Empiricism, Normativity, and the Burdens of Judgment

Here's another little point I want to make about the idea of empiricism in politics and social science. The “empirical” results in the social sciences are often not gained through direct observation, but through often quite loose measurement techniques involving all sorts of proxies and approximations. So, when I hear somebody say “Real wages have declined since 1970” or something like that, and I snort with absolute incredulity, and they point me to some BLS chart, and I demand to know how we calculate real wages, and then I'm told about the CPI, and I assert that the methodology of the CPI must be flawed because it is obvious by casual observation that real wages have increased a great deal, who's the empiricist? Were economic liberals being good empiricists when they bitched about the Boskin report?

Alex Tabarrok (like his co-blogger, also at least as good an economist as Michael Kinsley) makes this sort of measurement method contesting move this morning on the WSJ Econoblog:

Is America on a consumption binge? In 2003 the savings rate was a paltry 1.2% — the lowest rate since the Great Depression, when savings briefly went negative. But before we tighten our belts we need to know that the standard measure of personal savings (from the National Income and Product Accounts) is highly flawed.[empasis added]

Savings in the NIPA are defined as income minus consumption. But what is consumption? The NIPA defines education expenditures as consumption, but try as I might to keep my students' attention with the occasional joke, I think few would report that they are paying me for entertainment value alone. Education expenditures ought to be defined as investment financed from savings.

The NIPA also measures savings on an annual basis. But suppose that you are asked what your savings are. You probably don't add up this year's income and subtract this year's consumption, instead you add up your stock of savings; the value of all of your assets including equities, bonds, net housing value, cash and so forth. The latter measure is the right one if we want to measure provision for the future.

The flow and stock measures of savings can easily move in different directions. Indeed, a major reason that the yearly savings rate has declined is precisely because the value of assets has increased.

A declining savings rate, therefore, can easily signal positive things about people's net asset position. Similarly, we have in recent years experienced a boom in productivity. If individuals expect the boom to continue, it may be quite rational to reduce current savings.

If I insist that the CPI is screwy, or Alex insists that NIPA is screwy, it's probably because we think it fails to adequately measure observed economic phenomena, or fails to take into account what our best theory says a measurement device ought to be taking into account. Imagine that the motivation for our complaints about measurement comes from our ideological commitments. So be it. Empricism care not about our motivation, as long as we are doing our best to save the phenomena.

My broader point is that the social sciences have a normative upshot. We care about how we measure and reason about the social world because we want to make the social world better. Our normative commitments will inevitably guide the way we generate hypotheses and affect our choice of methodological tools for testing those hypotheses. But this is not counter to empiricism. These are the mechanisms through which empiricism in the social sciences operate. Chait's hackish claims of virtuous wertfreiheit empiricism betray a kind of self-satisfied naivete that actually threatens the kind of epistemic virtue necessary for empiricsm in social sciencces by encouraging a lack of reflection about the scientific producer's and consumer's necessary engagement with and commitment to norms in the social domain.

The empirical, descriptive task in the social sciences is not separable from the normative task. I believe that empirical investigation in the social sciences ought to be as neutral among values as possible. But the question of what constitutes a neutral investigative stance is not itself a question of social science, is essentially contestable, and the argument for wertfreiheit is itself a contestable normative argument. Science in general works because scientific practices and scientific communities embody certain norms and epistemic virtues. Science does not proceed through the application of a Baconian algorithm. Social science studies human action and human coordination, phenomena that cannot even be adequately described without mentioning the normative, goal-seeking nature of practical reasoning or normative features of coordination. (The idea of a market failure, for example, which Chait makes use of, is a normative notion about coordination.) And it seems neither plausible nor desirable for the social scientific community to to pretend indifference to the norm-ladeness of their domain.

Just how and to what extent the social scientist should engage with or be guided by normative conceptions in the conduct of empirical inquiry are very hard questions. Discuss at will.