The interregional impact of infrastructure capital.

• Author: Moomaw, Ronald L.

1. Introduction and Problem Definition

   Increasingly, economists and state and local policymakers see a broader link between declining regional competitiveness and the deterioration of highways, bridges, water and sewer systems, and other public infrastructure. Supporters of active industrialization policies point out that one way a region can regain its competitive edge is to upgrade its infrastructure. The Clinton Administration, citing the lack of infrastructure investment as a cause of sluggish national economic growth, advocates massive spending hikes to repair existing public capital and to invest in new public capital.

   The groups supporting this approach rely on recent empirical evidence showing that public infrastructure investment generally makes a significant contribution to productivity and output [2; 4; 5; 12; 16; 18]. Existing studies, however, report disparate results about the size of the effect of public capital on economic growth. Aschauer [2], using aggregate time-series data, estimated output elasticities with respect to public capital ranging from 0.39 to 0.56. Munnell [18] updated Aschauer's study and reported an elasticity of 0.33 for labor productivity with respect to public capital. The magnitude of the estimated elasticities and the associated marginal product of public capital in these two pioneer studies are generally accepted as extraordinarily large. The debate on their size is extremely lively.

   Two fundamental issues dominate the debate. First, Aaron [1] draws attention to the implausible nature of the earlier results by comparing the marginal productivities of public and private capital implied by these studies. He finds that the implied marginal productivity of public capital is five times greater than that of private capital. Second, McGuire [14] and Eisner [6] contend that the empirical evidence from the aggregate time-series data is not compelling. McGuire argues that the established statistical relationship may not imply a causal connection between infrastructure and economic performance. Eisner also questions the causal link by suggesting that one cannot expect an increase in public capital at the beginning of a year to translate into more output at the end of the year. He recommends specification of the model with a lag structure to measure the time-series relationship between output and public capital more accurately.

   To provide information additional to that obtained from national data, Munnell [18] undertook a regional analysis of the relationship between public infrastructure and economic performance. Using data for the 48 contiguous United States, Munnell found a regional-level output elasticity of public capital of 0.15 compared to her national-level estimate of 0.33. She also found the output elasticity of private capital to be twice the size of public capital.(1) A number of other studies using state-level data confirm that aggregate public capital (as well as individual components) has a positive and statistically significant effect on regional economic performance [3; 7]. Still, additional estimates derived with different data or from different specifications is sorely needed so that the results have greater credence.

   Providing more evidence on the relationship between the stock of physical infrastructure and key economic constructs at the state level is of practical-importance for several reasons. The most compelling reason is that states compete among themselves for private investment. Congested state highways, structurally deficient bridges, and overburdened water and sewer systems impose real economic costs on firms within a state and weaken the state's competitive edge. Detours and traffic congestion add to shipping costs because of additional wages paid, greater fuel consumption, and general time delays. Restrictions on connections to water and sewer systems may discourage or halt local construction activity. Second, many state and local governments face budget problems that restrict their ability to finance infrastructure investment. Because of the budget problems and because of the necessity of increasing maintenance expenditures on existing infrastructure, states must use great caution in their infrastructure investments. Given this situation, existing estimates may not be sufficient to critique past local policy decisions and evaluate plans to meet future capacity requirements of infrastructure end-users.

   The present analysis offers several variations on Munnell's study [18]. We generate state-by-state estimates of the impact of public capital on output rather than an estimate for the "average" state in the sample. We do so by estimating a translog production function with labor, private capital, and public capital as separate inputs. This avoids the bias inherent in a Cobb-Douglas specification (such as utilized by Garcia-Mila and McGuire [7]), which yields the same constant elasticity for each state. In contrast, the translog specification permits the retrieval of numerical estimates of the output elasticity of public capital for each state.

   State-by-state information is useful in several ways. For example, given that states can augment highway capacity to enhance their productive environment, specific estimates for highways can be used in calculating transportation benefit-cost ratios for highway investment. An assessment of the effectiveness of using highway expenditures to improve the competitive business climate within specific states is not possible with the "average" information gleaned from earlier studies. Consider a case where a state needs to offer greater incentives to remain competitive with neighboring states: a ranking of empirical estimates of the output elasticities of water and sewer systems can provide a tool for judging the efficacy of this type of public capital investment in that state relative to its competitors. A comparison of such output elasticities may also reveal the need for a critical review of past policy decisions and spur better evaluation of future expenditure plans. The magnitudes of the output elasticities will dictate whether it makes sense to invest in a specific type of infrastructure in a particular state.

   Herein we provide state-by-state estimates of the output elasticity of aggregate public capital stock and its three components - highways, water and sewer systems, and "other" capital. Estimates are generated for three separate years (1970, 1980, and 1986) by specifying a state-level translog production function that relates the gross state product measure of output to the inputs: infrastructure capital, private capital, and labor. We offer a regional comparative analysis of the output elasticities with respect to the various types of public capital stock over the time periods. Next, we extend this work by conducting a simple test of the "Hansen effect",(2) which proposes that the effectiveness of public capital varies with a region's level of development and its ratio of public to private capital. The next section of the paper presents the analysis: methodology, data, and empirical results. A final section summarizes our findings and focuses on their implications.

2. Analysis

   Methodology

   Prompted by the foregoing discussion, we specify two econometric models. Model I evaluates the impact of public capital stock on a state's economy by specifying a state-level translog production function that relates gross state product (Q) to private capital (K), aggregate public capital (G), and labor (L). Model II includes private capital, labor, and three measures of infrastructure capital - highway (HY), water and sewer systems (WS), and other (O). Because the translog specification does not impose constant output elasticities, our estimates permit comparative analysis of these parameters (across states) and provide detail needed for more specific policy recommendations. The differentiation of highway capital (HY), water and sewer systems (WS), and other capital (O) as separate...