Socioeconomic Scale Development

Standardized scores are calculated for each component score prior to combining them into a single scale. Standardized scores, often referred to as "Z" scores, indicate the number of standard deviations above or below the mean where a particular observation falls. They are calculated by dividing the difference between a particular observation and the mean by the standard deviation. Standardization facilitates comparison of scores from distributions. To ensure that outlier values do not have undue influence over the distribution range of any score, each standardized score is then normalized to a base of 100, using 2 standard deviations as reference points. Each of the individual component scores are combined into a single socioeconomic scale which is also normalized to a base of 100.

X = Sum S^1–6 [ ((S / Z) - A) * (100 / (B - A)) ]

                5

     where: X = Socioeconomic scale

     S = Standardized scores of each of 5 scale components.

     Z = 2, if S is persons in poverty or poverty intensity; otherwise, 1.

     A = - 2 (2 standard deviations below the mean).

     B = 2 (2 standard deviations above the mean).

The socioeconomic scores are reported on a seven-point categorical scale with a 1 indicating the lowest socioeconomic score and a 7 indicating the highest. The ordinal scale is derived from the continuous scores, divided into categories based on the number of standard deviations from the mean of the scale. A rating of 1 indicates a very low socioeconomic score and includes those scores at least two or more standard deviations below the mean (i.e. standard deviation => -2); 2-low (standard deviation > -2 and <= -1); 3-medium-low (standard deviation > -1 and <= -0.5); 4-medium (standard deviation > -0.5 and < 0.5); 5-medium-high (standard deviation >= 0.5 and < 1); 6-high (standard deviation >= 1 and < 2); 7-very high (standard deviation >= 2).

While income is a commonly used indicator of socioeconomic status and well-being, it is not included in the socioeconomic scale for two reasons: (1) most of the variables in the scale are closely correlated with income; and (2) income measures available from the Census data are problematic. The Bureau of the Census reports income in a variety of tables and formats. Comparisons of census-reported aggregate income with other income sources indicate that census-reported income is considerably underestimated, particularly for interest and dividend, and public assistance income (Stewart, 1996). Adjustments can be made to compensate for these discrepancies, but they can only be applied to aggregate income, making many of the census income tables, which report income within finite categories, unusable. Due to the level of variation in interest and dividend and public assistance income between block group aggregations, these correction factors have a significant effect on the ordering of relative income between aggregations. As an additional complication, analysis of aggregate income in previous studies (see Doak and Kusel, 1996) indicates that pockets of extraordinarily high income households can significantly distort the real distribution of income within aggregations making average income measures — whether household, family or per capita — inappropriate as relative indicators of the socioeconomic status of individuals within a particular area. Average income measures are best suited only as a means for expressing total income in an area in relative terms.

Since aggregate income, or average income derived from aggregate data, is the only measure that can be effectively adjusted for discrepancies involving interest and dividend and public assistance income, and since average income appears to severely distort the relative ranks of aggregation by actual income, direct income measures were not included in the socioeconomic scale.