"Facts Carefully Marshalled" in the Empirical Studies of William Stanley Jevons

William Stanley Jevons earned his well-deserved status as a “pioneer of modern economics” (Black 1981, 1) because of his role as a measurer and because of his contributions to economic theory. In both areas he struggled against the weight of John Stuart Mill's authority (Peart 1995, 1996a). This essay commences with the contrast between the mid-nineteenth-century method of Mill, who resisted using measurement techniques in economics, and that of Jevons, who endorsed measurement both in principle and by example. Mill allotted a rather inconsequential role to measurement in social science, one that precluded reliance on statistical data to establish causal relations. Jevons, on the other hand, generally regarded phenomena as governed by one (and, less often, several) “constant” cause, interfered with by “disturbances” ([1874] 1909, 339), and he urged the researcher to negate the influence of these disturbing causes by using techniques such as the method of means. In his own measurement projects, he maintained that omitted causes might be treated as “balancing” in the drawing of a mean (340), so that, for instance, the general variation of prices might be attributed to a single cause, the gold influx ([1865] 1884, 120).

Early proponents of statistical analysis, as opposed to those who used statistics for mere description, faced a major problem: they did not possess statistical measures of causal association (namely, measures of correlation). Nor did they have formal models to show how multiple causes might link in complex ways. Instead they relied on the analysis of tables and the taking of various kinds of averages to identify genuine causal relationships in their nonexperimental data (McKim 1997, 4; Morgan 1997, 74; 1990, 4; Klein 1997, chaps. 4 and 5). Although Jevons was well aware of the complexity of social and economic phenomena, he downplayed this complexity when he measured economic phenomena and presumed the existence of simple causal frameworks. There was also, as we shall see, continuing resistance to the sort of formal modeling that would be adequate for complex causal relationships. In order to situate Jevons's measurements in the context of contemporary work, we shall consider contemporary reactions to his measurements.

It is helpful in what follows to consider Jevons's empirical work along a spectrum: At one end we find those studies for which the causal framework was accepted before Jevons set out to measure the relationship. At the other are those projects for which he offered a less well accepted or new causal mechanism and which he tried to support with measurements. In cases where Jevons relied on established economic (or geological, or physical) relationships, he set out to confirm or measure the strength of the relationship and was able to apply his method of “wide averages” with relative success. In the case of proposed causal mechanisms less familiar to his contemporaries, we find him struggling to “marshall” the facts in support of his explanation.¹

Mill and Jevons on the Role of Measurement

Since Jevons's Principles of Science (1874) was largely directed at Mill's method as set out in Mill's Logic (1843), it seems appropriate to commence with some brief remarks on Mill. In the Logic as well as in the 1848 and subsequent editions of the Principles of Political Economy, Mill was reluctant to endorse the possibility of deriving empirical laws in social science ([1843] 1973, 908–9; [1848] 1965, 142–52; Peart 1995). Instead, Mill urged the scientist to investigate the existence and influence of all causal factors in the study of specific outcomes and to evaluate the merits of the theoretical framework in terms of its ability to explain that outcome (Hollander and Peart 1999). Measurement was not a means of inferring either the existence of causal relationships or the size of effects attributable to specific causes (Peart 1993, 1995).

In his Principles of Science, Jevons ([1874...