Claudia Goldin and Robert Margo, in their 1992 research of ...



A review of the research:

The Great Wage Compression: The US Wage Structure at Mid-Century

By: Claudia Goldin and Robert Margo

Lorie B. Owen

ECON 467

May 2, 2008

Claudia Goldin and Robert Margo, in their 1992 research of The Great Compression: the Wage Structure in the United States at Mid Century, attempt to identify specific causes and effects of the narrowing of the wage structure following the Great Depression. The 1940’s – 1960’s proved to be one of the narrowest wage structures in U.S. history. Subsequently, 1970’s reversed the narrowing and began the expansion of the wage structure.

Using a data sample of white males, 18 to 64 years of age, working more than 39 weeks per year, was used to calculate the variance of the wage structure. The log function was used in the regression because the data set does not follow a linear pattern. Analyzing the difference between the 90th percentile and 10th percentile over the 1940-1960 periods, Goldin and Margo determined that in fact the wage structure was narrower following the Great Depression. To further inspect this hypothesis they calculated the variance of the log of weekly wages, over each period. Their results supported the hypothesis that a compression existed in the wage structure.

|Variance of the log of weekly wages |

|1940 |0.325 |

|1950 |0.259 |

|1960 |0.275 |

In order to effectively isolate specific factors of interest they compared data for education, skills and demographic differences. The 1940 – 1960 data sets were taken from the Public-Use Microdata Sample (PUMS) from the U.S. Census Bureau. Most of the data is for only white men; however it was broaden under certain circumstances in order to compare this study with more recent studies. Education was broken down into three categories non high-school graduates, high-school graduates and some college. Even though most individuals during this time frame did not complete college, it is included as a means of comparison with future studies.

The ratios of weekly wages among educational categories show that the ratio between college graduates and high school graduate was the highest in 1940, with high school graduate to eighth grade closely following. This indicates that the highest jump in wages occurred between high school and college in 1940. However, also in 1940 the jump between eight grade and high school is larger than the increase in wages between high school and some college. This analysis was also compared using a range of experience levels from 1-40 years. This portion of the ratio investigation indicted that for most education categories that an increase in experience decreased the wage gap.

Occupations were also isolated for means of comparison. Looking at white collar, blue collar, professional, clerical, craft, operative, laborer and service professions compared to all non-farm wages on a weekly basis, to reveal that over the 1940 to 1960 period the greatest wages were seen in white collar and professional occupations. The authors acknowledge that breaking occupations down by experience level would also present a great insight to the analysis process; however they do not include that breakdown in this research.

Furthermore, Goldin and Margo’s research looks into the decomposition of the wage structure. By synthesizing the variance discovered in the regression, they found that “about 30 percent of the decrease in the variance in the log of weekly wages was due to a decrease in the variance of the residuals and 70 percent to a decrease in the variance of the quantities and market prices of observed covariates” (Goldin and Margo, 1992 p.12). Simply put this means that the variance or deviation of the data set can be attributed to the actual data points and to other market factors changing simultaneously. Although this draws some general conclusions the model by Juhn, Murphy and Pierce in 1989 looks more in depth at the decomposition.

Yit=Xitβt+µit; µ=Ft-1(θit|Xit)

Where Yit equals log of weekly wages, i equals individual, t equals year, and Xit equals individual characteristics. Using this model they sorted factors into three categories; changes in quantities of skills, changes in the prices of skills, and changes in unobservable factors. One error of this model is that it may show sensitivity to the order in which these factors change. This model reveals that changes in prices outweigh the changes in quantity, utilizing the 1940 base year. Operating under the 1950 base year, residuals are no longer significant in effects.

World events and factors to were considered to play a vital role in the compression of the wage structure. Specific effects from WWII, which were precise to this time frame, as well as effects of the Great Depression, a decline in skill levels were considered. Goldin and Margo suggest that because “during the Great Depression was disproportionately borne by the least educated and the lesser skilled” (Goldin and Margo, 1992 p.16). This would suggest that wage inequality would have increased. However, in 1933 the National Industrial Recovery Act (NIRA) was instituted which implemented the minimum wage law, which played some but not a large role on the distribution and magnitude of the wage inequality. Furthermore, they suggest that the command economy, encumbered with educational changes, such as the result of the GI bill, increased the supply of skilled workers relative to the demand for unskilled workers, increasing the wage rate.

Kuznets and Goldsmith in 1953 and 1967 suggest that the narrowing of the wage distribution started prior to 1940, but gained momentum shortly thereafter. For this study they collected income tax data. Most of their focus was the top percentage of the distribution. Because of their inclination it has often been a criticism that their data was biased because of the nature of those in the upper portion of the wage distribution, mainly white collar, highly educated individuals.

Following the criticism of Kuznets and Goldsmith’s studies, Goldin and Margo in their 1991 research considered the differences in white collar returns to educational investments versus those of lower unskilled rankings. The ratio they constructed show some gains in lower skill wages during the early 1920’s, but change shortly after the onset of the Great Depression. However, these gains were only brief in nature. Between 1933 and 1934 these gains reversed. One suggestion of this reversal is that the NIRA interfered in the natural market self correction.

WWII played an extremely large role in the wage compression. During WWII the U.S. fell under a command economy. Because of this regime, along with the implementation of the National War Labor Board (NWLB) price controls were initiated. Hourly wages were only allowed to move within a very finite range for each occupation as designated by the NWLB. Assuming this does have an impact on the wage distribution, the 1940’s would seem bias, because a lag was supposed in the census data set, by one year per decade. If WWII had not have had a significant impact on the wage structure, after the war a reversal to previous wages would have occurred. Because this reversal did not occur for over fifteen years, one can agree with a large degree of confidence that in fact WWII did play a vital role in the compression of wages.

Goldin and Margo’s final conclusions were that the compression of the upper portion of the wage distribution took place during the WWII period. The compression of the lower portion took place in the prewar time frame, and experienced a large decrease in the post war period. Any industry which was primarily female based could not be expected to behave in the same nature as the majority of the trend. This is true because prior to WWII, the labor force had very few female workers. During the war time female labor participation increased exponentially. Therefore this comparison of the pre and post war wage rates proves to be faulty and unable to follow the same trend as the majority of industries.

The primary cause of the Great Wage Compression is the combination of many events. These events because of their short term nature prove to be difficult to quantify pre and post data for means of comparison. Goldin and Margo concluded that all of the above mentioned causes did play a role in the Great Wage Compression.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download