Abstract: Capital stocks are notoriously hard to measure, and errors in them bias production function estimates. I develop an estimator that corrects for measurement error in capital built with the perpetual inventory method, using firms' initial intermediate inputs as an instrument for true initial capital. In Chilean manufacturing data, standard methods underestimate the capital elasticity by up to 50%, and up to 31% of estimated productivity variance in standard methods reflects measurement error in initial capital rather than true heterogeneity between firms.
Firm data commonly record output and intermediate inputs as annual totals, even though firms adjust inputs within the year. I show that discrete-time production function estimators are inconsistent when input levels vary within the period of observation, and I propose a corrected estimator. Modern estimators rely on timing assumptions about how far in advance firms must plan changes to each input, but even models that allow inputs to adjust slowly tend not to grapple with the possibility that slow adjustment causes input use to vary within a period. Within-period variation can also arise from price fluctuations, especially for inputs with volatile prices, such as energy.
I develop a model in which the firm adjusts intermediate inputs flexibly in continuous time, while capital and labor change only gradually over the year. Aggregating output and intermediates discards information on input timing and introduces an aggregation error correlated with inputs, making standard estimates of the labor and capital elasticities inconsistent. My correction relies on piecewise interpolation, performs well in Monte Carlo exercises, and readily incorporates data observed at mixed frequencies, such as monthly energy prices.
The common assumption of firm risk-neutrality is a convenient mathematical simplification but may not accurately reflect firm behavior. For instance, even if firm owners are risk-neutral, the principal-agent problem could result in risk-averse firm behavior if management is risk-averse. One way to infer firms’ risk preferences is to examine the their decisions over how much inventory to hold. Risk-averse firms uncertain about future prices and productivity can hedge against future negative shocks by building up inventories. I develop a model of firm inventory behavior under which a risk-averse firm’s production function can be identified from the firm’s first-order condition on inputs. Risk aversion itself can then be estimated using the firm’s first-order condition on inventory quantity.