Appendix D
Positional Influence and the Pork Barrel:
A Multivariate Regression Model

I constructed a multivariate linear regression model to test the assumption that a legislator's positional influence correlates positively with increased public construction expenditures in his or her home prefecture. The dependent variable in the regression is the appraised annual value of total public construction started per prefecture as measured in per capita terms (as reported in a monthly sample made by the Ministry of Construction). Specifically, the analysis sought to determine whether public works spending increased in the home prefectures of the construction minister, the head of PARC's construction division, the prime minister, and retired high-level officials of the Ministry of Construction. To control for the possibility that a positive correlation might accompany the appointment of a legislator to any major post, the analysis also included individuals appointed to the supposedly "pork-less" positions of foreign minister, labor minister, and head of PARC's labor division. To account for the potential gap between the time of appointment and the delivery of distributive policy benefits, "lags" of one through five years were incorporated

Aside from these positional influence factors, three additional political variables were used. First, the number of Lower House legislators per capita—a measure used to determine whether the relatively "overrepresented" prefectures reap a proportionately larger harvest of public construction expenditures than do the "underrepresented"

prefectures. Second, the number of LDP MPs per capita—used in a similar vein; notably, the LDP tends to attract strong support from rural districts, many of which have experienced a net decrease in population since the last significant reapportionment in 1947. Third, a measure of whether public construction expenditures tended to increase during years with general elections for the Lower House—intended to test the popular assumption that the LDP tried to boost government spending on construction during election years essentially in order to "buy" voter support.

Five independent demographic and economic variables were also incorporated into the model: (1) annual estimates of population were factored in to determine how public construction spending responds to changes in the population of the prefectures; (2) the aggregate value of manufactured goods shipments—to see whether increased business activity accompanies expanded public works spending; (3) per capita income; (4) per capita agricultural income; and (5) increases in per capita payments of national taxes—to see if these were rewarded with a proportionate expansion of public construction spending.

The regression analysis focuses on the period from 1964 to 1988 and proceeds down two pathways. The first set of regressions involve data on all of Japan's forty-seven prefectures except Okinawa, which did not revert to Japanese control until 1972. Including the lag effects, the analysis involves a total of nearly 60,000 data points. The second set of regressions contains most of the relevant variables for the nine prefectures-wide districts (zenkenku )—Fukui, Yamanashi, Shiga, Nara, Tottori, Shimane, Tokushima, Kochi, and Saga—districts in which legislators' positional influence and the possible impact of positional influence on credit-claiming are clearer than in prefectures that house more than one electoral district. However, all the prefectures-wide districts are located in the rural hinterland, where LDP candidates generally cast a disproportionately long shadow, and thus these prefectures are not representatives of all prefectures.

The regression analyses suggest that the share of the variance explained by the political varibles—including the positional influence factors—is markedly less than much of the received wisdom would have it. While public works spending increases in the prefectures represented by legislators serving as construction minister beginning in

the third year following their appointment, the size of the increase is not as high as might have been anticipated. Likewise, there was no statistically significant positive correlation either for legislators appointed to the prime ministership or for those appointed to the head of PARC's construction division (except in the case of the data concerning the nine prefecture-wide districts). Similarly, the results do not sustain the widespread belief that the districts of retired Construction Ministry officials are showered with increased public works spending following their election.

These results are preliminary, and a full test of these relationships would require additional analysis. For example, it may be that legislators in positions of influence in the public works subgovernment exercise their influence not to funnel projects into their districts but instead to ensure that specific projects are awarded to particular firms, possibly firms that have contributed funds to their political war chest. Further research is also needed to account for the informal influence of powerful individuals such as Kanemaru Shin, long reputed to be "the don of all dons" in Japan's public works subgovernment.