This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: 139Spring 09 Midterm 2Name ..................................................................DO NOT TURN OVER UNTIL TOLD TO DO SONotes:1. Use a calculator where necessary.2. Show your work, place your answers in the spaces provided and make use of the cheat sheet.3. If you want to make any additional assumptions to simplify a problem, please state them clearly.5.Keep verbal answers concise(no more than 4 sentences, and bullet points are &ne).5. There are 75 points in total (1 point per minute)Background:The &rst two questions in this exam are motivated by the estimation ofproductionfunctions, a common activity of applied economists.A production function measures how a &rmtransforms inputs (such as labor and physical capital) into output. The simplest production functionis a CobbDouglas, whereYi="iK&1iL&2iwhereYis some measure of valueadded in sales by &rmi(the value of output less the cost of material inputs),Kiis a $ measure of the &rm¡s physical capital(e.g., machines) andLiis its $ labor input."ire¢ects the productivity of the &rm (those with higher"is produce more for given inputs).For estimation it is standard to take logs and add a constant.This giveslogYi=&+&1logKi+&2logLi+"iA &rm with a higher"iis said to have "higher productivity" or to be "more productive".We canthen add other controls to the speci&cation to pick up additional e/ects which may matter. The datain the questions comes from French &rms surveyed in 1973, 1978, 1983 and 1988.11. You begin by estimating three di/erent speci&cations using only data from a single year (1978).You use OLS and calculate heteroskedasticity robust standard errors. The coe¢ cients and standarderrors are shown in the following table. The sales variable has been converted to a value added form,and prior to take logs the sales, labor and capital variables were all measured in millions of $s.(1)(2)(3)(4)Dep. Varlog(sales)log(sales)log(sales)log(sales)log(capital)0.5050.5060.4170.422(0.026)(0.027)(0.021)(0.022)log(labor)0.4670.4720.5090.505(0.035)(0.037)(0.028)(0.030)log(R¡D)0.0060.0530.061(0.019)(0.013)(0.014)Computer1.9591.970(0.051)(0.054)Government0.028(0.010)Constant2.7762.7853.0403.074(0.081)(0.082)(0.062)(0.076)Obs.890890890890R20.910.910.960.96Std. errors in parentheses are robust to heteroskedastictyIn the &rst speci&cation (column (1)) you only include log(labor), log(capital) and a constant asregressors, with log(sales) as the dependent variable.Questions (a)(c)(ii) all refer to the resultsin this column.(a) [3 points] Does the coe¢ cient on the constant have an interpretation in this regression?Ifyou£re not sure, explain on what additional information your answer would depend.SOLUTION: Yes¤maybe. It would re¥ect expected logsales of a &rm having zero logcapital (i.e.,capital = $1M) and zero loglabor (i.e., labor = $1M). So to know whether to interpret it you need toknow if there are &rms in this range of the data. As it happens both of these amounts are well withinin the range of the data so the constant coe¢ cient does have some meaning here.in the range of the data so the constant coe¢ cient does have some meaning here....
View
Full
Document
This note was uploaded on 08/03/2011 for the course ECON 139 taught by Professor Alessandrotarozzi during the Spring '08 term at Duke.
 Spring '08
 ALESSANDROTAROZZI
 Econometrics

Click to edit the document details