Internalization of externalities and local government consolidation: empirical evidence from Japan

Bradford and Oates (1974) provide a well-structured survey.

See, for example, Anselin (1988) or Lesage and Pace (2009). On the application of the spatial econometric method, see Brueckner (2003).

Total public spending in Japan was 24% of its gross domestic product (GDP) in 2010, while municipal spending was 7% of GDP.

Akizuki (2001) described the relationship between the central and local governments as “controlled decentralization.”

See, for example, Council of Local Authorities for International Relations (CLAIR 2009), MIC (2006, p. 41), and Yokomichi (2007).

One of the objectives of the “Great Heisei Merger” was to deal with residents’ demands for the provision of public services beyond traditional boundaries.

They included a grace period for local governments to avoid unconditional grant reduction resulting from a merger and a subsidy for the principal and interest of local bonds issued by municipalities to finance additional costs stemming from a consolidation.

As this approach indirectly measures publicness and the obtained results are problematic in terms of jurisdiction size, alternative approaches have been developed to rigorously estimate price elasticity and congestion (Reiter and Weichenrieder 1997). Nevertheless, subsequent studies mostly agree with the results presented by Bergstrom and Goodman (1973) and Borcherding and Deacon (1972) and show that the approach remains adequate for estimating the demand for local public services.

By minimizing a cost function subject to the Cobb–Douglas production function \({G_i} = aL_i^\beta K_i^{1 - \beta }\) , where denotes labor and denotes capital, and briefly rearranging, Eq. (5) is obtained. The Cobb–Douglas assumption is standard in this literature. See, for example, Bergstrom and Goodman (1973), Borcherding and Deacon (1972), and Reiter and Weichenrieder (1997).

“Spatial lag,” introduced by Anselin (1988), is a weighted average of random variables (dependent variables, in many cases) at neighboring locations.

Another econometric approach that provides consistent estimates is the instrumental variable method developed by Kelejian and Prucha (1998) and Kelejian and Robinson (1993). See, for example, Brueckner (2003).

Many previous studies of mergers show that economies of scale frequently prevail in public schooling and that efficiency gains are a typical driving force behind local government mergers (e.g., Austin 1999; Gordon and Knight 2009).

See, for example, CLAIR (2009).

Several empirical studies have addressed behavior akin to economies of scale with regard to population, called “economies of size” (e.g., Andrews et al. 2002; Duncombe and Yinger 1993), based on the fact that the larger the population size of a municipality, the more it could reduce its costs for public service production, for example, by efficiently utilizing its own public facilities. Among them, some studies have drawn attention to this type of economies of scale attained through jurisdictional integration (e.g., Brasington 1999, 2003b; Brink 2004; Gordon and Knight 2009). In these studies, economies of scale in population are defined by a decrease in per capita expenditure after consolidation. For our theoretical model, this aspect can be represented as \(\partial {e_i}/\partial {N_i} = \partial \left( {{c_i}{G_i}/Ni} \right) /\partial {N_i} < 0\).

Our empirical model basically follows the methodology proposed by Gordon and Knight (2009, p. 761), which assumes that average expenditure is simply composed of scale economy and diseconomy factors in terms of population and defines efficiency gains as the difference between pre- and post-consolidation expenditures per capita. Note, however, that our average expenditure specification is slightly different from theirs, to reflect the fact that the log of per capita expenditure is almost U-shaped in the log of population for Japan. However, there is some similarity too, from the viewpoint of allowing the possibility of scale economies and diseconomies. In this setting, including logs of population and squared population as controls in the second step of the regression makes it impossible to identify the coefficient of scale economy effects; for this specification, the coefficient of efficiency gain in the second step cannot be determined uniquely using the estimates of the differences in the logs of population and in the logs of squared population. Accordingly, we approximately estimate the size of efficiency gains due to consolidation by utilizing the fitted values of average expenditure.

Although racial composition is an important feature that may affect preference for local public goods (Brasington 2003b), it is not taken into account here because Japan is composed mostly of people of the same ethnic group.

See Miyazaki (2014) for theoretical background for using these covariates.

“Generated regressor” is defined as an imputed value from an auxiliary statistical model. Wooldridge (2002), for example, referred to estimations using a generated regressor.

There are primarily two approaches to correct the biases. The first is the two-step econometric model, in which unobserved variables are replaced by their estimates or predictions from an auxiliary model and standard errors in the second step are calculated from the correct asymptotic covariance matrix for the two-step procedure. The other is the full information maximum likelihood (FIML) model, in which auxiliary and main models are jointly estimated, yielding efficient estimates and asymptotically non-biased estimates of standard errors. Although efficient, the FIML model suffers disadvantages of complex and costly computation, and more specifically, a specific joint distribution of the disturbances of the auxiliary and main models. Hence, this study utilizes the two-step procedure to estimate the consolidation-decision model with the fitted values of the ATS.

The median ratio of taxable income is used as a proxy for the median tax ratio.

See, for example, Reiter and Weichenrieder (1997).

The government census is collected at 5-year intervals.

Financial support for mergers was subtly reviewed almost annually from 1999 to 2006 in order to encourage municipal mergers.

Measurements and units of the political proximity variables are provided in Panel B of Table 1.