Decomposing Learning By Doing in New Plants
December 1992
Working Paper Number:
CES-92-16
Abstract
The paper examines learning by doing in the context of a production function in which the other arguments are labor, human capital, physical capital, and vintage as a proxy for embodied technical change in physical capital. Learning is further decomposed into organization learning, capital learning, and manual task learning. The model is tested with time series and cross section data for various samples of up to 2,150 plants over a 14 year period. Word Perfect Version
Document Tags and Keywords
Keywords
Keywords are automatically generated using KeyBERT, a powerful and innovative
keyword extraction tool that utilizes BERT embeddings to ensure high-quality and contextually relevant
keywords.
By analyzing the content of working papers, KeyBERT identifies terms and phrases that capture the essence of the
text, highlighting the most significant topics and trends. This approach not only enhances searchability but
provides connections that go beyond potentially domain-specific author-defined keywords.
:
investment,
production,
productive,
manufacturing,
industrial,
growth,
earnings,
technical,
gain,
employee,
organizational,
managerial,
employed,
labor,
shift,
specialization,
producing,
development,
productivity size,
wages productivity
Tags
Tags are automatically generated using a pretrained language model from spaCy, which excels at
several tasks, including entity tagging.
The model is able to label words and phrases by part-of-speech,
including "organizations." By filtering for frequent words and phrases labeled as "organizations", papers are
identified to contain references to specific institutions, datasets, and other organizations.
:
Internal Revenue Service,
American Statistical Association,
Longitudinal Research Database,
Center for Economic Studies,
New England County Metropolitan,
Bureau of Economic Analysis,
Longitudinal Business Database
Similar Working Papers
Similarity between working papers are determined by an unsupervised neural
network model
know as Doc2Vec.
Doc2Vec is a model that represents entire documents as fixed-length vectors, allowing for the
capture of semantic meaning in a way that relates to the context of words within the document. The model learns to
associate a unique vector with each document while simultaneously learning word vectors, enabling tasks such as
document classification, clustering, and similarity detection by preserving the order and structure of words. The
document vectors are compared using cosine similarity/distance to determine the most similar working papers.
Papers identified with 🔥 are in the top 20% of similarity.
The 10 most similar working papers to the working paper 'Decomposing Learning By Doing in New Plants' are listed below in order of similarity.
-
Working PaperDecomposing Technical Change🔥
May 1991
Working Paper Number:
CES-91-04
A production function is specified with human capital as a separate argument and with embodied technical change proxied by a variable that measures the average vintage of the stock of capital. The coefficients of this production function are estimated with cross section data for roughly 2,150 new manufacturing plants in 41 industries, and for subsets of this sample. The question of interactions between new investment and initial endowments of capital is then examined with data for roughly 1,400 old plants in 15 industries.View Full Paper PDF
-
Working PaperManagerial Efficiency, Organizational Capital and Productivity🔥
March 2003
Working Paper Number:
CES-03-08
The paper focuses on the impact of managerial efficiency on output. Three sources of managerial efficiency are identified: (a) superior initial managerial endowments, (b) the accumulation of managerial knowledge and skills through learning and (c) the impact of an effective market for managerial resources internal to the firm. All three are explicitly measured by appropriate variables and their impact is examined in the context of variously specified production functions. The empirical analysis is carried out with data for approximately 5,000 new manufacturing plants in the United States over the 1973-92 period. It is found that variation in managerial endowments is an important explanatory variable for output with all other relevant inputs controlled. It is further found that the survival of plants with superior managerial efficiency, and the death of those with inferior efficiency, explains a substantial fraction of total factor productivity change in the manufacturing sector of the U.S. economy. There is also clear evidence of the significance for efficiency of internal markets as well as evidence of learning as plants age. Learning and superior managerial resources of old plants largely offset the benefits of capital goods of later vintage of new plants.View Full Paper PDF
-
Working PaperThe Demand for Human Capital: A Microeconomic Approach🔥
December 2001
Working Paper Number:
CES-01-16
We propose a model for explaining the demand for human capital based on a CES production function with human capital as an explicit argument in the function. The resulting factor demand model is tested with data on roughly 6,000 plants from the Census Bureau's Longitudinal Research Database. The results show strong complementarity between physical and human capital. Moreover, the complementarity is greater in high than in low technology industries. The results also show that physical capital of more recent vintage is associated with a higher demand for human capital. While the age of a plant as a reflection of learning-by-doing is positively related to the accumulation of human capital, this relation is more pronounced in low technology industries.View Full Paper PDF
-
Working PaperThe Life Cycles of Industrial Plants
October 2001
Working Paper Number:
CES-01-10
The paper presents a dynamic programming model with multiple classes of capital goods to explain capital expenditures on existing plants over their lives. The empirical specification shows that the path of capital expenditures is explained by (a) complementarities between old and new capital goods, (b) the age of plants, (c) an index that captures the rate of technical change and (d) the labor intensiveness of a plant when it is newly born. The model is tested with Census data for roughly 6,000 manufacturing plants that were born after 1972.View Full Paper PDF
-
Working PaperThe Survival of Industrial Plants
October 2002
Working Paper Number:
CES-02-25
The study seeks to explain the attrition rate of new manufacturing plants in the United States in terms of three vectors of variables. The first explains how survival of the fittest proceeds through learning by firms (plants) about their own relative efficiency. The second explains how efficiency systematically changes over time and what augments or diminishes it. The third captures the opportunity cost of resources employed in a plant. The model is tested using maximum-likelihood probit analysis with very large samples for successive census years in the 1967-97 period. One sample consists of an unbalanced panel of about three-fourths of a million plants of single and multi-unit firms, or alternatively of about 300,000 plants if only the most reliable data are considered. The second is restricted to the plants of multi-unit firms in the same time span and consists of an unbalanced panel of more than 100,000 plants. The empirical analysis strongly confirms the predictions of the model.View Full Paper PDF
-
Working PaperIndustry Learning Environments and the Heterogeneity of Firm Performance
December 2006
Working Paper Number:
CES-06-29
This paper characterizes inter-industry heterogeneity in rates of learning-by-doing and examines how industry learning rates are connected with firm performance. Using data from the Census Bureau and Compustat, we measure the industry learning rate as the coefficient on cumulative output in a production function. We find that learning rates vary considerably among industries and are higher in industries with greater R&D, advertising, and capital intensity. More importantly, we find that higher rates of learning are associated with wider dispersion of Tobin's q and profitability among firms in the industry. Together, these findings suggest that learning intensity represents an important characteristic of the industry environment.View Full Paper PDF
-
Working PaperProductivity Adjustments and Learning-by-Doing as Human Capital
November 1997
Working Paper Number:
CES-97-17
This paper measures plant-level productivity gains associated with learning curves across the entire manufacturing sector. We measure these gains at plant startups and also after major employment changes. We find: 1.) The gains are strongly associated with a variety of human capital measures implying that learning-by-doing is largely a firm-specific human capital investment. 2.) This implicit investment is large; many plants invest as much in learning-by-doing as they invest in physical capital and much more than they invest in formal job training. 3.) This investment differs persistently over industries and is higher with greater R&D. 4.) Consistent with a learning-by-doing interpretation, the human capital investment is much larger following employment decreases than increases. We conclude that learning-by-doing is a major factor in wage determination, technical progress and asymmetric employment adjustment costs.View Full Paper PDF
-
Working PaperThe Impact of Vintage and Survival on Productivity: Evidence from Cohorts of U.S. Manufacturing Plants
May 2000
Working Paper Number:
CES-00-06
This paper examines the evolution of productivity in U.S. manufacturing plants from 1963 to 1992. We define a 'vintage effect' as the change in productivity of recent cohorts of new plants relative to earlier cohorts of new plants, and a 'survival effect' as the change in productivity of a particular cohort of surviving plants as it ages. The data show that both factors contribute to industry productivity growth, but play offsetting roles in determining a cohort's relative position in the productivity distribution. Recent cohorts enter with significantly higher productivity than earlier entrants did, while surviving cohorts show significant increases in productivity as they age. These two effects roughly offset each other, however, so there is a rough convergence in productivity across cohorts in 1992 and 1987. (JEL Code: D24, L6)View Full Paper PDF
-
Working PaperThe Structure Of Production Technology Productivity And Aggregation Effects
August 1991
Working Paper Number:
CES-91-05
This is a sequel to an earlier paper by the author, Dhrymes (1990). Using the LRD sample, that paper examined the adequacy of the functional form specifications commonly employed in the literature of US Manufacturing production relations. The "universe" of the investigation was the three digit product group; the basic unit of observation was the plant; the sample consisted of all "large" plants, defined by the criterion that they employ 250 or more workers. The study encompassed three digit product groups in industries 35, 36 and 38, over the period 1972-1986, and reached one major conclusion: if one were to judge the adequacy of a given specification by the parametric compatibility of the estimates of the same parameters, as derived from the various implications of each specification, then the three most popular (production function) specifications, Cobb-Douglas, CES and Translog all fell very wide of the mark. The current paper focuses the investigation on two digit industries (but retains the plant as the basic unit of observation), i.e., our sample consists of all "large" manufacturing plants, in each of Industry 35, 36 and 38, over the period 1972-1986. It first replicates the approach of the earlier paper; the results are basically of the same genre, and for that reason are not reported herein. Second, it examines the extent to which increasing returns to scale characterize production at the two digit level; it is established that returns to scale at the mean, in the case of the translog production function are almost identical to those obtained with the Cobb-Douglas function.1 Finally, it examines the robustness and characteristics of measures of productivity, obtained in the context of an econometric formulation and those obtained by the method of what may be thought of as the "Solow Residual" and generally designated as Total Factor Productivity (TFP). The major finding here is that while there are some differences in productivity behavior as established by these two procedures, by far more important is the aggregation sensitivity of productivity measures. Thus, in the context of a pooled sample, introduction of time effects (generally thought to refer to productivity shifts) are of very marginal consequence. On the other hand, the introduction of four digit industry effects is of appreciable consequence, and this phenomenon is universal, i.e., it is present in industry 35, 36 as well as 38. The suggestion that aggregate productivity behavior may be largely, or partly, an aggregation phenomenon is certainly not a part of the established literature. Another persistent phenomenon uncovered is the extent to which productivity measures for individual plants are volatile, while two digit aggregate measures appear to be stable. These findings clearly calls for further investigation.View Full Paper PDF
-
Working PaperLarge Plant Data in the LRD: Selection of a Sample for Estimation
March 1999
Working Paper Number:
CES-99-06
This paper describes preliminary work with the LRD during our tenure at the Census Bureau as participants in the ASA/NSF/Census Research Program. The objective of the work described here were two-fold. First, we wanted to examine the suitableness of these data for the calculation of plant-level productivity indexes, following procedures typically implemented with time series data. Second, we wanted to select a small number of 2-digit industry groups that would be well suited to the estimation of production functions and systems of factor share equations and factor demand forecasting equations with system-wide techniques. This description of our initial work may be useful to other researchers who are interested in the LRD for the analysis of productivity growth and/or the estimation of systems of factor equations, because the specific results reported in this memo suggest that the data are of good quality, or because the nature of the tasks undertaken provides insight into issues that arise in the analysis of longitudinal establishment data.View Full Paper PDF
