The 2023 Nobel Committee awarded the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel to Claudia Goldin. Among her many contributions is the emphasis on the non-monotonic trend in women’s employment over the course of US history, as the result of structural change and evolving social norms regarding women’s responsibilities for home and family. This column documents 150 years of transformation in US hours and connects the trends to gendered change in the economy.
Most high-income countries witnessed a spectacular increase in women’s participation in the labour market during the second half of the 20th century. Rising women’s participation, however, is not a universal phenomenon. In fact, women’s employment has declined in the developing world in recent decades, as it did in high-income countries during other historical windows. Consequently, the relationship between women’s employment and GDP per head exhibits a non-monotonic pattern resembling a U-shape (as postulated by Boserup 1970 and Goldin 1990, among others). In contrast, men’s employment tends to decline consistently across different stages of development.
In a recent study (Ngai et al. 2024), we examine the relationship between gender trends in work and economic development through the lens of two processes: structural transformation – across the broad sectors of agriculture, manufacturing, and services – and the marketisation of home production. For this purpose, we build a consistent measure of men’s and women’s work in the US from 1870 to 2019, with an emphasis on the measurement of unpaid work in family businesses in the pre-1940 period. Alongside the correct characterisation of women’s contribution to the economy, the measurement of unpaid work in family farms matters for the estimation of agricultural productivity and structural transformation.
We establish trends on persons in work using data from the Census of Population from 1870 onwards. One clear drawback of census surveys is that – at least until 1940 – they include only paid occupations in employment statistics, thus massively undercounting women’s unpaid work in family businesses, mostly family farms. We therefore estimate the incidence of unpaid employment based on the share of persons living on farms and the (gainful) occupation of their household head. In particular, if a woman lives on the farm, does not report a gainful occupation, and is married to a self-employed farmer, we classify her as an unpaid family worker, motivated by evidence that farms relied heavily on family labour (Ruggles 2015). Indeed, early time-use data show that rural homemakers were working long enough hours on the family farm to be considered employed by modern definitions. Based on adjusted estimates (shown in Figure 2), the employment-to-population ratio for women falls from 56% in 1870 to 45% in 1940, reflecting the decline of unpaid work on farms, which virtually disappeared by 1960. The growth to 74% in 2019 is associated with the rise in services, employing 66% of women in 2019, corresponding to 89% of those in work. Men’s employment declines throughout the sample period, from 96% in 1870 to 84% in 2019, reflect declining agriculture until the 1960s, and declining manufacturing afterwards, partly offset by the rise in services.
Employment trends are accompanied by important variation in hours worked, both over time and across sectors. We combine information from several early sources of hours for the late 19th and early 20th century, which typically cover paid hours in specific occupations, sectors, or geographies. Thanks to early labour regulations and a structured work week, coverage for the manufacturing sector is reasonably systematic, and information on hours seems more reliable than for other sectors. Coverage is more sparse for the broad service sector and quite limited for agriculture.
Figure 1 Adjusted employment rates and industry shares: 1870–2019
a) Women
b) Men
Notes: This replicates Figure 2 in Ngai et al. (2024).
Finally, we obtain estimates of unpaid hours on the farm by drawing on early time-use surveys known as the Purnell diaries, which primarily sampled rural homemakers between the 1920s and the 1950s. According to the diaries, the typical homemaker on a farm would devote about 15.5 hours to farm activities per week in the 1920s, down to 7.5 hours in the 1950s, not including the time necessary to cater to farm employees or lodgers, if present. Due to sampling methods and the difficulty of recording unpaid agricultural work, hours series from early time-use surveys are inevitably affected by more severe measurement error than recent surveys. Drawing from the available sources, we also show realistic lower- and upper-bounds for unpaid hours.
Combining data on bodies and (paid and unpaid) hours from the sources described, we obtain a mid-range value of 21 hours per week for women’s market work in 1880 across the three sectors, down to 15 in 1940, reaching 21 again in 1980 and further rising to about 28 in recent years. The resulting U-shape in women’s hours shown in Figure 2 has a shallower left branch than the corresponding body-count in Figure 1, because hours spent in unpaid farm work – prevalent over the earlier period – are markedly lower than in regular, gainful employment.
Figure 2 Adjusted market hours per person, 1880–2019
a) Women
b) Men
Notes: This replicates Figure 6 in Ngai et al. (2024).
To account for the simultaneous evolution of men’s and women’s work and industry structure, we model a multisector economy in which individuals consume output from three sectors – agriculture, manufacturing, and services – and allocate their time to market work, home-production, and leisure. Consumers have a taste for variety; hence, the three types of goods are poor substitutes for consumption. In addition, the presence of a minimum food requirement in the consumption of agricultural produce implies that the demand for agricultural output is less income elastic than demand for manufacturing and service outputs. Services can be produced both in the market and in the home, with market- and home-produced varieties being close substitutes for each other. Productivity growth is uneven across sectors, being higher in agriculture and manufacturing than in services. Within the broad service sector, productivity grows faster in the market than in the home, as the scale of market production is more conducive to labour specialisation and technology adoption.
As outputs from the three sectors are poor substitutes, faster productivity growth in agriculture and manufacturing leads to structural transformation and a rise in services, via Baumol’s relative price mechanism. Conversely, faster productivity growth in market than home services, which are good substitutes for each other, leads to marketisation of home production. The simultaneous evolution of hours of work and industry structure can be summarised in two main phases. At early stages of development, when the agricultural sector is large, structural transformation is the main force at play, leading to declining agriculture, rising market and home services, and rising leisure via income effects. This implies a decline in market work, via the rise in both home services and leisure. At later stages of development, once the agricultural share is small, structural transformation mostly shapes labour reallocation from manufacturing into services. At the same time, a large service economy implies an important marketisation process, reallocating work from home to market services and raising market hours.
Patterns of gender specialisation determine the relative strength of these forces for men and women. Both men’s and women’s market hours fall along the initial phase of agricultural decline, while agriculture is still the core employer for both genders. But the later phase of manufacturing decline and service growth has differential impacts on men’s and women’s hours. As men and women specialise in manufacturing and services, respectively, the first channel is more relevant for men, implying a decrease in men’s hours, while the second channel is more relevant for women, implying an increase in women’s hours. Under the combination of structural transformation, marketisation, and gender specialisation, the evolution of women’s market hours displays a U-shape, while men’s market hours monotonically decline.
In a quantitative illustration of the proposed mechanisms, we establish that our model can reasonably reproduce the observed trends in men’s and women’s work under plausible combinations of relevant parameters, including the timing of the turning point in women’s market hours. In addition to the core processes of marketisation and structural transformation, we consider the evolution of within-sector gender intensity – reflecting gender-biased shifts in labour demand such as the mechanisation of agriculture or brawn-saving technologies in manufacturing – as well as social norms. The calibrated structural transformation and marketisation forces can account for the overall decline in market hours for both genders from 1880 and 1950, but only one quarter of the rise and decline, respectively, in women’s and men’s market hours from 1950 to 2020. While pre-1950 gender trends reflect almost exclusively the reallocation of labour across sectors, post-1950 trends are driven to a larger extent by within-sector forces.
source : voxeu