Sean Miner and Faculty Mentor: Quinn Galbraith, HBLL, Family Life and Sociology

Since 1980, the Associate of Research Libraries (ARL) has annually published a comprehensive salary survey that provides useful information on librarian salaries, separating them by institution, personal characteristics, and other factors. At the beginning of each of the surveys, ARL has documented many trends occurring in the United States and Canadian research libraries. Concerning the trend of the gender salary gap, ARL mentions the following:

“In keeping with previous years, the 2013-2014 data show that salaries for women in US ARL university libraries have not yet met parity with that of men. In 2013-2014 the overall salary for women was 96.3% of that of men for the 115 ARL university libraries (compared to 96.22% in 2011-2012). This suggests a slow, long-term trend towards closure of the gender gap in ARL libraries”

We wanted to explore the reasons behind the gender wage gap at research libraries. We realize there may be many legitimate reasons why there is a difference in pay, including years of experience, education, the specific position that a man or woman chooses to work, and whether a librarian works part-time or full-time. We could then discover how much of the gap is due to reasonable explanatory factors. Other factors may or may not be considered discrimination. We also wanted to discover if there was a gap between minorities and non-minorities and do a similar analysis if there was a gap.

We were given special permission to use the salary survey published annually by ARL. The relevant variables that we used from the salary survey include: salary, job, institution, sex, race, years of experience, rank, and a percent of full-time worked. Some of the variables that we thought might be useful in explaining the salary gap were institution, job, years of experience, and percent of full-time worked.

The particular institution at which a given librarian works can explain a great deal of variation in salaries since each institution will pay according to its associated cost of living. Also, specific institutions might have other reasons why they pay more or less than other institutions that we were unaware of.

Different types of jobs within libraries would be expected to have different levels of pay since there would be different levels of skill associated with each job. Also, before acquiring the data, we examined prior ARL salary survey publications and noticed that there were unequal demographic distributions of workers in each of the type of job categories that are used by the ARL. For example, there seemed to be a higher proportion of men in jobs that were more technical. Because of these unequal proportions across different job types, we felt that the job type might be a useful variable in explaining the salary gap.

The other two obvious variables that might explain the salary gap included years of experience and percent of full-time worked. It would seem likely that the years of experience variable might be a likely reason for a portion of the salary gap between minorities and non-minorities: non-minorities in the United States had an average of 2 more years of experience than minorities.

Linear regression was used to analyze the data and a step by step approach was used. The demographic variables (gender & race) were first used as the sole explanatory variable. This would essentially describe the raw wage gap. On average, we found that male librarians were paid about $3000 more than females, which was about a 3.5% difference. Comparing minorities to non-minorities, the salary gap was about the same, around $3000. After this explanatory variables were added to the model. This would allow us to see an adjusted salary gap. Essentially it would allow us to see how much we would expect one demographic to be paid above another demographic holding certain variables constant such as years of experience, type of job, etc.

For analyzing gender, we first added both the years of experience and the percent of full-time worked variables to the model. After adding these variables in the adjusted salary gap became about 3% instead of 3.5%. We then added regional factors, such as which school they worked at, and whether they worked at a law or medical library (these types of libraries have different salary structures). After adding these factors, the adjusted salary gap actually increased to the original gap, around 3.5%. The last factor we added to the model was type of job. After adding this the gap decreased quite a bit, to around 2.2%. This is the final adjustment we made. Therefore, we determined that about 40% of the gender salary gap in the library workforce were due to factors such as years of experience, full-time status, regional factors, and the type of job worked.

For analyzing race, we used the same pattern as analyzing gender. However, the first model included gender and race. After adjusting for gender, there was about a 1.7% difference in salaries between minorities and non-minorities, with non-minorities having the higher salary. After adjusting for experience and part-time status, there was only a 0.8% difference in salaries. This reduction in the salary gap was what we had expected. Next, we added on regional factors which reduced the salary gap to 0.5%. Lastly we added job type, which actually reduced to gap to -0.8%. In other words, adjusting for years of experience, full-time status, regional factors, and job type, we would expect minorities to be paid about 0.8% more than non-minorities.

We also expanded the process to each race, rather than minority and non-minority. However, the smaller sample size for some of the races decreased the power of this study. When doing this, the adjusted salary gap against Caucasians closed to zero for almost every race. The only exception was Hispanics, where the salary gap was about 3.2% even after adjusting for the factors discussed above.

Overall, the salary gaps that have been discussed in the libraries can partially be explained by factors such as years of experience, part-time status, regional factors, and the type of job within the library. However, there is still a significant portion of the salary for female workers that is not explained. Also, salaries for Hispanics compared against Caucasians were still significantly lower, even after adjusting for other explanatory factors. While there are still unexplained salary differences among these groups, it cannot be concluded that this is due to overt discrimination. There could be other explanatory factors not included in this data that would either decrease or perhaps increase the gap.