The Linked Employment Cost Index: a first look and estimation methodology

Measuring Union and Nonunion Wage Growth: Puzzles in Search of Solutions

The United States Department of Labor's Bureau of Labor Statistics produces two measures of the cost of compensation, neither of which is easily understood-not even by compensation professionals. The Employment Cost Index (ECI) measures pure wage rate and compensation cost changes, while the second measure, Employer Costs for Employee Compensation (cost levels) provides information on compensation cost trends not available from the ECI and is generally a better measure of the changes affecting the welfare of workers. Differences between the cost levels and ECI indices in year-to-year changes can be explained by differences in the way the two measures are constructed-that is, the set of weights used and the way the data are linked from quarter to quarter. This article summarizes data collection for these two measures and notes the major trends reflected. It also discusses the differences between the measures and how they can be used, both separately and side-by-side.

ion from wage rigidities. There are important exceptions including work at the Federal Reserve Board by Erceg, Henderson, and Levin (2000), but the approach of the more recent Golosov-Lucas paper with perfectly flexible wages is more common. In my view, wage determination is still a source of inflation dynamics, though not in the same rigid ways as the models in the Econometrics of Price Determination Conference. While the (S, s) model makes sense for prices it does not seem accurate for the timing of most wage changes. I think that the study of wage determination needs to be put on the front research burner again. A very important task for future research is an exploration of microeconomic wage data in the BLS's Employment Cost Index, comparable to what Klenow and Kryvtsov (2005) are doing with the CPI. My guess is that the characteristics of the micro wage data will look much different This content downloaded from 207.46.13.183 on Thu, 21 Apr 2016 05:12:45 UTC All use subject to http://about.jstor.org/terms JOHN B. TAYLOR : 199 from the price data. Moreover, the micro wage data will permit one to discriminate between different staggered wage-setting models. As Levin et al. (2005) show, the shape of the distribution of wage contracts in staggered wage-setting models matters significantly for monetary policy. Econometric policy evaluation models are still too rarely used for forecasting. In my view one of the best tests of a model is its ability to forecast, but most forecasting models are still of the reduced form variety, and only loosely connected with structural models, as the Stock and Watson paper at this conference illustrates. Efforts to use the models for forecasting-this will probably require more details for the simple three equation models-would be very worthwhile. 5. IMPACT ON ACTUAL POLICY What has been the impact on policy of the development of policy evaluation models described here? One might ask if it was influential in building the consensus that led to the end of the Great Inflation. Perhaps, the finding that inflation could be reduced with less disruption than traditional models suggested helped reduce some policymakers' reluctance to take the steps to end inflation, but it seems to me that people supported Paul Volcker's disinflation efforts mainly because they were fed up with inflation. Perhaps more relevant have been the simulations of alternative monetary policy rules. Clearly, it would not have been possible to even consider such rules systematically if it were not for these models, but the impact of the policy rule research itself is more difficult to prove. Policy rule research has, of course, been useful in comparing policy in different countries and in different time periods, or to characterize good policy versus bad policy. It has enabled central bank staff to work with interest rates as an instrument and to do so with an analytical framework. It has shown that certain monetary policy principles are important, such as reacting in a preemptive fashion to forecasted increases in inflation. It has shown that targeting asset prices can lead to poor results. And it has been helpful anytime forward-looking expectations are important, as in formulating a communication policy during the recent considerable period or measured pace phases. It is hard to imagine these types of policy issues being analyzed rigorously without the models with rational expectations and sticky prices and wages. 6. FUTURE POLICY ISSUES There are many important policy issues that these types of models or their successors can be called on to analyze in the future. Understanding the reason for the recent reduction in pass-through coefficients for exchange rates and energy prices is important for monetary policy; the models that incorporate sticky prices, market power, and rational expectations, such as Golosov and Lucas (2005), can be helpful This content downloaded from 207.46.13.183 on Thu, 21 Apr 2016 05:12:45 UTC All use subject to http://about.jstor.org/terms 200 : MONEY, CREDIT AND BANKING here. Determining when appropriate deviations from policy rules should start and when they should end is another important practical topic. Finding better ways to incorporate asset prices into policy formulation is yet another. Nailing down the key reasons among the many conjectures for the recent improvements in output volatility is also important for future policy; in my Homer Jones Lecture (Taylor 1998) I argued that it was improvements in monetary policy, but many other explanations have been raised. But whether the future policy issues are on this list of policy examples or not, it is important to keep future monetary research focused on policy issues. I think this brief history of econometric policy model development clearly illustrates that economic research is most exciting and productive when it is policy driven.

Employment Cost Index

The Employment Cost Index measures the change in wage rates for a fixed set of jobs over time. The separate indices for industry and occupation groups grew at varying rates during the 1980s. However, their growth rates have been much more similar since then, which implies that the increase in the wage rates of low-wage workers relative to middlewage workers during the 1990s was due to changes in the mix of jobs. Had all workers remained in their jobs, the relative wage of low-wage to middle-wage workers would have remained constant.

Recent Trends in Compensation Practices