Growth and defense: pooled estimates for the NATO alliance, 1951-1988.

• Author: Macnair, Elizabeth S.

1. Introduction

   In two provocative and influential contributions, Emile Benoit [4; 5] uncovered a net positive association between defense spending and economic growth for 44 less-developed countries (LDCs) during the 1950-65 period. Apparently, a larger defense burden, as measured by the share of national income devoted to defense, may have promoted growth for these countries. This finding was controversial and generated an extensive literature that either found fault with Benoit's methodology,(1) or else investigated the relationship between growth and defense with alternative methodologies. One such methodology relied on an aggregate production function to derive an equation for the sources of growth owing to technological change, labor growth, investment, productivity differences among sectors, and externalities. This supply-side approach was associated with Feder [13] and Ram [22]. When a Feder-Ram model was used, researchers typically discovered either a small positive effect or else no impact of defense on growth. Except for Alexander [1], the Feder-Ram approach has been applied to large cross-sections of LDCs, to cross-sections of LDCs and developed countries, or to individual countries over time.

   The purpose of this paper is fivefold. First, we extend the three-sector Feder-Ram model to allow for the influence of defense spillin externalities as the defense expenditures of a nation's allies affect the nation's own defense and civilian sectors in either a positive or negative fashion. Spillin externalities can stem from input complementarity or substitutability, procurement practices, joint ventures, or operational considerations. Second, we check the data for unit roots and, when appropriate, cointegration. These tests help us to avoid a dynamic misspecification. Third, we estimate a three-sector Feder-Ram growth equation for each of ten NATO allies for the 1951-88 period. Fourth, we report alternative pooled time-series, cross-sectional estimations of this Feder-Ram model for these sample nations using a variety of error component specifications that include fixed effects, one-way random effects, and two-way random effects. Fifth, we apply a host of tests to identify the most appropriate pooled estimation, which turns out to be the two-way fixed-effects model. Finally, we draw policy conclusions.

   By studying a well-defined cohort of allied nations, our Feder-Ram growth study differs from anything in the literature. The majority of our pooled results are consistent with the expectation that investment, labor growth, and the expansion of nonmilitary government expenditures stimulate growth. For our pooled estimates, the growth of defense has a positive and significant impact on growth on the supply side. Because demand-side influences, which are apt to be negative owing to crowding out of investment, are not included, this result is not sufficient for us to conclude that defense is an engine for growth. A modest peace dividend is, nevertheless, anticipated as the composition of the government budget is changed in favor of nondefense spending in the post-cold-war era, since the growth-promoting influence of nondefense public spending exceeds that of defense spending. The growth of defense spillins exerts a small negative impact that is statistically significant in some of the pooled estimates. This finding suggests an input substitutability among the allies' defense efforts. With the exception of labor growth, the alternative pooling techniques yield similar estimates for the coefficients. Because the individual time-series estimates perform more poorly than the pooled estimates, we conclude that the Feder-Ram model appears better at identifying the average behavior over a well-defined cohort than in explaining the behavior of an individual nation.

2. Previous Supply-Side Literature

   To establish the novelty of our contribution, we provide a brief review of Feder-Ram literature and results.(2) In a much-cited contribution, Feder [13] formulated a production function model that related economic growth to investment, labor growth, and the export sector expansion. The Feder model was motivated by the belief that the export sector has important externality and productivity implications for LDCs' economies. Externalities arise because the export sector in many LDCs employs advanced technology that provides growth-enhancing spillovers of knowledge and techniques to other sectors. Moreover, the export sector in LDCs is expected to be more productive than other sectors. Feder's [13] empirical analysis identified these positive externalities and productivity differences. Ram [22] then applied the Feder methodology to determine whether the government sector, in general, yielded positive externalities and contained significant productivity differences conducive to economic growth. For Ram's estimates of the two-sector model, the government sector had a growth-enhancing influence for the cross-section of 115 countries in the study. Time-series analysis of individual nations also supported Ram's hypothesis that the government sector augmented growth in LDCs.

   The positive influence of government expenditures on growth suggests that a similar stimulant for defense might be present. To investigate the supply-side influence of military activity on growth, Biswas and Ram [6] developed a two-sector model consisting of a military and non-military sector. Even though these authors found a weak growth-enhancing military impact for a sample of middle income LDCs, no Benoit effect was discovered for the low income LDCs or for the entire sample.

   Using time-series data, Atesoglu and Mueller [2] applied the two-sector model of Biswas and Ram [6] to the United States, and demonstrated that U.S. military expenditures had a significant, but very small, growth-promoting impact for 1949-89. In a subsequent study, Mueller and Atesoglu [20] disaggregated technological change in the two-sectors from the growth of expenditures. By including technological change, these authors could separate the impact of defense on growth into two components: the relative size of the defense sector and the rate of change in defense spending. Both components were positive and significant, so that defense was shown to have a small but positive influence on growth.

   Alexander [1] developed the most elaborate extension of the Feder-Ram analysis to a four-sector model that included an export sector, a defense sector, a nonmilitary public sector, and a civilian (rest-of-the-economy) sector. A complex structure of externalities among sectors was specified; the nonmilitary sector could influence both the defense sector and the export sector, while the defense, export, and nonmilitary public sectors could affect the civilian sector. Alexander included nine industrial countries (Australia, Austria, Belgium, Canada, Denmark, Finland, the Netherlands, New Zealand, and Sweden) in a pooled time-series, cross-sectional estimation for the 1974-85 period. A growth-promoting influence was found for the export and nonmilitary public sector, but not for the defense sector.

   Although the Alexander study is an important contribution, there are a number of shortcomings with his methodology. First, he never established the stationarity of the macroeconomic aggregates due to limited degrees of freedom. Second, Alexander did not test among alternative pooling techniques to ascertain the most appropriate error structure. Third, he did not account for any allies' defense spillins that could have an impact on the supply side. Fourth, Alexander pooled over a rather diverse group of developed nations. Fifth, his result that investment was a negative and insignificant influence on growth is counterintuitive and against most theoretical paradigms. Our study of the NATO allies addresses these concerns and, in doing so, presents an improved methodology.

   Our position on pooling is somewhat between the positions of those who do not pool over nations [2] and those who pool over large diverse samples. We believe that pooling can provide insights about well-defined cohorts that possess similar economic and political conditions. Pooled estimates give an average picture for a group of nations, and improve on the efficiency of individual country estimates by having many more observations. When variables are defined in terms of growth rates, there may be little annual variation in the variables, and this may make pooling necessary. By pooling over sample data from ten NATO allies for the 1951-88 period, we obtain some of the best empirical results (according to priors) to date.

3. Theoretical Model

   Our empirical tests are based on a three-sector Feder-Ram model containing a civilian sector, a nonmilitary public sector, and a military public sector. We exclude the export sector, because it is unlikely in mature developed economies to provide externalities to other sectors. Additionally, the export sector is anticipated to use similar technologies as those in the civilian sector and elsewhere, so that productivity differences are not relevant. Unlike earlier three-sector models, we allow externalities to arise from the two public sectors and from defense spillins from allies.

   There are two inputs - capital (K) and labor (L - allocated among the three sectors at each point in time, so that

   [K.sub.c] + [K.sub.n] + [K.sub.d] = K, (1)

   [L.sub.c] + [L.sub.n] + [L.sub.d] = L, (2)

   where lower case subscripts refer to the civilian sector (c), the nonmilitary public sector (n), and the defense sector (d). The following three production functions apply:

   N = N ([K.sub.n],[L.sub.n]), (3)

   [Mathematical Expression Omitted],

   and

   [Mathematical Expression Omitted],

   in which N denotes nonmilitary public-sector...