The Covariance Structure of Earnings in Great Britain : 1991 - 1995 Part 1 Xavier

In this paper I decompose the monthly earnings of male full-time employees into their permanent and transitory components, using the British Household Panel Study for the period 1991-95. Cross-section evidence points at a positive message: the increasing earnings inequality trend of the 1980s is interrupted. Notwithstanding this, I find that a substantial proportion of individual earnings differences at a point in time have a permanent character. Changes in total earnings variance, however, are mostly accounted for by changes in the transitory component, which comprise the effects of random shocks which persist longer than one period but that depreciate very quickly. Unlike recent studies on the covariance structure of earnings, I also consider the effects of observed characteristics on the covariance structure of log earnings. Human capital related observables account for some of the permanent component of log earnings, whereas more job related variables also impinge on transitory differences. Finally, the implications of such a structure, which contemplates individual heterogeneity and serial correlation, are illustrated for the case of low earnings probabilities by cohorts. If the low earnings threshold is set at half the median earnings, single period probabilities are higher for the youngest cohort (those born after 1960) and oldest cohort (those born before 1940); joint probabilities for a pair of consecutive periods are fifteen times larger than the product of single year probabilities for those born between 1941-60, and three and six times larger for the youngest and oldest cohorts respectively. The probability of low earnings in a given period for an individual born between 1941-60 chosen at random is fifteen times higher if he was in the low earnings state in the previous period than if he was not. This paper is part of my PhD thesis at the Institute for Social and Economic Research, University of Essex. I would like to thank Stephen Jenkins for his many and valuable comments and discussions. My thanks also to Frank Cowell, John Ermisch and Richard Dickens. The usual disclaimer applies. Correspondence to: Universitat Autònoma de Barcelona, Dept. Economia Aplicada, Edifici B, 08193 Bellaterra. Spain. E-mail: [email protected] Nontechnical Summary Individual (log)earnings can always be decomposed into a permanent component which represents the fraction of earnings that, on ‘average’ over time, a given individual enjoys over his life cycle and a transitory component measuring deviations of the actual earnings from this ‘average’ at each period. Permanent and transitory earnings differences have different economic implications. Permanent differences imply little or no earnings mobility and possibly a rigid labour market in where the starting position is crucial for determining individual’s life-cycle earnings profile and their position in the earnings distribution over time. Alternatively if cross-section inequality is largely due to short run transitory differences, over time, the burden of inequality can be seen as being shared more equally among individuals. From a policy perspective, the degree of persistence in individual earnings differences tells us to what extent low earnings is a lasting or one-off experience. This paper estimates error component models using the British Household Panel Survey (BHPS) to assess the relative importance of permanent and transitory components in male earnings over the first half of the nineties in Great Britain. The preferred error component model is then used to predict the probability of observing a given individual below a low earnings threshold. Low earnings dynamics are also considered and the effects of heterogeneity and serial correlation are highlighted. My findings suggest that earnings differences across individuals are largely due to transitory shocks to individual earnings that tend to persist for a short period of time (two to three years). Notwithstanding this, the permanent differences account for some one fourth of overall earnings variation. Unlike recent studies on the covariance structure of earnings, I also consider the effects of observed characteristics on the decomposition of log earnings. Human capital related observables, such as education, account for some of the permanent component of log earnings, whereas more job related variables also impinge on transitory differences. Finally, the implications of such a structure, which contemplates individual heterogeneity and serial correlation, are illustrated for the case of low earnings probabilities by cohorts. If the low earnings threshold is set at half the median earnings, single period probabilities are higher for the youngest cohort (those born after 1960) and oldest cohort (those born before 1940); joint probabilities for a pair of consecutive periods are fifteen times larger than the product of single year probabilities for those born between 1941-60, and three and six times larger for the youngest and oldest cohorts respectively. The probability of low earnings in a given period for an individual born between 1941-60 chosen at random is fifteen times higher if he was in the low earnings state in the previous period than if he was not.