: The personal consumption expenditure deflator. The data are seasonally adjusted, quarterly, available from 1947:IQ to 1999:IIIQ, 1992 = 100. Productivity: Nonfarm business. The data are seasonally adjusted, quarterly, available from 1947:IQ to 1999:IIIQ, 1992 = 100. Relative inflation in food and energy prices: CPI inflation less the growth rate of the CPI excluding food and energy. Unit labor costs: Nonfarm business, unit labor costs. The data are seasonally adjusted, quarterly, available from 1947:IQ to 1999:IIIQ, 1992 = 100. Wage and price control dummy variables: The first dummy variable takes the value 1 durring 1971:IIIQ to 1972:IVQ and 0 otherwise. The second dummy variable takes the value 1 during 1973:IQ to 1974:IVQ. Wages: Nonfarm business, total compensation. The data are seasonally adjusted, quarterly, available from 1947:IQ to 1999:IIIQ, 1992 = 100. The data were transformed by taking logs, and quarterly growth rates are annualized. 11 FEDERAL RESERVE BANK OF CLEVELAND References Blomberg, S. Brock, and Gregory D. Hess. “Politics and Exchange Rate Forecasts,” Journal of International Economics, vol. 43, no. 1/2 (August 1997), pp. 189-205. Clark, Todd E. “Do Producer Prices Help Predict Consumer Prices?” Federal Reserve Bank of Kansas City, Research Paper no. 97-09, December 1997. Diebold, Francis, and Roberto Mariano. “Comparing Predictive Accuracy,” Journal of Business and Economic Statistics, vol. 13, no. 3 ( July 1995), pp. 253 – 63. Gali, Jordi, and Mark Gertler. “Inflation Dynamics: A Structural Econometric Analysis,” Journal of Monetary Economics, vol. 44, no. 2 (October 1999), pp. 195–222. Ghali, Khalifa. “Wage Growth and the Inflation Process: A Multivariate Cointegration Analysis,” Journal of Money, Credit, and Banking, pt. 1, vol. 31, no. 3 (August 1999), pp. 417– 31. Gordon, Robert J. “The Role of Wages in the Inflation Process,” American Economic Review Papers and Proceedings, vol. 78, no. 2 (May 1988), pp. 276–83. Hogan, Vincent. “Explaining the Recent Behavior of Inflation and Unemployment in the United States,” International Monetary Fund Working Paper, no. 98/145, 1998. Huh, Chan G., and Bharat Trehan. “Modeling the Time-Series Behavior of the Aggregate Wage Rate,” Federal Reserve Bank of San Francisco, Economic Review, no. 1 (1995), pp. 3–13. Layard, Richard, Stephen Nickell, and Richard Jackman. The Unemployment Crisis. Oxford and New York: Oxford University Press, 1994. Mehra, Yash P. “Unit Labor Costs and the Price Level,” Federal Reserve Bank of Richmond, Economic Review, vol. 79, no. 4 (Fall 1993), pp. 35–52. , “Wage Growth and the Inflation Process,” American Economic Review, vol. 81, no. 4 (September 1991), pp. 931– 37. Orphanides, Athanasios, and Simon van Norden. “The Reliability of Output Gap Estimates in Real Time,” Finance and Economic Discussion Series, Board of Governors of the Federal Reserve System, Research Paper no. 99-38, August 1999. 13 POLICY DISCUSSION PAPERS NUMBER 1, APRIL 2000 Rissman, Ellen R. “Sectoral Wage Growth and Inflation,” Federal Reserve Bank of Chicago, Economic Review, July/August 1995, pp. 16–28. Staiger, Douglas, James H. Stock, and Mark W. Watson. “The NAIRU, Unemployment, and Monetary Policy,” Journal of Economic Perspectives, vol. 11, no. 1 (Winter 1997), pp. 33–50. Stock, James H., and Mark W. Watson. “Forecasting Inflation,” Journal of Monetary Economics, vol. 44, no. 2 (October 1999), pp. 293– 335. , and . “A Simple Estimator of Cointegrating Vectors in Higher Order Integrated Systems,” Econometrica, vol. 61, no. 4 ( July 1993), pp. 783 – 820. 14 pd papers FEDERAL RESERVE BANK OF CLEVELAND RESEARCH DEPARTMENT P.O. BOX 6387 BULK RATE U.S. Postage Paid Cleveland, OH Permit No. 385 C L E V E L A N D , O H I O 4 410 1 Return Service Requested: Please send corrected mailing label to the Federal Reserve Bank of Cleveland, Research Department, P.O. Box 6387, Cleveland, Ohio 44101.