Academic journal article American Economist

Wage Growth and Inflation in the United States: Further Evidence from Johansen's Cointegration Approach

Academic journal article American Economist

Wage Growth and Inflation in the United States: Further Evidence from Johansen's Cointegration Approach

Article excerpt

1. Introduction

There has been considerable disagreement among economists about the inflationary impact of increased wage rates. A standard hypothesis is that autonomous wage changes are a significant component of inflation determination. This is basically the wage-push hypothesis of the Phillips curve. As wages move in one direction, it is expected that the price level would move in the same direction in the long-run.

Several empirical studies, based on the implicit assumption that their time series data are stationary and integrated of order I(0), have examined the relationship between wage growth and inflation [see, Gordon (1988); Zeira (1989); Helpman and Leiderman (1990); Nymeon (1991); Mehra (1977); Tanner (1993)]. Some researchers have presented evidence in support of the hypothesis that wages and prices are correlated. [See, Zeira (1989); Helpman and Leiderman (1990); Nymeon (1991); Tanner (1993)]. Other authors, however, have suggested that wage growth and prices are not related [see, Gordon (1988); Mehra (1977)].

We suggest that these mixed results may be attributed to the implicit assumption of the various studies that the variable series used in their models are stationary. However, many macroeconomic time series data have been shown to be nonstationary in their levels but stationary after first differencing, thus integrated of order one, I(1). [see, Nelson and Plosser (1982)]. As a result, the use of such nonstationary series in an ordinary least squares (OLS) regression model would lead to spurious results, unless a linear combination of the series are shown to be cointegrated. The use of cointegration analysis would allow for the avoidance of such spurious regression results [see, Granger and Newbold (1974); and Engle and Granger (1987)]. Where there is evidence to support cointegration, the indication is that a long-run economic equilibrium relationship exists among the relevant variables. But, we first, examine the time-series properties of the macroeconomic data used in the model to ascertain whether the variables series in our model are integrated of same order. This is because cointegration requires that all variable series in a model be integrated of the same order [see, Engle and Granger (1987); Dickey and Fuller (1979) and (1981); Murthy, Ukpolo and Mbaku (1994)].

The two-step cointegration technique developed by Engle and Granger (1987) was adopted by Mehra (1991) to examine the relationship between wages and prices, using quarterly data of the United States for the period 1959:1 to 1989:3. Evidence was presented to support long-run movements between changes in the growth rates of wages and prices. However, it has been shown that this two-step cointegration technique is weak in testing and estimating cointegrating relationship where there are more than one explanatory variable in the model. Instead, the use of Johansen's maximum likelihood cointegrating technique is recommended because it is more powerful in estimating parameters of a model where the possibility of multiple cointegrating vectors exist [see, Johansen (1988); Hall (1989); Johansen and Juselius (1990); Orden and Fisher (1993); and Banerjee et al., (1993)]. No study can be found that has adopted Johansen's maximum likelihood procedure in testing for the relationship between wages and prices.

This paper uses Johansen's maximum likelihood cointegration technique to examine the relationship between wage growth and inflation in the United States during the period 1975:11994:3. First, a unit root testing procedure is conducted, using the augmented Dickey-Fuller (ADF) test, to ascertain the stationarity of the relevant series. [see, Fuller (1976); Dickey and Fuller (1981)]. But, it has been shown that the number of lags chosen in an ADF test could cloud the power of a test [Gordon, (1995)]. As a result, following Gordon (1995), we test for stationarity by looking at the behavior of the ADF statistic over different lag lengths. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.