6.1.2Urbanization Economies vs. Localization Economies

A question that has generated many empirical studies is whether localization economies or urbanization economies have the stronger effects. Sveikauskas (1975), Segal (1976) and Moomaw (1981) have attempted to show that gains in productivity increase with city size, which illustrates urbanization economies. Shefer (1973) examined the effects of local industry size on productivity (localization economies) but ignored the effects of population (urbanization economies). He found that localization economies have unstable effects on productivity over time. Carlino (1978) rejected this result in favor of urbanization economies, arguing that “population is a worthless surrogate for business agglomeration economies.” Population reflects urban diseconomies, such as congestion, which offset business agglomeration economies. Finally, in order to clarify the debate, Moomaw (1983) suggested a refinement of the estimates measuring urbanization and localization economies. He asked whether urban agglomeration economies or diseconomies dominate.

The methodology employed by Moomaw consisted of regressing the logarithm of value added per worker on population and other variables. His data covered 2-digit manufacturing industries for the year 1977. He found a positive and significant coefficient for population for most industries, which refutes Carlino’s proposition, and shows that population does matter in agglomeration economies. These studies have led to additional work on measuring the contribution of agglomeration economies to labor productivity, in a static and dynamic fashion.

Henderson (1986) studied urbanization and localization economies in U.S. and Brazilian cities, further developing Moomaw’s work. Instead of a specific production function, he used a flexible functional form. He found that two-stage least squares estimation had the effect of strengthening estimates of localization economies and weakening estimates of urbanization economies. Unlike Moomaw’s results, Henderson’s results showed stronger effects of urbanization economies.

Sveikauskas, Gowdy, and Funk (1988) studied urbanization and localization economies in the food processing industry. Focusing on one industry only allowed the authors to get better data. They used a translog production function, allowing for increasing returns to industry size in order to measure localization economies. Their results indicate that, when the extent of nearby agricultural production is included in the production function, the economies of scale coefficient is not significantly different from 1. The coefficient for SMSA population is positive and significant. These results contrast with the ones that would have been obtained using Henderson’s method. Thus, productivity estimates seem to be very sensitive to the production function specification, measurement of industry scale effects, and the particular data used.

Garnick and Renshaw (1980) and Hawley and Fogarty (1981) argued that the productivity advantages of urban areas have declined relative to these enjoyed by other locations. Carlino (1985) explained this phenomenon by the changes in production method made in communication and in transportation. He used a CES production function, assuming the labor market is in equilibrium and estimated the following equation:

lnWit + lnLit = lnA + α1Nit + α2N2 it + β lnQit + ho(1-β) lnLit + h1(1-β) lnL2 it(6.1)

where N is population, Q total manufacturing value added per establishment, W is total manufacturing payroll per employee, L is manufacturing employment per establishment, h= h0 + h1t is the returns to scale parameter, β is the elasticity of the wage with respect to changes in output. Data are for 80 SMSA, for the 1957-1977 period. His results indicate that h= 1.13 + 0.002t, which means that scale economies were present but increased slowly.

Moomaw (1985) then extended Carlino’s work by regressing value added in manufacturing per productive and nonproductive worker hour on SMSA population, value added minus payroll, and dummies for regions. The 1967 and 1977 data come from 18 2-digit manufacturing sectors. Moomaw used a fixed-effects model (dummy for 1977) and found little evidence of change in Hicks-neutral productivity advantages of large relative to small SMSAs.

Henderson (1997) estimated dynamic externalities in a panel analysis framework, which allows separation of externalities from fixed effects and identification of a lag structure. He found strong evidence of localization economies. Urbanization economies effects are smaller, but their effects persist to the end of the time horizon, whereas localization economies’ effects usually disappear after six years.

The next section describes in further detail the different techniques used to measure spatial differences in productivity and the role of externalities.