Assignment:
1. Professor Ray Fair of Yale University has proposed the following statistical model for predicting presidential election outcome based on recent macroeconomic performance.
His model involves running the following regression:
, Vote = γ0 + γ1party + γ2Person + γ3Duration + γ4War + γ5Growth + γ6 Inflation + γ7Goodnews
where Vote is the vote share of the Democratic party's candidate. Party is the current party of the incumbent, = 1 if a Democrat and = -1 if a Republican. Person is equal 1 if the incumbent candidate is a Democrat, and = -1 if the incumbent candidate is a Republican. Duration is equal to 0 if the incumbent party has already served one term, 1 if served two consecutive terms (-1 if incumbent is Republican), 1.25 if served three consecutive terms (-1.25 if incumbent is Republican), 1.5 if served four consecutive terms (-1.5 if incumbent is Republican). War = 1 in 1920, 1944, and 1948 elections, 0 otherwise. Growth is the growth rate in per capita real GDP in the three quarters previous to the election. Inflation is the average inflation rate in the 15 quarters prior to the election. Goodnews the number of quarters in during the current term in which per capita real GDP growth exceeded 3.2 percent. Each of the economic variables is multiplied by the value of Party.
a. What values (positive or negative) would you expect the parameters of the above regression to be and why?
b. The Fair Model worksheet in the Excel file 4386HW3F16.xls contains data for elections of 1916 through 2012. Use the data in the spreadsheet to estimate the above equation. Are the coefficients statistically significant? Do they conform to your initial expectations about their sign?
c. Using the coefficients estimated in your regression, show the predicted outcome of the model. What does the model predict Democratic candidate Hillary Clinton's vote share should be? Based on current public opinion pools, what is Clinton's expected vote share? Why do think the model is giving a different prediction (if at all) than the opinion polls this year? (Be sure to cite your source of the opinion poll).
2. The Political Business Cycle (PBC) theory predicts that policymakers will try to manipulate the economy in order to be reelected.
a. What type of cyclical behavior would the PBC predict for output and inflation?
b. To test the PBC model, we estimated the following regression for real GDP growth:
?yt = y1 + y2Y2t + y3Y3t + y4Y4t ,
where ?yt is real GDP growth, Y 2t is a dummy variable = 1 during the 2nd year of the presidential term and zero otherwise, Y3t is a dummy variable = 1 during the 3rd year of the presidential term and zero otherwise, and Y 4t is a dummy variable = 1 during the 4th year of the presidential term and zero otherwise.
If the PBC model is true, what would you expect the value of γ4 to be relative to the other parameters?
c. Using the data contained in worksheet PBCc of the excel file 4386HW3F16.xls, estimate the equation in part Q2b. Are the parameter estimates consistent or inconsistent (in a statistical sense) with the predictions of the PBC model and why?
d. Worksheet PBCd. of the excel file 4386HW3F16.xls contains data to estimate the
following regression for inflation:
?∏t = y1 + y2 Y2t + y3Y3t + y4Y4t +α1π t-1
We include lagged inflation in our regression because there is substantial serial correlation in inflation data. Are the resulting estimates for the γ 's consistent with the predictions of the PBC model and why?
5. The partisanship model of political-macroeconomics suggests that economic performance may differ depending which party is in power.
a. Use the data in worksheet PARTIa to estimate the following regression:
?yt = y0 + y1REPt
where ?yt is real GDP growth and REPt = 1 if current president is a Republican and zero otherwise. Based on the regression results, does there appear to be a statistically significant difference between the parties with respect to real GDP growth? Explain. Do the results change if you drop the period 2009-15 from your sample?
b. Use the data in worksheet PARTIb to estimate the following regression for inflation:
?∏t = y0 + y1REPt +α1π t-1.
Does there appear to be a statistically significant difference in inflation depending on which party is in power? Explain.
c. Suppose the economy is characterized by a simple expectations augmented Phillips curve:
y1 = y*t +α (∏t-∏te).
Suppose that under Republican presidents πt = πR , while under Democratic presidents πt = πD . Suppose that Republicans want lower inflation than Democrats so that πD > πR .
Because people must set their expectation of inflation before they know the election output, expected inflation for the period just after the election is πet = pπR + (1- p)πD .
The term p is the probability that the Republican candidate wins the election while (1- p) is the probability that the Democratic candidate wins the elections.
Later in the term people know which party is going to be in power so that πet = πR if a Republican is in power and πet = πD if a Democrat is in power.
Use the above Phillips curve and expected inflation to show which party has the highest level of output just after the election, i.e. what is output if the Republicans win the election and what is output if the Democrats win the election. Later in the term, does it matter for output which party is in power?
d. Worksheet PARTId contains data to estimate the following regression:
?yt = γ1 + γ2DEM2ndt + γ3REP1stt + γ4REP2ndt ,
where DEM 2ndt = 1 if in the 2nd half of a Democrat term and zero otherwise, REP1stt = 1 if in the 1st half of a Republican term and zero otherwise, and REP2ndt = 1 if in the 2nd half of a Republican term and zero otherwise.
Are the parameter estimates in the output growth equation consistent with the above partisanship model? Explain.
e. Worksheet PARTIe contains data to estimate the following regression:
πt =y1 + y2DEM2ndt +y3REP1stt +y4REP2ndt +α1π t-1
Are the parameter estimates (the γ 's ) consistent with the above partisanship model? Explain.
Attachment:- Compatibility mode.rar