17.871Spring 2006Take Home Assignment # 2Handed out March 9Due at the beginning of class March 231. The following table reports the total level of federal spending from FY 1980 to FY 2003,in nominal (i.e., non-inflation-adjusted) dollars. Letting Y denote spending and X theyear, run the following two regressions: (Eq. 1)YXuttt=+ +ββ01and (Eq. 2)lnYXuttt=+ +ββ01Estimate each regression equation and report how you would substantively interpret inβ1each case.FiscalYearExpenditures ($millions)FiscalYearExpenditures ($millions)FiscalYearExpenditures ($millions)1980 590,941 1990 1,253,165 2000 1,788,7731981 678,241 1991 1,324,369 2001 1,863,7701982 745,743 1992 1,381,655 2002 2,010,9701983 808,364 1993 1,409,489 2003 2,157,6371984 851,853 1994 1,461,8771985 946,396 1995 1,515,8021986 990,430 1996 1,560,5351987 1,004,082 1997 1,601,2501988 1,064,455 1998 1,652,5851989 1,143,646 1999 1,701,891Turn in a “do file” that would do any of the variable transformations needed to produce theanswer you provide, plus the “log file” that shows the results of the regressions you ran.22. African Americans vote at a lower rates than whites, even though it’s now a generationsince the passage of the Voting Rights Act. Some argue that this is because old habits diehard in the South, where the most stringent forms of Jim Crow were imposed. Othersargue that this is purely due to the economic disadvantages that African Americanscontinue to labor under.Address this question using a regression. On the class web site you will find a Statadataset named voting2004.dta. This data set was taken from the Voting and RegistrationSupplement to the November 2004 Current Population Survey. It contains informationabout whether people voted in the 2004 election, plus some selective demographics.The codebook is in the file voting2004.txt. Note that the variables of interest are rarelycoded in an immediately useful form. So, an initial issue is how to recode the variablesyou find in order to do the analysis you want to do.I expect you to do a multiple regression to address this question. Choose at least twoindependent variables, and possibly more. You may find it useful to run two regressionsto address this question.The one hint I will give you is about the dependent variable. The dependent variableshould be a binary, or dummy variable, equal to 1 if the person voted, 0 otherwise. Youwill notice that the variable in the data set that carries this information, PES1, is equal to1 if the answer was yes, 2 if the answer was no, and three different negative numbers formissing values. You are going to have to do something about this.After settling on a regression that helps you answer the question the way you feelcomfortable with, write two pages, or so, double-spaced, that do the following:(1) Describe the independent variables and why you chose them. Describe how yourecoded the original variables in the data set to produce the final variables youused in your analysis.(2) Describe the regression you settled on, what the coefficients were, and what thecoefficients mean.(3) Discuss how the results from your multiple regression differ from running a seriesof bivariate regressions and why they differ. (If you do more than one multipleregression, you may answer this question with respect to only one of the multipleregressions.)Finally, turn in a printout of the “do files” that would allow us to reproduce anyregressions or other analyses (like graphs or tables) that you report in your write-up. Also include a copy of the “log file” that is produced when you run these “do files.”3File locations:data: /mit/17.871/www/2006/voting2004.dtacodebook:
View Full Document
Unlocking...