| This study examines the potential role of frontier efficiency calculations in productivity analysis. For a sample of manufacturing plants taken from Census of Manufacturing data, two different techniques are evaluated for measuring plant performance: one, productivity-based, and the other, based on how plants compare to the frontier-efficient technology within their industry. Data Envelopment Analysis (DEA), a linear programming procedure, is used to estimate nonstochastic production frontiers for each industry. Plant efficiency relative to the DEA frontier is contrasted with more traditional relative productivity calculations based on the Tornqvist Index.;Results show that it makes little difference which approach is used in terms of ranking plant performance: rank correlations of the two measures are very high. It is also found that, whichever measure of performance is used, plant success or failure is persistent over 5-year, 10-year, and even 15-year periods.;Nevertheless, industry efficiency estimates compare fairly well with observed industry structural characteristics. Average efficiency levels are consistent with a number of industry-level variables, such as: the number of plants, average plant age and size, price-cost margins, inventories held, specialization ratios, production worker wages and hours worked.;Efficiency measurements prove most enlightening when they are used to decompose productivity growth into technical change and efficiency change components following the methodology of the Malmquist Index. Regression results show significant independence of these two components. Surprisingly, the technical change component, at least for this dataset, is the less variable component, and better explained by the data. Efficiency change, although noisier than technical change, is seen to be more important for increasing total productivity. Results indicate significant advantages to plants that attempt to increase productivity by improving efficiency rather than innovating to improve technology. That is, "catching up" to the best producers in the industry pays greater dividends than pushing out the production frontier through basic research. |
