Percent Black and Lynchings Revisited

Abstract

Interestingly, the two sentences quoted from me by Tolnay et al. were contained in a mere footnote in connection with a power threat argument that relied much more heavily on several other sorts of data, namely the Dixiecrat vote in the 1948 national elections (assumed to be primarily a white vote) and the percentage of blacks who were not registered to vote in the 1950 election (as an indicator of white mobilization). In both of these instances, the choice of denominator was relatively straightforward. Not so with respect to lynch rates, as the present debate makes abundantly clear. Indeed, it was partly this ambiguity in one's choice of denominator that led me to confine my remarks to that single footnote. In addition, however, there is the question of the proper choice of unit of analysis. Should it be the state, the county, a community, or some other unit? If the unit becomes too small, one encounters instability in lynch rates. But if it is too large the symbolic nature of the act may be unclear to all concerned, though this will depend on the extent and nature of the publicity surrounding the lynching. We encounter here, as well, a conceptual and methodological problem in the use of so-called variables. At base, we have a ratio (lynchings divided by some population-related number) being used as an indicator of a propensity or risk factor. Tolnay et al. choose to formulate the problem so as to eliminate the denominator problem, but this does not really resolve the issue since an appropriate equation then needs to be formulated to explain the number of lynchings L. In this connection, I fail to understand the logic behind their use of an equation in which the population size N merely appears in a single additive term on the righthand side. The proper use of a ratio-variable model requires reasoning and assumptions based on a micro theory in those instances where the underlying rationale refers to individual motivations and propensities, here involving discriminatory behaviors based on a perceived threat. If our theory is formulatd at the micro level, but allows for contextual influences (here the minority percentage), then in the ideal we might prefer to have a direct measure of such a propensity or risk, rather than a relative