15 968 Problem Set 1 Due Date October 1 1999 This problem set is an empirical replication extension of McGahan Porter s results on Australian data You can download a stata dataset australia dta at the following URL http web mit edu 15 932 www phd 15943 html The data is from Furman 1998 1 an unbalanced panel of 5 variables at the business unit level of analysis The variables are the following cusip a firm identifier year varies from 1992 to 1996 seg industry classification SIC code f seg the number of segments the firm is involved in r the rate of return on assets for each firm in each of its line of business in each year it was recorded For example company x represented by CUSIP 23 had a rate of return of 13 in 1992 in industry 3089 You are permitted to work in groups on the problem set up to 4 people While one program output per group is acceptable please do provide individual write ups of your findings Question 1 Furman follows M P s methodology for his analysis of the profitability of Australian firms with the caveat that he does not perform a correction for first order serial correlation a Replicate Furman s results by using a set of fixed effects regression see table 5 1 attached Why are the cells in italic left blank Fill in table 5 2 b Implement the AR 1 correction on the Australian data Does it change the previous results and why Compare and interpret these final results with M P s findings on US data Question 2 Focus on a single cross section by restricting your analysis to the year 1994 How are the results affected In light of the discrepancies in the findings of M P and Schmalensee with US data is this surprising Why Question 3 This sample compiles the rates of return of public companies Assume for a minute that you had a dataset wich included compatible data on private companies along with the data you presently have How would you expect the relative importance of business unit effects industry effects and corporate parent effects to change Why Question 4 Some theories of within industry firm performance differentials invoke the notion of CEO leadership Assume that this dataset also had CEO dummies Could you test leadership explanations that way Explain Problems pierrot mit edu or 253 6620 or 547 5805 Does Industry Matter differently in Different Places Working Paper Sloan School of Management MIT 1 Please note that this dataset is being used with the permission of Jeff Furman he requests that you contact him if you would like to use the data for any purpose other than that of this problem set jfuman mit edu Table 5 1 Results for Australia ANOVA statistics from ordinary least squares estimation on core model R2 R2 Year Effects Industry Effects 0 0029 0 2185 BusinessSpecific Effects 0 7151 Year Industry Effects 0 2194 Year BusinessSpecific Effects 0 7154 Year Corporate Parent Effects 0 1365 Year Industry BusinessSpecific Effects Year Industry Corporate Parent Effects Year Corporate Parent BusinessSpecific Effects R2 0 3316 Year Industry Corporate Parent Business Segment Effects R2 Table 5 2 Fraction of explained variance R2 attributable to each type of effect Australia Year Industry Corporate Corporate Parent Effects 0 1348 Parent Business Specific Full Model 100 0