HOME VALUES AND DEMOGRAPHIC CHANGE OAK PARK, ILLINOIS 1984-2002
Because of the controversial nature of the results of this study, it will not be presented in its entirety at this site until it has been published in a refereed journal of urban studies.
A synopsis of the research, with discussion of the quantitative methods employed, will be presented so that other scholars have the opportunity to duplicate the work or utilize its techniques for similar studies in other regions.
PROPERTY VALUES, TAXATION, AND DEMOGRAPHIC CHANGE IN OAK PARK, ILLINOIS, 1984-2002
Leslie M. Golden
University of Illinois, Chicago
Submitted December 4, 2002
Summary. The values of sales prices of 10.800 homes in Oak Park, Illinois, over a 19 year period were studied as a function of 10 demographic variables. We found significant functional dependences with [edited]. A positive secular variation is attributable to [edited].The effect on home values of demographic changes is strongest with a two-year time lag, indicating that policy adversely affecting home values can be reversed on that time scale.
Oak Park, Illinois, a land-locked suburb of population 53,000 immediately west of Chicago, has been heralded nationally by its civic leaders as a model of progressive social policy and integration. Beginning in the 1950's, the Chicago neighborhoods to the east experienced rapid racial change, large unemployment, high crime, housing decay, high rates of illegitimate births, and a gang culture. In response, Oak Park beginning in the 1960's evolved from a conservative community to one embracing open housing and racial integration in an effort to prevent similar decay.
Oak Park was marketed to homeseekers in the liberal Chicago communities of Hyde Park and Lincoln Park, resulting in several major demographic changes. The wealthy class of small business owners was replaced by a wealthy class of professionals. This was accelerated by the opening of the Chicago campus of the University of Illinois in 1965 and the expansion of the nearby Cook County-University of Illinois-Rush St. Luke Presbyterian and Loyola University medical centers.
In addition, the poor working class of factory workers was replaced by the poor single parent black family. This was accelerated by the razing of the nearby Hawthorne Works of Western Electric, the general loss of small manufacturing jobs in the Chicago area, and the decay of the adjoining neighborhoods of Chicago. Third, the aforementioned marketing led to the influx of large number of homosexuals, giving Oak Park the largest concentration of that group in the Chicago area other than Hyde Park and Lincoln Park.
Several effects resulted from these changes. First, the population of Oak Park decreased from 63,000 in the 1950's, when it carried the title of "World's Largest Village," to its current size. Second, the middle class was severely reduced in size resulting in a bi-modal distribution of incomes. Third, Oak Park's crime rate rose precipitously in the 1980's with the spillover of crime from Chicago to the east, until it ranks among the highest 10% of communities in crime rate in Cook County. The annual crime rate since about 1992 has oscillated by 5 - 10% year about a saturation level.
With these changes came increased demand and need for public services and an explosion of property taxes. Oak Park earned the distinction of having the highest property taxes in the state of Illinois and the highest property taxes normalized by assessed valuation in the United States.
THE DATA AND ANALYSIS
The values of home sales prices were obtained for 10,800 homes sold in Oak Park from 1984 through 2002. For purposes of analysis, Oak Park [edited]. Median values for those homes sold [edited] in each year are calculated and are used as the dependent variable. Because mean values of home sales are skewed from the sale of a few expensive homes, median values are considered a more meaningful measure of sales trends than mean values.
The independent variables consist of any data that is tabulated on an annual basis [edited]. These include [edited]. The year of the data is used as an explicit time variable. Accordingly, no need exists for correcting any of the data for inflation. Regressions were, however, also performed correcting all dollar-denominated variables by the consumer price index to determine actual home appreciations. The total tax levy was determined by multiplying the total equalized assessed valuation of Oak Park property by the tax rate, each of which is available on an annual basis.
Among the data that were not available on an annual basis were those of [edited]. As noted above, recent crime values have not shown any significant trend. Such figures are notorious for their politically-motivated non-reliability and do not satisfy criteria for scholarly analysis.
A standard multi-variate regression was performed on the data using a modified Gauss-Jordan technique. The dependent variable is the annual mean sales price. The independent variables are those noted above. Because the values of the parameters are not expected to have instantaneous effects on the home values, a time lag in the home sales values was introduced as a parameter. That is, the mean home sales values in year 1988, for example, are regressed against the values of the independent variables of 1987, yielding a time lag of 1 year, 1986, a time lag of 2 years, 1985, a time lag of 3 years, and 1984, a time lag of 4 years, as well as against the values in 1988.
Because the values of each of the independent variables have different units and widely different magnitudes, an intuitive interpretation of the values of the various regression coefficients is difficult. For example, [edited]. By normalizing the values of the independent variables by their respective 1984 values, the independent variables will be pure numbers of the same order of magnitude. The resulting regression coefficients will all have units of dollars (the unit of the dependent variable), thereby allowing their straightforward interpretation. Of course, the significance of a given independent variable is indicated by the magnitude of its regression coefficient compared to the magnitude of its standard error, and simply scaling the values of the independent variables will not change this significance.
Numerous regressions with various combinations of independent variables were performed. The more independent variables employed, the larger the standard errors. Accordingly, these regressions were performed with an eye to eliminating those independent variables on which the mean sales values had insignificant functional dependence. Also, because the [edited], regressions including the former are considered redundant and are not presented here.
It was found that of the independent variables, no variation was found with [edited]. The results presented here accordingly exclude analyses employing those independent variables.
The results are given as the coefficients of the least squares fit,
hi = ai + bi x1 + ci x2 + di x3 + ei x4 + fi x5
+ gi x6,
where hi is the annual mean home sales value for [edited], and
x1 = time in years,
Tables 1 -5 present the values for the coefficients based on a time lag of 0, 1, 2, 3, and 4 years, respectively, [edited]. The standard errors in the variables are given in parentheses. The right-hand columns provide the standard error of each regression. Table 6 presents the same results for [edited].
[edited] provide the annual home sales values as a function of time and the values predicted by the regression curve for the analysis based on a 1 year time lag. Figure 19 presents the same results [edited]. Figure 20 presents [edited]. Figure 21 shows the number of total number of homes sold as a function of time.
In addition, histograms of the regression coefficients are presented. Figures 22 - 26 present the distribution of coefficients bi, ci, di, ei, fi, and gi for time lags of 0, 1, 2, 3, and 4 years, respectively. The error bars represent the rms of the standard errors of the regression coefficients [edited].
Comparing Tables 1, 2, 3, 4, and 5 indicates that the dependences on the independent variables become strongest after a time lag of 2 years, and weaken thereafter. Figures 22 - 26 and Figures 27 - 45 demonstrate this "elasticity." In the latter, the absolute values of the regression coefficients of each independent variable are plotted as a function of time lag, for time lags of 0 through 4 years. These plots generally show a peak near 2 years, showing that the median home values are most sensitive to the independent variables after a time lag of 2 years. This implies that the effect of a given independent variable on home values can be reversed on a similar time scale.
As with any regression analysis, the functional dependences between the dependent variable and any given independent variable need not indicate a causal relationship. A hidden variable may be the cause of the variation. Interpretation of these results is therefore subjective, and is presented in an accompanying article.
This research was supported by generous grants from the Byron and Tweetie Duffy Trust, and the Illinois Taxpayer Education Foundation.