INTRODUCTION

Most of the information that we have about income distribution is cross-sectional in nature; there are statistics about, for example, income levels, poverty rates, and the extent of inequality for a given year or for a series of years.

The period of time over which income mobility is assessed is a fundamental issue, and different choices have led to two relatively distinct literatures. On the one hand, there is the subject of how an individual’s income changes between one year and another during their lifetime; on the other hand, there is the subject of income change between gener­ations of parents and children. We use this distinction between intragenerational and intergenerational income mobility as an organizational device in this chapter, reflecting the division in existing literature, but we shall also attempt to draw out the features of the measurement of income mobility that are common to both topics while also highlighting dimensions of them for which different approaches to analysis are appropriate.

Conceptual issues are addressed first because clarification of them is an essential pre­liminary to any discussion of measurement principles, data sources, and assessment of empirical evidence.

We review the measurement of income mobility in Section 10.3, focusing on the generic case in which there are data on income at two points in time, whether this be two years (as in the intragenerational mobility literature) or two generations (as in the intergenerational mobility literature). This is the most commonly examined situation. Thus we are interested in not only summarizing a single bivariate joint distribution of income but also comparing such distributions across time or countries to say whether mobility is greater or smaller. We explain various descriptive methods for situations in which income data are either continuous or grouped into categories. First we discuss graphical devices and methods that may be used to undertake mobility comparisons with­out resort to choice of a particular mobility index (so-called dominance checks). Second, we consider scalar indices of mobility ranging from regression coefficients and correla­tions through to other more specialized developments.

By considering measurement from a generic point of view, we aim to show how there might be greater cross-fertilization between the intra- and intergenerational mobil­ity literatures in approaches to measurement. At the same time, we highlight how the different measurement approaches relate to different concepts of mobility identified in Section 10.2.

Evidence about income mobility is the subject of the next two sections: Section 10.4 considers intragenerational mobility; Section 10.5 considers intergenerational mobility.

The final section, 10.6, provides brief concluding remarks and makes some proposals concerning where the returns to future research efforts are the greatest.

Earlier research on income mobility has typically focused on either within- or between-generation topics. For surveys of intragenerational measurement issues, we build on Jenkins (2011a) who, in turn, draws heavily on other surveys such as by, e.g., Atkinson et al. (1992), Burkhauser and Couch (2009), Fields and Ok (1999a), Jenkins and van Kerm (2009), and Maasoumi (1998). For intergenerational mobility, important earlier reviews are provided by Solon (1999), Bjorklund and Jantti (2009), Black and Devereux (2011), and Piketty (2000). Many of the reviews just cited appear in volumes with “Handbook” in their title. Indeed extensive surveys of cross-sectional approaches to income distribution were provided throughout the Handbook of Income Distribution, Volume 1 (Atkinson and Bourguignon, 2000). It is timely and appropriate to give income mobility similar attention.

Although the chapter draws heavily on the work of others, it also has some distinctive features besides simply being more up-to-date. One aspect is our goal to try to integrate the discussion of intra- and intergenerational mobility insofar as this is possible, while also highlighting what aspects of each topic are intrinsically different and deserving of separate attention. Other aspects include our coverage from conceptual issues through to data, issues of empirical implementation, and evidence.

The emphasis of this chapter is on the measurement of income mobility, broadly defined. Of course, it is also of interest to not only describe how individuals’ incomes change from one time period to another but also to explain the patterns observed.

There is some discussion of intragenerational models of earnings dynamics, nonethe­less, in Section 10.3 because estimates from “variance components” models have been used to derive measures of mobility in the form of income risk. Other types of modeling approaches are reviewed by Jenkins (2000), who also discusses more general issues con­cerning the modeling of intragenerational income dynamics. These are further elaborated by Jenkins (2011a, chapter 12).

One important distinction is between reduced-form and structural empirical models, each of which has different strengths and weaknesses. The former are empirically grounded rather than derived from a well-developed theoretical model that implies spec­ifications, the parameters of which are estimated from the data. The advantage of a struc­tural approach is that there is a close relationship between parameter estimates and behavioral model parameters, and so interpretation is improved and one may be able to say more about underlying causes. The problem with a structural approach is that clear-cut implications for model specification and proofs of relationships can often only be derived by massive simplification—simplification that compromises claims that the model describes empirical reality. The tension between reduced form and structural approaches has existed for a long time and is likely to remain. The reason for the tension is obvious—approaches combining structure, practicality, and feasibility are very difficult to develop. The problem is that a model is needed not only for the dynamics of labor earnings for an individual but also the earnings and possibly other income sources of other individuals in a multiperson household, and the dynamics of household structure itself also needs to be modeled (Jenkins, 2011a, pp. 368-369).

Exactly the same tension has arisen in empirical modeling of intergenerational income dynamics, where there is also a need to consider not only multiple income sources but also demographic factors. The structural (“optimizing”) approach is epitomized by Becker and Tomes (1986) and the reduced-form (“mechanical”) approach by a series of papers by Conlisk (1974, 1977, 1984).^[573] The relative merits of the two approaches are lucidly discussed by Goldberger (1989), with a “reply to a skeptic” provided by Becker (1989).

10.2.