-- RalfMartin - 27 May 2007

PublicationForm edit

Title TFP without Capital Stocks
Author Ralf Martin
Year 2005
Abstract I propose a new way of dealing with the fact that often only in-
vestment but not capital stocks are reported in business micro data
sets. Capital stock information is essential to calculate measures of
total factor productivity (TFP). Because data on investment is usu-
ally available, on the other hand, the standard practice is to apply
some perpetual inventory method (PIM). This is not without prob-
lems however. A PIM calculation requires many assumptions. Besides
assumptions on depreciation, we need to come up with guesses of the
initial values. If plants are subject to random sampling so that they
are not observed in all periods they are alive then we equally have to
guess their investment levels in those periods. Standard practice is
to do these various steps in fairly ad hoc ways which are not neces-
sarily mutually consistent. Moreover, once an estimate of the plant
level capital is obtained it is often treated in subsequent productivity
calculations like a variable that is directly observed although it is pre-
sumably subject to serious measurement error. The paper addresses
these issues by developing a novel framework which allows the simul-
taneous estimation of TFP and capital stock using simulation based
inference methods. I also develop a test which investigates if the re-
strictions implied by more standard ways of computing the capital
stock matter empirically.
DownloadInfo Ceriba mimeo
Topic revision: r1 - 27 May 2007, RalfMartin

Copyright © by CeRiBA.
Ideas, requests, problems? Send feedback