Economic Inflation Data, United States, 1948-1995 Submitted by Stephen J. Guastello The data set consists of monthly readings of economic inflation in the United States for the years 1948 through 1995. These are annualized values. Inflation data are regularly available from the Bureau of Labor Statistics, and are in the public domain. The data set contains both the Producer Price Index (PPI) and the Consumer Price Index (CPI). Most public statements about inflation refer to the CPI. Nonlinear dynamics analysis for this data set were published in Guastello (2000, 2001). The analysis was a response to some public policy debates concerning the possible presence of stability in inflation and its relationship to unemployment. Statistical tests for chaos, fixed-point attractors, and bifurcation structures indicated the presence of chaotic attractors for inflation rates with nonzero epicenters. An annotated SPSS control program, which operates the data set, appears below. It is written in mainframe syntax, which is how the data were analyzed for the publications. The annotations built into the COMMENT lines inform the user about the functions of the various groups of program statements. The fourth program line defines the name of the variables and the spaces that they occupy within a line of data. REFERENCES Guastello, S. J. (2000). Attractor stability in unemployment and inflation rates. Proceedings of the Japan Association for Evolutionary Economics (pp. 23-26). Tokyo: Chuo University/JAFEE. Note: This publication is included with the Structural Equations tutorial on this website. Guastello, S. J. (2001). Attractor stability in unemployment and inflation rates. In Y. Aruka (Ed.). Evolutionary controversies in economics (pp. 89-99). Tokyo: Springer-Verlag. SPSS: title u.s.inflation 1948-1995 file handle datafile name ='inflate-k.dat' data list file=datafile records = 1 /1 month 1-2 year 4-5 ppi 10-14 cpi 18-22 COMMENT THE NEXT LINE DEFINES TIME FROM THE MONTH-DATE DATA. IT IS ONLY USED IN THE LINEAR CONTROL MODEL AND NOT IN THE NONLINEAR EXPONENTIAL MODELS. compute time = (year-48) + (month/12) COMMENT THE NEXT LINE WAS USED IN A PRELIMINARY ANALYSIS TO DETERMINE (ESTIMATES OF) LOCATION AND SCALE. HERE LOCATION WAS SET TO 0.0 FOR THEORY REASONS, AND SCALE WAS SET EQUAL TO THE STANDARD DEVIATION OF VALUES OVER TIME. *frequencies variables = ppi cpi gap/statistics=all COMMENT THE NEXT 3 LINES ARE ORDINARY REGRESSION COMMANDS FOR PPI AND CPI AS FUNCTIONS OF TIME. THEY COMPRISE A LINEAR COMPARISON MODEL. TO SEE THE RESULTS, REMOVE THE * AND RUN SPSS. I SKIPPED THIS STEP ON THE DEMONSTRATION. *regression/missing=pairwise /variables= time ppi cpi /dependent = ppi cpi /enter time COMMENT THE NEXT 2 LINES TRANSFORM THE Y INTO Z BY TAKING PPI, SUBSTRACTING LOCATION (0.0) AND DIVIDING BY SCALE compute nppi = (ppi)/5.08 compute ncpi = (cpi)/3.31 COMMENT THE NEXT 2 LINES DEFINE Z1, GIVEN THAT THE LITERAL OBSERVATION IN OUR DATA SET WILL BE Z2. THE LAG LENGTH IS 3 MONTHS, OR ONE ECONOMIC QUARTER. FOR AMUSEMENT, CHANGE 3 TO 12 AND RUN SPSS. compute nppi1 = lag(nppi, 3) compute ncpi1 = lag(ncpi, 3) COMMENT THE NEXT 3 LINES DO NONLINEAR REGRESSION FOR THE SIMPLE NONLINEAR EXPONENTIAL MODEL AND THE PPI model program a=1 c=2 compute pred = exp(a*nppi1) + c nlr nppi with nppi1 COMMENT NOW WE DO IT AGAIN FOR THE CPI model program a=1 c=2 compute pred = exp(a*ncpi1) + c nlr ncpi with ncpi1 COMMENT THE NEXT 3 LINES DO THE NONLINEAR REGRESSION FOR THE NEXT EXPONENTIAL MODEL LEVEL, BIFURCATION MODEL WITH BIFURCATION VARIABLE UNKNOWN, WITH THE PPI model program a=1 b=2 c=2 compute pred= b*nppi1*exp(a*nppi1) + c nlr nppi with nppi1 COMMENT NOW WE DO THE BIFURCATION UNKNOWN MODEL AGAIN FOR THE CPI model program a=1 b=2 c=2 compute pred= b*ncpi1*exp(a*ncpi1) + c nlr ncpi with ncpi1