|
COPYRIGHT 2001 MIT Press Journals
I. Introduction
ECONOMISTS have long argued that best-practice technology is embodied in new capital so productivity of a manufacturing plant should be associated with its entry year, or vintage. A plant's age, on the other hand, impacts productivity for different reasons. As plants age, managers accumulate experience, gain from learning by doing, undertake new investments, or achieve economies of scale, all of which can improve plant-level productivity.
This suggests that two forces--a vintage effect from higher productivity of recent entrants and a survival effect from increased productivity of surviving plants--interact to drive productivity growth for an industry and jointly push out the productivity frontier. Because many cohorts from different periods coexist at each point in time, however, the competitive process should lead to a rough convergence in productivity across cohorts as the countervailing influences of vintage and survival balance.(1)
This paper explores and measures vintage and survival effects using cohort data from nineteen U.S. manufacturing industries from 1963 to 1992. We compare the relative productivity of cohorts of different vintages (entry year) and different ages (number of years since entry) to answer several well-defined questions about productivity growth. First, do more-recent cohorts enter with higher productivity than earlier cohorts? Second, do surviving plants of a particular vintage become more productive as they age? Third, how do these two effects trade off at a particular time? That is, do the improved capital, technologies, and practices behind vintage gains create an advantage for younger plants, or do older, surviving plants show higher productivity from the experience, learning, and scale effects associated with age?
The paper is largely empirical, and we structure our analysis around a carefully specified framework that systematically attempts to control for exogenous factors. Disembodied technical change, improved management techniques, and demand or supply shocks, for example, could affect the productivity of all plants in ways unrelated to the specific impact of vintage or age. Because failure to control for these factors can bias estimates of the vintage and survival effects, we use industry-wide variables such as total industry output, average industry productivity, and change in total output to control for general, time-related factors to identify the vintage and survival effects. Our results suggest that both effects are large and important.
The 1992 cohort of new entrants was 51% more productive than the 1967 cohort in its entry year of 1967. Even after controlling for industry-wide factors and input differences, the vintage effect is greater than 50% from 1967 to 1992. Because this analysis compares successive cohorts of new entrants in their first period, we conclude that the improved capital, technology, and operating practices embodied in new plants and equipment are an important source of productivity growth.
Analysis of a single cohort of surviving plants over time shows that age is also a good predictor of productivity. The 1967 cohort of surviving plants, for example, increased productivity by 57% between 1967 and 1992. Although much of these gains were associated with industry-wide factors, productivity growth of 19% can be independently attributed to age-specific factors such as experience and scale. These surviving plants showed steady gains and improved their relative standing in the productivity distribution.
Finally, the data suggest that vintage and survival effects roughly offset each other in a given year. In 1992, all surviving cohorts that entered prior to 1992 showed average productivity within 6.5% of the industry average. Only the 1992 cohort was far from the industry average (14.2% below), reflecting the important processes of competition and selection as competitive forces have not yet forced the lowest-productivity plants from the market. Nonetheless, this sorting process appears quite rapid. The 1987 cohort, for example, entered with productivity 10% below the industry in 1987, but selection and survival gains allowed the surviving plants to improve to only 2.7% below the industry average by 1992. As low-productivity plants failed and surviving plants improved with age, the cohort made large relative productivity gains.
These results highlight the complexity of the productivity dynamics that contribute to industry-level productivity growth. As new plants continually enter with higher productivity, competition weeds out the poor performers, and survivors improve with age, the productivity frontier is pushed out, and average productivity rises. Only by looking at relative productivity across different cohorts at different points in time can we understand these important sources of productivity growth.
II. Productivity and Microdata
Understanding productivity growth is obviously an important topic, but numerous measurement and conceptual problems make it difficult and controversial. One robust finding from microstudies, however, is that distributions of plant-level productivity show enormous heterogeneity both within and across industries. Jensen and McGuckin (1997), for example, report productivity ranges of 4 or 5 to 1 as the norm within four-digit manufacturing industries.
A second robust finding is the growth and survival of efficient producers while inefficient producers decline and exit. Haltiwanger (1997)--rebuilding on Baily, Hulten, and Campbell (1992) and Bartelsman and Dhrmyes (1998)--attributes approximately 20% of U.S. manufacturing productivity growth to the entry of new plants and the exit of unsuccessful plants. New entrants exhibit average productivity levels well below those of incumbents, and a large fraction of these new entrants eventually fail. Haltiwanger (1997) estimates an additional 40% of manufacturing productivity growth is due to reallocations of output among surviving plants, with the remaining 40% reflecting industry-wide factors shared by all plants. Thus, approximately 60% of manufacturing productivity growth is associated with the growth of successful producers at the expense of the less successful rivals.
These facts imply evolution of the productivity distribution--which reflects a process of "creative destruction"--is a key factor in overall productivity growth. Plant vintage and plant age are two factors that contribute to this evolution. Recent empirical work, however, has paid little attention to distinguishing the two. With the notable exception of Bahk and Gort (1993), who focus directly on learning by doing and the separate impact of age and vintage, previous empirical work has typically treated these two characteristics as synonymous. Baily et al. (1992), for example, argue that the "the age of the plant is an obvious way to measure vintage" (p. 197). For some purposes this is reasonable, but a deeper vintage and age story lies behind the evolution of plant-level productivity. This paper explores the interaction of these two key factors by isolating and measuring the impact of productivity changes for new entrants (vintage effect) and surviving incumbents (survival effect) in U.S. manufacturing plants from 1963 to 1992.
III. Definitions and Hypotheses
We define a plant's vintage, v, as the year a plant first produces; a plant's age, a, as the number of periods the plant has been in operation; and time, t, as the period in which a plant is observed. For each plant, therefore, t [equivalent] [Upsilon] + a. We begin with a traditional production function:
(1) [Y.sub.t,i,a] = [f.sub.[Upsilon]]([K.sub.t,i], [H.sub.t,i], [M.sub.t,i], [a.sub.i], t),
where [f.sub.[Upsilon]](*) is the production function for plants of vintage [Upsilon],
[Y.sub.t,i,a] is real gross output,
[K.sub.t,i] is capital,
[H.sub.t,i] is labor hours,
[M.sub.t,i] is intermediate inputs, and
[a.sub.i] is age, all for plant i in period t.
Average labor productivity for plant i in period t of age a, [y.sub.t,i,a], is
(2) [y.sub.t,i,a] = [Y.sub.t,i,a] - [M.sub.t,i]/[H.sub.t,i],
where [y.sub.t,i,a] is value added per hour worked.(2)
To untangle the relationship between entrants, survivors, and productivity, we define a vintage effect as
(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
and a survival effect as
(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],
where [y.sub.t,s,a] is employment-weighted labor productivity of all plants in industry s at time t of age a.
The definitions appear similar, but they analyze different cuts of a larger data set. The vintage effect in equation (3) compares productivity across cohorts of new plants (a = 0) to measure how productivity evolves before age-related factors...
Read the full article for free courtesy of your local library.
|
Bookmark this article
Print this article
Link to this article
Email this article
Digg It!
Add to del.icio.us
RSS
