Robert Erikson: Descriptions of Inequality
Robert Erikson: Descriptions of Inequality
Abstract and Keywords
Ysander explores the implementation decisions in Erikson's paper, which broadly encompasses (1) the choice of welfare indicators, where Ysander points to the problem encountered in ‘capability’ measurement especially when the measurement is made across countries and across time; (2) the choice of distribution space, where Ysander notes that the survey‐limited‐distribution issues to questions of discrimination; and (3) the choice of measure of distributional change, where Ysander stresses that in order to understand the reason behind a change, one must trace how one position, or set of capabilities, or lack thereof, conditions the next.
Erikson's paper presents both a survey method for measuring level of living and some empirical results from using the method on Swedish households. Both are interesting, but what I personally find most intriguing is what lies in between—the problems cropping up in implementing the theoretical concepts.
I will try to exemplify this by commenting on three broad implementation decisions, as they appear in the Swedish survey. They deal respectively with the choice of indicators, the choice of distribution space, and finally the choice of measure of distributional change.
Let me start with the choice of welfare indicators. As Erikson rightly stresses, this choice—always somewhat arbitrary—will reflect the concerns and interests of the investigator. This is certainly true in this case, where the Swedish government commissioned the survey. What the government wanted was not simply to know how to keep various voter groups happy in order to retain power, but a detailed description of living conditions, both as a background to planning distributive policies and to mobilize support for these policies. Hence the choice of detailed descriptive indicators instead of more general evaluative questions.
It is interesting to note that in searching for suitable indicators, the designer, Sten Johansson, had an ideal in mind, seemingly very close to Sen's concept of ‘capability’. It is, of course, hard to be quite sure about what ‘capability’ rightly means. But it is here enough if we think of it as an opportunity set of actions, achievements, or functionings—a set which may of course forestall any question of choice by having just one member. ‘Capability’ thus simply becomes a measure of options.
What is even more interesting is the fact that in an overwhelming number of cases, the investigators failed to observe and measure anything which could, with the best will in the world, be called a capability. They had to make do with particular individual achievements, from which they could try to construct or estimate capabilities. This could, of course be due to lack of imagination and/or analytical flexibility on the part of the investigators. I rather suspect, however, that it also shows that capabilities are often rather elusive things to catch, that it is difficult to measure directly the counterfactual part which has to do with what a person might be or do—or might have been or have done. Economists may be confused here because they are used to dealing with one of the few aspects of life where the opportunity set can be reasonably well defined and measured.
If you try to look more closely at how capabilities are dealt with in the survey, you find that they usually seem to be implicitly defined in terms of some a priori behavioural model which spells out what the probability is that a certain capability or set of options will manifest itself in certain observable achievements. (p.85) A certain degree of political capability, for example, means a certain probability that you will get politically organized, make speeches, and write to the newspapers. If you then observe the achievements of a specific individual, you can try a posteriori to estimate his capability. The reliability of the results will depend on how good or generally acceptable your a priori model is. It can be a risky business. If you are fond of marching in demonstrations or pestering your MP with phone calls, is this a sign of political capability or the reverse? Depending on prevailing political conditions, engaging in these activities may be a sign of political freedom–or frustration at finding all other channels of political influence blocked. In the Swedish survey these a priori models are very simple–and never made explicit. They deal mostly with dichotomies and usually seem to assign equal probability to all relevant achievements.
There are a number of conceptual problems involved here–apart from the unavoidable multidimensionality and the arbitrary division of options. One obvious problem has to do with the interdependence of individuals' choices in different areas. Whether or not you make use of your options for political action will, for example, depend on your capabilities in areas like education and income. The kind of a priori model you need should thus be formulated in terms of joint probabilities–a rather forbidding task.
A second difficulty arises when you try to use the indicators to measure differences in capabilities between different countries or different times. What options are implied by actual behaviour will always depend on the current social and institutional setting. Disposable income, for example, is a rather empty welfare concept if there is nothing to buy. The options indicated by observed actions can shrink drastically in the face of political regulations or social sanctions. You thus need a lot of complementary information on the institutional framework and the division of rights to draw any conclusions as to fundamental capabilities from the indicators used. Even within the comparatively stable and homogeneous Swedish society, Erikson runs into difficulties in making comparisons over a thirteen‐year period. Achievements like having a secondary education or having seen a doctor mean different things in terms of job opportunities or physical capability after the big expansion in higher education and out‐patient care during the 1970s.
Another connected conceptual problem has to do with social security. A high income, for example, has quite different welfare implications when it is reasonably secure from when it is part of a social lottery and can disappear at any time. You really need to measure not just present options but contingent future options. How well are your present options ‘insured’ against risks of various kinds? Even in the Swedish context, changes in the coverage of social insurance, etc., have been considerable, and if you extend the method to cover international comparisons the risk scenario would certainly have to be taken into account as a major component of welfare.
These various problems may raise some doubts about the possibility of making the capability concept operational for sophisticated welfare measurement. They (p.86) may at least provide an argument for sticking to simple and basic capabilities.
Let me pass to the second kind of implementation decision: the choice of measure of welfare distribution.
When we discuss equality most of us probably think of welfare distribution among individuals. In the Swedish survey, however, what was studied was distribution for the three observation years among the fifty‐four socio‐economic groups, which result from combining sex, age, occupational class, and region. What this means is that it does not really deal with equality as such since it ignores all the differences between individuals which are the result of luck and choice. What it is concerned with is that part of welfare distribution which can be ascribed to social discrimination in the four given dimensions. You might of course question whether you can talk of discrimination in terms of occupation and region as being on a par with discrimination by sex and age, since the choice of occupation and region is at least in part voluntary.
More interesting and surprising is the fact that the distribution problems are here narrowed down to problems of discrimination, which seems to fit rather badly with the original political background of the study, as stressed by Erikson. It is hard to make this minimal conception of social justice compatible with the aim of equality pursued by the commissioning government.
Erikson's own explanation is that nothing more is possible, since the material is too shaky to allow anything but dichotomies into capable and non‐capable. If taken at face value, this would mean that we would for ever abstain from investigating and discussing more general problems of equality in welfare. I do not think that is necessary even in dealing with this survey material. Admittedly, we will never get nice continuous indices of individual capabilities. But there is certainly room for sorting individuals into an ordered number of classes.
Let me finally deal with the third kind of implementation decision, which has to do with the choice of measure of distributional change.
What is done in the Swedish survey is simply to compare the distribution at different times. I would like to argue that this is seldom enough. In order to evaluate the change, understand it, and plan new policies, you must go a step further and study the change for the individuals making up the distribution.
We cannot really evaluate a distributional change without knowing, for example, if the formerly underprivileged have kept their place or if they have exchanged places with the privileged. We cannot hope to understand why a change occurs if we cannot trace how one position, or set of capabilities or lack of them, conditions the next. And this is what we need to know if we are to design a distributive strategy.
Now the survey material offers an unusual chance of looking into individual change, since for each subsequent observation year some 85 per cent of the respondents belong to a panel left over from the earlier survey. It does seem a pity to pass up this chance.
Problems do arise, however, when you try to apply concepts like capability or equality to the life careers of individuals. These concepts are usually defined (p.87) within a static framework and are not unambiguous when applied to processes of change.
Every time an individual makes a choice or passes into a new state, he or she generally modifies his or her capabilities at the same time. If the change is the result of a conscious choice we can talk about this change of capabilities as investment or disinvestment. This means that every achievement is generally associated with two kinds of capability–before and after, opportunities forgone and opportunities anticipated.
Again, it is far from easy to know how to evaluate a change in the distribution of capabilities in terms of equality, when an unknown part of the change is due to voluntary acts of choice. You may avoid the problem by retreating to the study of some narrower concept–like non‐discrimination. Otherwise, you will be faced with the impossible task of separating choice from luck over a historical sequence. I doubt whether there is any solution to that problem. If you want simply to give long‐term insurance to cases without any moral hazard, it is better to abstain from insuring them at all. Or to insure them in certain basic respects and forget about the hazard.
We can learn a lot about Swedish welfare problems from the Swedish survey. At the same time it highlights the need for further work on defining and making operational the exact aim of the exercise. There is no such thing as a generally applicable measure of welfare.