Estimating Wage Differentials for the Western Region
Using the March Current Population Survey (CPS) Supplement
by: David Bullard, Economist
" ... we expect a high school graduate on average to earn $5,391.77 more per year than a high school dropout."
There is considerable variation between the wages that workers earn. While this variation is most commonly attributed to the workers’ different occupations and industries, other factors, such as gender, race, level of education and age also play a role. This article uses the March supplement to the Current Population Survey (CPS) to estimate wage differentials for these characteristics for full-time, year round employees in private business.
Our approach to the analysis of wage differentials is modeled after a similar study of wage differentials in Arkansas(1). This article shows that education is a positive and highly significant predictor of earned income. Age (as a proxy for experience) and the hours worked are also positive and significant. The model suggests that women earn about $7,000 less per year than men with similar education working in the same occupational group and major industry. While earnings for Blacks and Asians are not significantly different than Whites in this region, American Indians and Alaskan Natives (Aleuts) earn about $2,600 less than workers of other races. Workers in the Mountain states and West North Central states tend to earn less than their counterparts in the Pacific region. Certain groups of occupations (Managerial and Professional Specialty) pay more than others while certain industries (most notably Mining and Transportation, Communications, & Public Utilities) pay more as well.
The March CPS Supplement asks respondents detailed questions about their work and income during the previous year in addition to the standard CPS questions about demographics, education, labor market activity, etc. It provides a rich source of information on the income and work activities of a large sample of the civilian noninstitutional population. The data file used for this analysis was downloaded from the Data FERRET (Federal Electronic Research and Review Extraction Tool), an Internet site maintained by the Census Bureau at http://ferret.bls.census.gov/.
The purpose of this article is to estimate the wage differentials paid to workers for different characteristics, such as gender, educational attainment, occupation, etc. Originally, this analysis was attempted using the 1998 CPS March supplement for Wyoming; however, the small sample size (n=313 workers) made analysis difficult and it was decided to use the 20 Western states of the Mountain, Pacific and West North Central Census Divisions(2). This gives a sample size of 9,471.
Economic Theory of Wage Differentials
Economic theory beginning with Adam Smith suggests that wage differentials will be primarily determined by differences in occupations(3). Smith’s theory states that wages will adjust so that the labor market for a particular occupation will be in equilibrium. He noted that occupations have many different characteristics; some are pleasant and don’t require unusual physical activity or long hours, while others require workers to work long hours doing heavy labor. They also vary by the preparation for entry into the occupation. Some require long periods of education or training, while others can be learned in a few minutes. Smith suggests that wages will adjust so that each occupation will have enough workers. Thus, unpleasant occupations will pay higher wages (ceteris paribus, i.e., all other things being equal) than pleasant occupations and occupations which require many years of education will pay more than occupations without such requirements.
A second, more recently developed theory applicable to wage differentials is the human capital theory of Gary Becker. In a nutshell, Becker’s theory is that people invest in "human capital" through education, and by increasing their skills, they make their labor more valuable. Thus, we normally expect workers with higher levels of education to be more productive and receive higher wages.
A competing theory about the value of education is the screening or signaling model. It hypothesizes that education doesn’t really teach anything or give people better skills, but it sorts out the most productive workers. It is assumed that since college is easier for smarter people, it costs them less effort and they are more likely to get a degree. Employers are looking for the smartest, most productive workers, and those who have passed the screening test of school are the ones for which they are looking. This theory explains why those who have graduated from college earn so much more than those who have had some college but did not obtain degrees.
Regression Analysis
As stated in the introduction, this analysis only involves subjects who worked full-time (35 or more hours) year round (50-52 weeks) during 1997. The sample is further limited to those who worked in private business and excludes most self-employed individuals. This analysis only includes workers from age 22 to 59, which are recognized as prime working years. The few subjects with earnings below $10,000 or above $100,000 are excluded as outliers.
The Table summarizes the results of the regression model for wages. The adjusted R2 is 0.369 suggesting that this model explains about 37 percent of the variation in wages. Throughout the discussion that follows, additional variables will be suggested that could account for some of the other 63 percent of variation. The dependent variable(4) explained by the regression model is wage earnings in 1997 expressed in dollars.
Age is included in the model as a proxy for experience. This is one of the two quantitative independent variables(5) included in the model. The coefficient is positive and highly significant. The coefficient suggests that on average, workers earn $332.09 for each year of age.
The second independent variable is usual hours of work. Only full-time (35 hours and up) workers are included, so the minimum value for this variable is 35, but the mean is 43.9 and the maximum is top coded at 99. The coefficient is positive and highly significant, suggesting that those usually working more hours earn more income.
Education
Respondents are classified according to the highest level of education completed. All categories are positive and highly significant. As shown in the Table, the coefficients for wages generally increase as the level of education increases. The interpretation of these coefficients is the difference in annual earnings of a person having this level of education compared to someone with less than a high school diploma. For example, we expect a high school graduate on average to earn $5,391.77 more per year than a high school dropout.
It should be noted that the two associate degree categories have different coefficients. This suggests that workers who hold an associate degree in an occupational or vocational program are earning $1,875 per year more than their counterparts who hold associate degrees in an academic program. One possible explanation is that since academic associate degrees are primarily intended to prepare people to enter four-year degree programs, having this as one’s highest degree suggests that one’s education is still not complete. In contrast, associate degree programs in occupational or vocational subjects give students training in a specific occupation and are seen as a terminal degree.
Occupation
All of the occupational dummy variables(6) are interpreted as the difference between the occupation in question and the occupations not included in the regression equation. These generally low-paying occupations not included are:
All of the estimated occupations have positive coefficients. That is, they pay more than the non-estimated occupations as a group. Executive, Administrative and Managerial Occupations pay the highest, followed by Professional Specialty Occupations. The occupational wage differentials generally reflect the amount of skill required in the occupation.
Industry
For various economic reasons, some industries pay higher wages across occupations than others. The omitted industries are:
Many of these reasons are related to Adam Smith’s theory of occupational wage differentials. For example, much of the work in the Construction industry is short-term, lasting only a few months. Therefore, construction workers are paid a premium for the short-term nature of their work and the high probability that they will be laid off.
The capital intensity of industries also accounts for some of the difference in wages. In cases where the capital to labor ratio is high, such as Mining, workers are more productive(7) and paid higher wages. In certain Retail and Service industries, the capital to labor ratio might be quite low, making workers less productive and giving them lower wages.
In addition, the economic effects of competition also keep wages low in certain industries, while allowing others such as Transportation, Communications, & Public Utilities (TCPU) to maintain monopoly power and increase profits and wages.
Gender
In an effort to identify the effect of gender on wages, a dummy variable for females is introduced into the model. This variable is negative and highly significant. It suggests that holding all other modeled factors constant, on average females earn $7,015 less than males in the 20 Western states. This is lower than other estimations, indicating that the male-female wage gap may be getting smaller. The Arkansas model of wage differentials based on the March 1993 CPS supplement estimated the female wage differential at $13,318. Other comparisons between male and female average wages have ignored education, hours worked, occupation and other important variables included here(8).
Several possible explanations exist for the difference between the earnings of men and women. First, the occupational groups included in this model are broad. An analysis using detailed occupations would probably have different results; certain occupations are dominated by either males or females. Estimating wage differentials for detailed occupations(9) would require a larger sample and more detailed information.
Historically, men and women have had different levels of attachment to the labor force(10), with women generally being less attached, using age as a proxy for experience can introduce some serious errors into the model. For example, comparing a 50-year-old man’s wages with a 50-year-old woman’s wages would be difficult if the man had been working steadily for 30 years but the woman had not.
Although educational attainment is included in the model, the subject or field of study is not included. Just as there are male-dominated occupations and female-dominated occupations, some majors tend to attract mostly men or mostly women. Average earnings vary widely by field of study(11).
Race
Interestingly, for the regions being covered by this analysis, wages are not significantly different for Blacks or Asians than Whites. While this does not necessarily mean that minorities aren’t subject to discrimination in the workplace, or that they don’t earn less than their majority counterparts, it does suggest that for these similarly situated minority workers (age, gender, education, occupational group, industry and hours of work), the difference isn’t large enough to be statistically significant.
However, using a 95 percent confidence level, American Indians do earn significantly less than other workers. The Table shows that this is the only variable not significant at the 99 percent level. Since many American Indians live on reservations where employment opportunities are limited, this result is not surprising.
Region
As previously mentioned, this study only includes 20 Western states. Earnings are significantly lower in the Mountain and West North Central divisions than in the Pacific division. Although there are no reliable figures which allow interstate cost of living comparisons(12), the author believes that some of the difference in wages may be accounted for by lower cost of living in Mountain and West North Central states than those on the West Coast.
Other regional differences could be related to detailed industry patterns. The industry groups used in the model are broad like the occupational groups. Since Pacific states have a different mix of industries (even within an industry group such as Durable Goods Manufacturing), they have different earnings patterns than Mountain or West North Central states.
The power of unions could also affect regional wages. None of the Pacific states have right-to-work laws, while more than half of the Mountain and West North Central states do. These laws tend to decrease the bargaining power of unions and limit their ability to increase their members' wages.
Conclusion
This analysis shows that education is a positive and highly significant predictor of earned income. Age (as a proxy for experience) and the hours worked are also positive and significant. The model suggests that women earn about $7,000 less per year than men with similar education working in the same occupational group and major industry. While earnings for Blacks and Asians are not significantly different than Whites in this region, American Indians and Aleuts earn about $2,600 less than workers of other races. Workers in the Mountain states and West North Central states tend to earn less than their counterparts in the Pacific region.
1 Christy Rollow and Bill Seyfried, "An Analysis of the Determinants of Wage Differentials in the State of Arkansas," Arkansas Business and Economic Review, Summer 1995.
2 The Census Divisions and the states they include are:
3 Occupation is defined as "a group of similar jobs found in different industries or organizations." An example of an occupation is accountant. Accountants are found in practically all industries. The Current Population Survey (CPS) uses the Standard Occupational Classification (SOC) scheme to classify occupations. The SOC classifies occupations based on the type of work performed.
4 The dependent variable is the variable that the model explains. For example, this model attempts to explain the annual wage earnings of full time year-round workers in private industry in 20 Western states.
5 The independent variables are the explanation for the level or change in the dependent variable. In this model, educational level is an independent variable that helps explain the level of wages.
6 A dummy variable is a qualitative independent variable used in multiple regression analysis. It is thus distinguished because the usual multiple regression analysis uses quantitative data for both independent and dependent variables.
7 Labor productivity is defined as output divided by the hours of labor input. Capital can make labor more productive. For example, a person using a computer can do much more work in an hour than the same person with only paper and a pencil. Similarly, a person using a backhoe, or earth moving equipment can move more dirt in an hour than the same person using a shovel.
8 Mary Beth O’Loughlin, "Gender Tenure and Wages," Wyoming Labor Force Trends, August 1997; Gregg Detweiler and Brett Judd, "The Relation of Age and Gender to Employment in Wyoming: Parts One and Two," Wyoming Labor Force Trends, May and June 1996.
9 This model uses 13 occupational classifications. A detailed occupational breakout (for example the Wyoming Wage Survey) includes several hundred different occupations.
10 Attachment to the labor force refers to how much and how often a person works for pay. A person who worked every week in the past year, or every week for the past several years is said to be more attached than someone who only works sporadically. Brett Judd found that in 1996, 16 percent of all Wyoming workers only worked one quarter, "The Wyoming Wage Record Classification System," Wyoming Labor Force Trends, March 1998.
11 Daniel E. Hecker, "Earnings of College Graduates: Women Compared With Men," Monthly Labor Review, March 1998; Rosalind R. Bruno, What’s It Worth? Field of Training and Economic Status: 1993, U.S. Census Bureau.
12 For an explanation of the difficulty of interstate cost of living comparisons see Gayle C. Edlin, "Selected Determinants of Elementary and Secondary Teachers’ Wages," Wyoming Labor Force Trends, September 1998.
West North Central: Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota
Mountain: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming
Pacific: Alaska, California, Hawaii, Oregon, Washington
Afterword
The preceding article shows that our labor market outcomes (specifically wages) are a function of both variables that we control and circumstances that we don’t control. Examples of variables we control include education, occupation and industry. Circumstances that we don’t control would be gender, race, age and the occupations and industries in our local area. Other variables included in the model are a combination of things we sometimes control and sometimes don’t. For example, some workers can influence their usual hours of work (perhaps by choosing a certain occupation or industry), while others cannot. The regional variable suggests that workers in the Mountain or West North Central Regions will earn less than similarly situated workers in the Pacific Region. While workers are free to move from one state to another, choosing to live in the Mountain Region usually means accepting lower wages than could be earned on the West Coast.
Table of Contents | Labor Market Information | Employment Resources | Send Us Mail
These pages designed by Gayle C. Edlin.
Last modified on
by
Valerie A. Davis.