THE WOMEN OF WORLD WAR II

On December 7, 1941, Japan attacked Pearl Harbor, thrusting the United States into a two-front war against Japan, Germany, and Italy. The war ultimately would require the service of 16 million Americans. To meet the labor demands of the war, the US government relied on a selective conscription scheme, first implemented in 1940. Additionally, the US Congress established the all-volunteer WACs in May 1942 to supplement the number of noncombat military personnel and additional auxiliary corps for the other services (Bellafaire Reference Bellafaire1993). Ultimately, almost 350,000 women would serve in the armed forces, including 133,000 WACs (figure 1). Women were critical for the war effort: female volunteers in the US Army drove and repaired military vehicles, served as clerks, worked in office jobs, and conducted intelligence collection, among other roles (Bellafaire Reference Bellafaire1993).

Figure 1 Female Volunteer Rates by County, 1943

Darker gray indicates lower volunteerism, lighter gray indicates higher volunteerism.

THE INFLUENCES ON FEMALE VOLUNTEERISM

We assume that individuals are rational when choosing to volunteer for military service. They weigh the costs and benefits of military service, maximizing their utility. The costs and benefits of service come from both internal and external motivators (Moskos Reference Moskos1977). Internal motivators include patriotism, religious beliefs, wages, a sense of duty, and educational opportunities. External motivators include structural forces such as military regulations that (dis)encourage service for certain demographic groups. For example, the US military limited the opportunities available to Black women during World War II to menial labor (Bolzenius Reference Bolzenius2018). External motivators also include social pressures that push people toward or pull them away from military service. For example, conservative communities promoted traditional gender norms and disapproved of women joining the military (Permeswaran Reference Permeswaran2008). The following discussion theorizes several of the costs and benefits that influenced women’s decisions to volunteer.

The first factor that shaped female volunteerism was the societal stigma around women serving in the military. Women feared the social sanctions from their communities—family and friends, political elites, and mass media—if they joined the military (Permeswaran Reference Permeswaran2008). We argue that these social sanctions were most potent in culturally conservative communities that observed strict gender norms. In these communities, women were encouraged to stay in the household and discouraged from engaging in nontraditional activities, including educational attainment and labor-force participation (Glass and Jacobs Reference Glass and Jacobs2005).

In a prewar context, an appropriate proxy for the intensity with which a community held these traditional gender norms was the propensity for its women to obtain higher education. Culturally conservative values are associated with lower female educational attainment (Glass and Jacobs Reference Glass and Jacobs2005). Many conservative communities perceived female higher education as violating “the ideal of true womanhood” (Graham Reference Graham1978, 759). As a result, communities that were more comfortable with women pursuing higher education would be more willing to accept women assuming other nontraditional roles, such as military service.

If a woman’s surrounding community objected to her enlisting, she likely would have felt conflicted about doing so. For women in progressive communities, volunteering was far less costly. Therefore, communities with more female college graduates would observe more volunteerism.

The second factor that shaped volunteerism was racial discrimination. Volunteering for the military was extremely costly for Black women, who had to overcome obstacles due to both her gender and her race. Blacks comprised only 1% of local draft board members during World War II, leaving Black women’s enlistment primarily in the hands of white men—who often did not believe that they belonged in the military (Murray Reference Murray1971).

Many communities adopted recruitment rules that were responsible for gatekeeping potential volunteers on the basis of race. Although the military had no educational requirements, it eventually implemented literacy requirements—through the Armed Forces Qualifications Test (AFQT)—that adversely affected Black women. As a result of racial disparities in the US public educational system, Black women scored significantly lower than white women on the AFQT, particularly in segregated Southern communities (Moore Reference Moore1991). When Black women did pass the AFQT, local draft boards often delayed their enlistment by several months, claiming a lack of adequate training facilities (Murray Reference Murray1971).

All of these challenges make it unsurprising that Black women never comprised more than 4% of the WACs. For Black women, attempting to enlist was extremely costly. Therefore, communities with more racial strife would observe less volunteerism.

The third factor that we examined was economic opportunity. During World War II, labor shortages led to significant increases in civilian wages (Schumann Reference Schumann2003). Increased wages at home primarily benefited men (Aldrich Reference Aldrich1989). Most women who entered the labor market did not take over the jobs of enlisted men or the jobs that men who stayed home wanted. Instead, most women competed for newly created jobs (Rose Reference Rose2018). These jobs were geographically concentrated, temporary, and lower paid; in 1942, a woman working in a civilian job earned 53 cents to a man’s dollar (Aldrich Reference Aldrich1989).

In 1943, Congress passed an equal-pay provision for the WACs. Men and women from all racial backgrounds received equal pay according to rank and, with a few modifications for WACs, almost the same benefits (Bolzenius Reference Bolzenius2018). For many women, these regulations provided opportunities that were unimaginable at home. Increased pay and benefits may have partially ameliorated the high societal costs described previously.

We anticipated that as wages rose in a county, an influx of men seeking higher wages would drive the local women to take lower-paid jobs; thus, joining the WACs became more attractive. We expected women in counties with higher wages to face more costs in the civilian workforce and to view military service as more desirable.

DATA AND METHODS

To test our hypotheses, we analyzed a novel dataset of all female volunteers in the US Army during World War II (1941–1945), compiled by the US Department of War and converted into the Soldiers of World War II dataset.Footnote ¹ Workers at local induction stations took the data from enlistees and entered this information on punch cards for use on early computers. Decades later, researchers at the National Archives and Records Administration combined these records into ASCII files. We cleaned these files and converted them into a format usable for analysis. This dataset contains every enlistee’s date of birth, state and county of residence, enlistment date, prewar occupation, education level, race, and ancestry.

To address the immediate research questions in our study, we created another subset of our enlistments to women who served. Although the records do not indicate the gender of the enlistee, we could obtain this information using the enlistees’ serial numbers. Female enlistees had an “A” placed before their serial number to indicate that they were a member of the WACs. In total, we had access to more than nine million service-member records, of which 133,843 were women in the WACs.Footnote ² Ideally, we would have compared predictors of volunteer rates at an individual level; however, crucial granular, individual-level partisan, demographic, and economic data do not exist. Therefore, we aggregated information to the county-year, resulting in a balanced panel of 15,580 observations for 1941–1945. This approach allowed us to compare the influence of various county-level attributes on that county’s female volunteer rate.

Dependent Variables

Our first outcome measure was the proportion of female volunteers in a county-year. We created additional subsets for white volunteers and Black volunteers.Footnote ³ Figure 2 presents the number of volunteers by year.

Figure 2 Female Volunteers, 1941–1945

Independent Variables

Consistent with the Education Hypothesis, women from counties with higher rates of female college education would have higher enlistment rates. The distribution of college-educated women by county is shown in figure 3. Consistent with the Race Hypothesis, areas with more overt racial segregation should have had lower volunteerism rates, particularly Black volunteers. We operationalized county-level racial segregation with the two reliable measures of racial segregation used by Logan and Parman (Reference Logan and Parman2017). The first is the isolation index, which measures the probability that a member in one group (whites) will interact with a member of another group (Blacks). The isolation index ranges from 0 to 1, in which higher values indicate lower segregation. The distribution of neighborhood isolation is illustrated in figure 4. The second measure is the dissimilarity index, which measures the percentage of a group’s population that would need to change residence for each neighborhood to have the same percentage of that group. The dissimilarity index also ranges from 0 to 1, in which higher scores indicate higher segregation. To test our Wages Hypothesis, we included county-level information on economic indicators, including wages in the four major US Census sectors: manufacturing, retail, wholesale, and service (Dodd and Dodd Reference Dodd and Dodd1973).

Figure 3 Female College-Graduation Rates by County, 1940

Darker gray indicates fewer graduates, lighter gray indicates more graduates.

Figure 4 Racial-Isolation Rates by County, 1940

Darker gray indicates higher segregation, lighter gray indicates lower segregation.

We controlled for two partisan indicators: county-level vote shares for President Roosevelt and the Democratic candidate for the US House of Representatives (Haines Reference Haines2005). We also included county-level demographic data, such as the percentage of white women and men, the percentage of Black women and men older than 21,Footnote ⁴ the proportion of the urban population, and the total population.Footnote ⁵

To analyze these data, we used ordinary-least-squares (OLS) regressions with state-year fixed-effects and clustered standard errors.Footnote ⁶ The first model was all female volunteers, the second model was all white female volunteers, and the third model was all Black female volunteers.Footnote ⁷ In reporting our analyses, we show estimated coefficients on separate figures—although they originated from the same models—to highlight results by hypothesis.Footnote ⁸