Explaining Inflation Professor Phillips Econ 240A Final Project Nicholas Burger Anthony Mader Elizabeth Mallon John Burnett Mickey Sun Ryan Carl Objective Determine if inflation can be explained by changes in the M3 money supply federal funds rate productivity and federal budget deficit surplus Regression model Dependent variable CPI 1982 100 Independent variables M3 money supply billions of dollars federal budget deficit surplus billions of dollars productivity index output hour federal funds rate H0 1 2 3 4 0 HA At least one 0 Data Collection Relevant data obtained at http research stlouisfed org fred Data analyzed quarterly Exploratory Analysis 10000 400 SURPLUS DEFICIT 8000 M3 6000 4000 2000 0 200 0 200 400 0 50 100 150 200 0 50 CPI 150 200 150 200 CPI 20 140 120 OUTPUT 15 FFR 100 10 5 100 80 60 0 40 0 50 100 CPI 150 200 0 50 100 CPI M3 and Output are directly proportional with CPI FFR and Federal Budget Deficit Surplus are oscillatory while CPI increases Results Model 1 T statistic highly significant for all variables but FFR High R2 value 0 980 and high Fstatistic 2781 589 Low DurbinWatson statistic 0 07 Results Model 1 250 200 150 100 50 20 0 0 20 40 60 60 65 70 75 Residual 80 85 Actual 90 95 Fitted 00 Model follows data well up to 1990 Increased deviation between actual and fitted coinciding with 1991 2001 expansion Results Model 2 First Model t statistic for FFR did not give evidence for a linear relationship between FFR and CPI We ran the regression without this independent variable to see if it significantly improved the validity of our model Results Model 2 T statistics are highly significant and R2 value unchanged at 98 F statistic improved to 4161 575 Durbin Watson statistic still indicates autocorrelation Results Model 3 We also attempted to correct for the apparent lack of correlation between CPI and FFR Changes in the FFR take time to effect the economy lag time of 9 18 months Therefore we shifted the FFR data forward by 9 18 months and regressed against CPI Results Model 3 The 9 12 and 18 month shifts produced t statistics for FFR of 0 488 0 412 and 0 3928 respectively The regression failed to improve the explanatory power of FFR on the behavior of CPI Results Model 4 We attempted to correct the autocorrelation present in our model We ran the regression using the change in each variable s value from the previous quarter Results Model 4 Coefficient for productivity is negative and the Durbin Watson statistic increased to 0 57 R2 decreased dramatically to 0 139 and Fstatistic dropped although still significant at the 5 level Results Model 5 The Last One In order to correct autocorrelation we developed another regression model We added an independent variable to the model that has a time ordered effect on the dependent variable Results Model 5 All variables are linearly related to CPI at the 5 significance level The R2 value and f statistic both increased The DurbinWatson statistic increased Results Model 5 200 150 100 50 10 0 5 0 5 10 15 60 65 70 75 Residual 80 85 Actual 90 95 Fitted 00 This final model follows the data most closely of all the regressions investigated as reflected by the actual fittedresidual curves Conclusions The CPI is negatively correlated with the federal funds rate and productivity while the CPI is positively correlated with the government budget deficit surplus and M3 money supply In order to achieve an accurate model for the relationship between the dependent and independent variables a timeordering variable must be introduced into the regression
View Full Document
Unlocking...