by Anna Giannuzzi, Harvard College ‘20
ABSTRACT: Since the collapse of the Argentine economy in 2002 the country has experienced a severe deceleration in female labor force participation. While the recent leveling of the female labor force participation rate may suggest a natural rate has been reached, trends from the microdata suggest this is not the case. Since 2003, there has been an evident shift in the compositional makeup of the female working age population. The number of low educated women in the female labor force has fallen greatly as the number of middle and highly educated women continues to grow. Additionally, between these three groups, labor force participation increases drastically with greater educational attainment. Therefore, as women become more educated and transition to formal labor markets the labor force should approach the participation rates of the middle and high educated population. In addition, cross-section analysis of the wage elasticity of the labor supply of married women since 2003 reveals the growing dominance of the substitution effect over income effect for this group. This attitudinal shift of the married female population could prove critical to the continued growth of a resilient female labor force.
Female labor force participation has been one of the most significant forces propelling economies around the world in the past century. In Argentina, the rise of the female labor force participation rate in the 1990’s coincided with GDP growth that was sustained until the collapse of the Argentine economy, which culminated in a default on the country’s public debt in December 2001 (figure 1, figure 2). According to the Argentine government’s national estimates, the female labor force rose from 35.2% in 1992 to 50.31% in 2003. However, after the economy’s GDP quickly recovered from the economic crisis the rapid rise in the female labor force abruptly came to a halt. In fact, it currently has a lower female labor force participation rate than most other Latin American-Caribbean countries and has the second lowest among peer countries (Gasparini & Marchionni, 2017; Mateo Diaz & Rodriguez-Chamussy, 2016). What is puzzling about the deceleration is that precipitators of increased female participation, such as a fall in the mean number of children per household and a rise in female education levels, were strong during this time (figure 3, figure 4).1 In this paper, I seek to address whether or not a resurgence of Argentina’s female labor force participation rate is in sight. By turning to household survey data I break down changes in the female labor force since 2003. First, I analyze the female labor force by levels of educational attainment. Data reveals labor force participation is substantially higher amongst higher educated groups and while these groups are growing in size, low educated female workers still comprise the majority of the working age population. Next, I dissect labor force participation trends and identify a relationship between the female labor force participation rate and male unemployment that falls away as educational attainment rises. Lastly, I analyze the elasticity of labor supply of married women to understand how the income effect and substitution effect have manifested since 1 Busso and Romero Fonseca (2015) discuss the effects of these factors in addition to increased technological innovation that helped wives with households chores and therefore made it easier for them to enter the workforce. the economic collapse. The results support the hypothesis that the female labor force participation rate has not reached a natural rate. The country appears to be emerging from a transition period as the labor force of married women shifts from a dominating income effect to a dominating substitution effect. This suggests that the traditional arguments of Cerruti (2000) and Gasparini and Marchionni (2017) are no longer sufficient for explaining female labor force deceleration. Instead, declining real wages throughout the economy since 2011 have hindered rising female labor force participation growth. In fact, in the absence of deteriorating real wages, a rate-of-change supply model predicts a increase of 1.813% in the married female labor supply since 2011. Therefore, falling labor demand appears to be counteracting positive compositional shifts in the female labor force.
Existing theories on the deceleration of the female labor force view changing economic and political settings following the economic default as the key forces counteracting positive compositional changes on the female labor force. The period from 1998-2002, known as the Argentine Great Depression, took a large toll on the country, with the poverty rate reaching 53% during this period (Cibils, Weisbrot, & Kar, 2019). Households were desperate for additional income and by entering the workforce, women could alleviate some of the pain. As economic conditions began to improve after the crash, ample social assistance was also provided under the Kirchner administrations that greatly aided the country’s poor (Garganta, Gasparini, & Marchionni, 2017). The improved financial standing of these households can be viewed as one of the main forces keeping women from joining the labor force. The economic literature supporting this claim is not limited to Argentina as the shift in female labor force trends occurred in many Latin American countries. By looking at the entire region it becomes clear how changing economic status affected female labor force engagement. Gasparini and Marchionni (2017) discuss the deceleration in female labor force participation across the region but state that the decline is particularly noticeable amongst married and “vulnerable” women which they define as “women with low levels of education, living in rural areas, with children, and married to low-earnings partners”. They claim the reason for this decline was an increase in transfer programs to low income households across the region and the improved economic status of their partners. In these cases, women were deterred from entering low quality jobs by their spouses. Similarly, Cerruti (2000) argues that female workers in Argentina in the 1990’s were subject to an “added worker effect” in which other household members join the workforce when primary earners are unemployed.2 This further supports the claim that as the economic standing of families began to improve, less women entered the labor force to compensate for the male head of the household’s lost income. The fall in labor force growth seems to support the theory of a U-shaped relationship between development and female labor force participation put forth by Goldin (1994). The theory claims that there is a correlation between female engagement in the labor market and economic development. In its simplest form, it means that in the poorest countries female labor force participation is high in order to provide additional family income. It then begins to fall in middle income countries as a strong income effect, promoted by a social stigma which looks down upon working in available manual labor jobs, starts to dominate. Furthermore, the theory suggests that eventually a switch occurs as as rising education and falling fertility rates make it easier for women to join the workforce. During this time, the income effect falls away and the substitution effect begins to dominate.3 At this point, women enter more socially acceptable white-collar jobs. Understanding where either a negative income effect or a positive own-substitution effect dominates the married female population can offer crucial information about whether the female labor force participation rate will continue to rise. If women decide to rely on the increased income of their spouses as opposed to deciding to work, then participation rates can expect to remain stagnant. Currently, Argentina sits toward the middle of this theoretical curve and begs the question of whether it will remain here or continue on a trajectory up the curve (figure 5). Discussion on whether or not Argentina’s female labor force has settled in center of this curve and reached its natural rate is limited. Yet, despite the limited attention the matter receives it is an increasingly pressing topic. The female labor force participation rate has stalled in the past decade and in looking at trends of the overall female labor market, no significant change appears to be approaching. Additionally, the country’s current economic situation is dire with inflation rising 43.7% in 2018 and the poverty rate rising six percentage points to 32% in the past six months (INDEC, 2019). Following the Depression, the country instituted an ineffective economic development program with an expanded social safety net that helped the impoverished population while creating an unsustainable cycle (The World Bank, 2018). These changes acted against growth fundamentals such as a strengthening labor force that had been developing prior to the collapse and derailed positive economic development. At the present moment, the country is plagued with low productivity and in need of a boost. If the female labor force were to match the male rate it would increase per capita growth on average by one percentage point per year and contribute to the social security system.4 5 An end to the period of stagnation the female labor force has endured could provide desperately needed assistance to the country’s economy.
The data used in this study come from the Encuesta de Permanentes Hogares Continúa (EPH-C) from the third quarter of 2003 to the second quarter of 2018. The survey is a representative household sample which is gathered using multi-layered and stratified design and is administered quarterly by the Argentina’s Instituto Nacional de Estadísticas y Censos (INDEC). The sample covers 31 large urban areas and represents over 60% of the country’s population. It is divided into two data sets each quarter, one for individuals and another for households. This survey replaced the previous survey used by the INDEC, Encuesta de Permanentes Hogares (EPH). The EPH was used up until the first half of 2003 and differed significantly from the current survey. It was administered to a smaller portion of the population, was only administered twice a year and had a differently structured questionnaire. Therefore, in an effort to maintain consistent labor force estimates without jumps in results due to methodological changes, the data analysis of female labor force trends is limited to 2003-2018. The data covers all quarters in the period with the exception of Q3 2007. Aside from this missing frame, the data set is ideal for analyzing trends in female labor force participation during and following the deceleration period. There are various elements that could confound results such as differing sample size from quarter to quarter, occasional input errors for data entries and biases towards lower income participants due to greater likelihood in responses from these households.
Trends in the Female Labor Force
Analyzing both general and education specific female labor force rates from 2003-2018 reveals informative trends regarding the female working population. Eligible members of the female labor force in this study are women who are over 15, are not students, are not making revenue leasing or renting property, are not retired or receiving a pension and are not disabled. Under this classification, the female labor force participation rate has followed the following trend since the Q3 of 2003 (figure 6). It has been subject to cyclical fluctuations, similar to the male labor force participation (figure 7).
However, the most notable trend is that the rate underwent a general increase until about 2011 when it ceased to rise. This description is in tune with the trends identified with the SEDLAC, National Estimate and ILO trends (figure 2) although the exact year in which the rise seems to end varies by source due to different definitions of labor force and different sample groups. The declining growth of the female labor force participation rate in this period appears worrisome upon first glance. However, it can be understood more thoroughly by breaking down these changes according to educational attainment group. Looking at the composition of the female labor force by low, medium and high educational attainment reveals vastly different trends within each and suggests the deceleration is transient. Individuals are classified as “low educated” if they reported not completing a secondary education on the EPH-C. “Medium educated” are those that completed secondary school but did not graduate from university. “Highly educated” are those that graduated from university. Women who participated in the EPH-C were grouped into one of these categories and used to understand residual changes of the female labor force since Q3 of 2003. The first important result from the data demonstrates how the composition of the female labor force has changed since 2003. Table 1 shows how a rising educational attainment has affected the labor force population. There has been a large decline in low educated workers since the beginning of this period. The percentage of low educated females that comprise the female labor force population fell from from 57.51% in Q3 2003 to 41.46% in Q2 2018 (-16.05%). Over the same period the percentage of medium educated females rose from 28.19% to 37.97% (+9.78%) and highly educated rose from 14.30% to 20.57% (+6.27%). A rise in educational attainment is widely considered one of the key cultural changes leading to an increase in the female labor force participation rate. As education rises, more sophisticated jobs are available to women and they are less likely to look down upon the position they accept.6 However, trends in the overall female labor force participation rate suggest the rising education of the female population does not have a significant impact. The next critical result from this data indicates why this is not the case. The second key result pertains to the labor force participation rate of different educational groups of the female working force population. Table 2 shows drastically different participation rates for each group. As education rises, the probability of participating in labor markets also rises. High educated women exhibit the largest participation rate which remains consistently around 89% and are subject to low variability. Medium educated women also participate at a greater rate than the female average, around 72%, and are subject to more variability than the high educated women. Lastly, low educated women experience by far the lowest participation rate at around 40% and it are the most variant group during the period of analysis. Additionally, the different rates appear to be motivated by different factors. Akin to the findings of both Cerruti (2000) and Gasparini and Marchionni (2017), fluctuations in the low educated female labor force population are correlated with fluctuations in male unemployment (figure 8). During periods of economic instability, as suggested by times when male unemployment is high, low educated females demonstrate the “added worker effect” by joining the labor force to counteract the lost household income of their spouses. Yet, the same relationship does not hold for higher educated groups as no clear correlation exists with male unemployment exists (figure 9). This begins to explain the differing attitudes towards work which develop throughout the general female population during the period after the economic collapse of 2002 that can be further explained through analysis of labor supply elasticities. Results A more fundamental understanding of what is driving Argentine women to work can be obtained by deriving estimates of labor supply elasticities for married women. While it is generally necessary for single women to work in order to generate an income and survive, married women are more likely to sit on the precipice of entering the labor force. Therefore, an attitudinal shift of married women can prove to be a key indicator of a rising female labor force. This information reveals a clearer vision for what the future of the female labor force participation rate. By looking at the income of married working women, the income of their spouses and the hours worked by these women from Q2 survey results of each year since 2004, estimates for the income elasticity and own-substitution elasticity can be found through cross-section analyses.7 Before beginning this analysis, it should be noted why a consideration of the income and own-substitution elasticities of the female labor force population is useful. Wage elasticities are normally examined over a much longer period of time. For instance, Goldin (1990) in her analysis of the married female labor force in the United States looks at how elasticities change over more than a century. Yet the period after the Argentine economic collapse merits a consideration of these elasticities given the incredibly rapid economic development that occurred in such a short amount of time. Note that the average deflated hourly wage of working married women in Q2 of 2003 was $1.43 and rose to $5.94 by Q2 2015 (table 3).8 Because of this, the wage elasticity of married females, η, defined as change in weekly hours worked over change in hourly wage, changes rapidly year-to-year (table 4). It is therefore constructive to look at the ratio of the income effect to the substitution effect in order to determine which dominates the average married woman’s decision to work. The labor supply elasticities for each year were derived using cross section analysis and the Slutsky equation.9 The Slutsky equation is defined as: η = ηs + α -ε. The equation relates the uncompensated wage elasticity, η, to three other terms: :ηs , own-substitution, or compensated wage, elasticity :-ε, income elasticity : α, wife’s full-time income divided by husband’s full-time income Cross-section estimates for the coefficients η, -ε and α can all be generated using information collected from the EPH-C survey. The uncompensated wage elasticity estimate is found by running a regression on married female hours worked per week over married female wage, which is defined as reported weekly income from primary job divided by hours worked at that job in a given week. Income elasticity is estimated by running a regression on married female hours worked per week over spouse’s wage, which is also defined as reported weekly income from primary job divided by hours worked at that job in a given week. Lastly, α is estimated by running a regression on wife’s weekly reported income from her primary job over her spouse’s weekly reported income from his primary job. The hope was that by only using information from primary work positions more consistent wage estimates would be derived however it should be noted that this could lead to errors in the estimates. In the end, the three coefficients found for Q2 in each year were used to calculate the own-substitution elasticity. It is important to note that the estimates for the ratio of the income elasticity to the own-substitution elasticity are noisy. Given consistent fluctuations in currency value, as seen by the constantly changing exchange rate to the U.S. dollar, rapidly rising real wages, effects from social welfare policies that greatly assisted low income families and potential biases in survey sampling from year-to-year, there many factors are at play. Yet, while the ratios do not follow a completely cohesive trend, they do tell a story. The income and own-substitution elasticities battle for dominance up until 2008, after which the income effect never overwhelms the substitution effect again. The rise of the substitution effect supports the effects of compositional trends with regards to differing attitudes towards labor force participation exhibited by different educational groups. Less married women being motivated by the economic status of their husbands and becoming more incentivized by their own wages suggests the construction of a more resilient female labor force that can will not shirk in periods of positive economic growth. If this is the case, then why has the female labor force participation rate seen a fall since 2011 according to both hourly estimates of married women and survey estimates for the entire population? To answer this question, one must turn to the changing real wages and real income of both married females and their spouses since 2011. In the period from 2011-2017 the real income of married females fell -1.878% while the average hours worked per week fell -3.933% (table 3). Additionally, the average uncompensated wage elasticity of married females was 3.090 (table 4). Using this information, a variant of the traditional supply and demand curves can be used to predict what labor force growth would have been in the absence of falling wages (figure 10). In this model, the horizontal axis is the rate of change of labor force participation (average hours worked per week) and the vertical axis is the rate of change of married female income (weekly earnings).10 The slope of the supply curve is the inverse of the average wage elasticity. The supply and demand functions can easily be derived from the following static supply equation: ℓs = [ S’ Ym -ε ] wη ℓd = Dw-δ where ℓ is married female labor force participation by weekly hours worked, Ym is husband’s weekly earnings, w is wife’s weekly earnings, -ε is the income elasticity, η is wage elasticity, δ is the elasticity of demand, S’ includes factors affecting labor supply except Ym and w and D includes all factors affecting demand except w. Taking logs of both sides of the equations and completely differentiating yields: ℓs * = [ S’ - εYm * ] + ηw = S* + ηw ℓd * = D* - δw where S* = [S’ - εYm* ]. With this linear equation and the previously calculated coefficient values, it can easily be found that the change in labor supply would have been +1.813% in the absence of wage deterioration. The elasticity of demand was not calculated for the purpose of this calculation however in order for the equilibrium to hold the demand curve requires a negative intercept and therefore declining labor demand.
A common explanation for the deceleration of female labor force participation in rapidly developing countries is too simple. Many economists have published works stating that the improving economic stature of poorer households removes the need for the wife to enter the workforce and as a result, female labor force participation falls. Although it is a salient explanation that can be applied to many countries in Latin America prior to periods of economic development, it must be applied with caution. While this story may hold true for the low educated portion of the female labor force in Argentina, the entirety of the story is far more complex. As education rises in Argentina, the correlation between labor force participation and periods of economic instability begins to fall away. In the past 10 years as educational attainment of female has risen rapidly, the population group which the original story pertains to has shrunk. There are greater forces at work that have curtailed the rise in the female labor force. A fall in economic productivity, translating to a fall in real wages, can be viewed as a primary factor preventing the rise in married female labor force participation. The mentalities dictating married women’s approaches to work have been shifting, likely due to a higher level of education and more occupational opportunities. With the necessary economic stimulus, Argentina’s could currently be pushing through the lower portion of the theoretical U-shaped curve of female labor force participation. However, the country is currently facing great economic strife that has prevented the market of female labor supply from reaching its potential. As opposed to viewing the female labor force as a potential cure to the country’s economic difficulty, the economic turmoil has instead ailed the growth of the female labor force. Although this may inhibit future rises of the female labor force, it is not the end of the story. Should the low educated portion of the female labor force continue to fall, convergence to both the medium and high educated labor force participation rates should also approach. Additional research can be conducted to understand what factors are driving the fall in wages across different industries to be able to more concretely interpret the effect on married women’s decision to work. Further insight can also be drawn by looking at compositional changes within industries in order to better understand what the future may look like. In conclusion, there is still much more to understand about the economic climate affecting the female labor market. All that can be said definitively is that the female labor force of Argentina still has room to grow.