The Added Worker Effect in the Household

Carolyn M. Moehling, The Ohio State University

NOTE: The summary which appeared in the October 1997 issue of The Newsletter of The Cliometric Society contained graphics and formatting which cannot be accurately reproduced in this version.

Economists have produced a large literature examining the effects of the labor market outcomes of men on the labor force behavior of married women. In this literature, a great deal of attention has been focused on the impact of men's unemployment on the labor market participation of their wives. In the basic model of family labor supply decisions, the shortfall in income due to husbands being laid off, coupled with the inability to borrow against future earnings, will force some women not currently in the labor market to enter the labor market and will increase the labor supply of those women already in the labor market. This is the so-called " added worker effect." By focusing on labor market behavior, however, we miss a potentially important effect of men's unemployment on the work of women. The household has been, and continues to be, an important arena for women's labor. In response to her husband's unemployment, a woman may not only adjust her labor supplied to the market but also her labor supplied to household production. In other words, there may be an added worker effect in the household as well as in the labor market. Moreover, in eras in which married women faced limited job opportunities and low wages, the unemployment of men may have had a greater impact on women's labor in the household than on women's labor in the market. Focusing only on labor market behavior, therefore, leads us to underestimate the effects of men's unemployment on the work of women as well as the economic contributions of women to their families. My objective is to examine the added worker effect in the household in different historical periods. This paper represents the first step toward that objective.

The Added Worker Effect in the Household
The added worker effect in the household is best understood in a simple model of household decision-making that allows for household production. In the terminology of Becker's seminal 1965 article, households derive utility from Z-goods, basic commodities produced by the household by combining labor and market goods. Examples of Z-goods are meals and clean clothes. A household maximizes its utility subject to its budget constraints and production functions for the Z-goods. The solution of this maximization problem provides not only the household's demand for market goods but also its members' labor supplies to the market and to household production tasks. In response to the unemployment of the husband, the household can limit the reduction in its Z-good consumption either by increasing its income from other sources or by reducing the money requirements of its consumption bundle. The household can achieve the first response by increasing the labor supplied to the market by other household members. This is simply the " added worker effect in the labor market." The household can achieve the second response if it can substitute labor for market goods in Z-good production. This is the " added worker effect in the household."
This discussion makes clear that a necessary condition for the added worker effect in the household is the substitutability of market goods and labor in household production. For many household tasks, the assumption of some degree of substitution between labor and money seems reasonable. Consider for example the production of a meal. At one extreme, a household can produce a meal with almost no labor but substantial monetary cost by going to a restaurant. At the other extreme, a household can produce a meal with a large input of labor and relatively small monetary cost by growing its own vegetables, raising its own livestock, baking its own bread, and canning vegetables at home. In the 1930s and 1940s, Margaret Reid demonstrated the substitutability of labor and money in household production by comparing the production choices of households of different income levels. Low-income households, Reid found, chose more labor-intensive production techniques than high-income households. In particular, low-income households were much less likely to send their laundry to commercial laundries and more likely to bake bread than to buy it from bakeries. By choosing more labor-intensive modes of household production, women in lower-income households were able to extend the resources of their families. Given the relationship between household production choices and income, it seems reasonable that women responded to the unemployment of their husbands and the shortfall of income that entailed, by adjusting their labor in the household as well as their labor in the market.
Evidence of changes in household behavior during the Great Depression provides support for the added worker effect in the household (Milkman 1976). In 1931, there was a boom in the sales of glass jars used in home canning and a fall in the sales of commercially canned products (La Follette 1934, 95). A study of consumption patterns in Milwaukee between 1929 and 1931 found that annual prepared soup consumption per family fell from 75 to 57 cans while flour consumption per family increased from 177 to 256 pounds, suggesting the substitution of home baked goods for bakery goods (Reid 1943, 140). Many studies of how families dealt with the Depression emphasize the change to more labor-intensive household production techniques to save money. These studies report that woman reduced their use of commercial laundries, canned foods at home, and even took up sewing to maintain their families' consumption in the face of diminished income flows (Morgan 1939, 33; La Follette 1934, 102).
The evidence from the Depression suggests that adjustments of household labor may be an important way in which families respond to unemployment and therefore, an important topic to study. However, the study of the added worker effect in the household faces several challenges. The first set of challenges comes from the complexities of household production. First of all, what we refer to as " household production" is really many different tasks. The monetary return of increasing household labor will vary greatly across these tasks due to the relative prices of market substitutes and the market inputs to home production, and even the skills of the household worker. Secondly, while household labor and market goods may be substitutes in household production, home-made goods and market goods are not always perfect substitutes. Families may prefer home-made bread or home-canned tomatoes to their commercially produced counterparts. In addition, many household workers derive positive utility from some household tasks such as cooking and baking. In such cases, choosing the more labor-intensive household production technique may represent leisure consumption rather than a cost-saving choice. These points are well-illustrated by the 1931 survey of 652 gainfully employed married women by Cecile La Follette (1934). La Follette found that of the working women in his sample, over 50% made jelly, 45% canned fruit, and 17% canned vegetables. When asked why they canned or made jelly, 38% of these women said it was because they liked home-canned products; 11%, because home products were better quality than commercial products; 22%, because they liked to can; and only 28%, to save money.
These non-monetary influences on household production techniques imply that households do not always allocate labor across household tasks to maximize the monetary return. For instance, a household in which the household worker enjoys baking bread may choose to send out its laundry and bake bread at home even though the cost savings of doing laundry at home are greater than those of baking bread. Therefore, the impact of unemployment on household labor may not just be an increase in the total number of hours worked in the household but also (or even instead) a re-allocation of hours across household tasks. More time would be allocated to tasks with greater monetary returns at the expense of tasks with lower monetary returns but high non-monetary returns. So the performance of some household tasks may actually decline in the face of the unemployment of the primary market worker, even though the household added worker effect is at work. Which household tasks will decline will depend on relative prices as well as the preferences and skills of the household worker, things difficult for the outside observer to measure. The implication of this for the study of the added worker effect in the household is that many dimensions of household production must be examined. The added worker effect will have different magnitudes and even different signs for different household tasks. The effect will likely be large and positive for tasks such as laundry work that have almost equivalent quality market substitutes, offer little enjoyment for the household worker, and have high monetary returns. The effect will be much smaller and may even be negative for tasks such as baking that produce much higher quality output than the market alternatives, provide enjoyment for the household worker, but have small monetary returns.
The second set of challenges for the study of the added worker effect in the household stem from data limitations. The data we would like to have are time spent in different tasks accompanied by actual measures of the output of those tasks such loaves of bread baked, jars of vegetables canned, jars of fruit preserved, and pounds of laundry washed. Such data are extremely rare, especially for historical periods. When such data exist, they typically are not accompanied by information on unemployment.1 The most readily available data that contain information on unemployment and provide insights into household production are data from household budget surveys. While these data cannot provide definitive measures of the output of household production, they can provide indicators of household production choices such as the expenditures on laundry sent out, the ratio of canned to fresh vegetables purchased, and the ratio of bakery bread to flour purchased.

Data and Empirical Results
For my examination of the added worker effect in the household, I use data from the Bureau of Labor Statistics Cost of Living Survey of 1917-1919. These data were used to construct the original weights of the Consumer Price Index and contain information on the income and expenditures of over twelve thousand families of wage-earners and low- and medium-salaried workers living in 99 industrial centers. Families included in the survey were required to contain an intact married couple and at least one child, to be not " slum or charity" families, to have no boarders, and to be able to provide all items of expenditure and the income of all family members. Each family was surveyed on its income and expenditures over a one year period. The interviews in different cities were staggered over a couple of years, however, so that the data from different cities cover different annual periods between July 31, 1917 and February 28, 1919.
The primary weaknesses of these data for my analysis stem from the fact that they pertain to a period in which the U.S. was at war. Dramatic changes in relative prices due to wartime shortages most likely had significant effects on household decisions. Household behavior may also have been altered by non-pecuniary factors during the war such as the government propaganda campaign aimed at reducing the consumption of wheat flour. Another problem for my analysis is that unemployment during these years was extremely low. The wartime boom in production led to high demand for labor and estimated unemployment rates of less than 3% for both 1918 and 1919 (Weir 1992, 341).2
Another weakness of these data for my analysis is that they do not contain information on unemployment as traditionally defined. Unemployment, as we traditionally define it, is time spent out of work and seeking work. The BLS data contain the number of weeks worked in the year preceding the survey date. Subtracting weeks worked from 52 gives us a measure of " nonemployment." Nonemployment will include unemployment as well as time spent out of the labor force for such things as vacations and illness. To compensate for this in my analysis, I consider the effects of three levels of the husband's nonemployment: greater than 4 weeks, greater than 13 weeks, and greater than 26 weeks.
I consider five indicators of household production choices: expenditures on laundry sent out of the home, the ratio of canned tomato purchases to total tomato purchases, the ratio of canned corn purchases to total corn purchases, the ratio of wheat bread to wheat flour purchased,3 and net income from garden and poultry. The first four variables are negative measures of home production: an increase in home production is indicated by a decrease in these variables. The last indicator, net income from garden and poultry, is a positive measure of household production.
In the estimated specifications, I control for various household characteristics that may have influenced household production choices. One very important determinant of household production choices was the labor market status of the woman of the household. The woman's labor market status is clearly endogenous to household production choices, but for now I treat it as exogenous. I estimate tobit specifications because all five of these indicators are censored at zero. Table 1 presents highlights of these results.
While the results in Table 1 do provide support for the added worker effect in the household, they also illustrate the complex nature of the relationship between male unemployment and household production choices. The most straightforward and strongest evidence in support of the added worker effect in the household comes from the specification for expenditures on laundry sent out of the household. Nonemployment of the husband for more than 4 weeks lowered laundry expenditures by $2.90. This is a substantial effect given that the unconditional mean of laundry expenditures was $11.75 and the conditional mean was $17.31. Households responded to the nonemployment of the husband by sending less laundry out and most likely increasing the laundry work done in the household. However, the specifications for the ratio of canned to total tomato purchases, the ratio of canned to total corn purchases, and net income from gardens and poultry, indicate a reduction in the relative importance of home production in response to the husband's nonemployment. The nonemployment of the husband appears to have increased the relative importance of commercially canned products in a household's consumption. This finding is not surprising, however, in light of the estimated negative effects of husband's log weekly wage and household non-labor income on relative canned good consumption. These negative income effects suggest that households viewed commercially canned products as inferior to home-canned or fresh vegetables. Increased canned good consumption in response to the husband's unemployment, then, may reflect a re-allocation of household labor to tasks such as laundry that had higher monetary returns. The negative effect of husband's nonemployment on net income from gardens and poultry may be interpreted in the same way. Husband's nonemployment had no statistically significant or sizable effect on a household's ratio of wheat bread to wheat flour purchases. In contrast to Reid's evidence from the 1940s, a household's relative consumption of bakery bread was not strongly affected by the husband's log weekly wage and household non-labor income. This result may reflect distortions in household production choices brought on by the war or may simply reflect that income was not an important determinant in the decision to buy or to bake bread during this earlier time period.

Final Remarks
The preliminary results presented in this paper suggest that male unemployment had significant effects on household production decisions in 1917-1919. I find that, at least in the case of laundry work, households reduced their use of labor-saving market alternatives in response to the unemployment of the male worker. This lends support to the hypothesis of an added worker effect in the household. Future work will attempt to take account of the endogeneity of the woman's labor market status and will consider how the responsiveness of household production choices was affected by household characteristics such as the presence of small children or additional adult females.

1Most of the data that exists on the output of household production pertains to farm households.
2The data from the BLS Cost of Living Survey of 1934-1936 would be better-suited for my analysis since they do not suffer from these problems and also have some information on household production choices like purchases of yard goods not contained in the earlier survey. However, the household-level expenditure data of the 1934-1936 survey are not currently available.
3All three ratios represent quantities purchased rather than expenditures.

La Follette, Cecile Tipton. 1934. A Study of the problems of 652 gainfully employed married women homemakers. Contributions to Education, No. 619. New York: Teachers College, Columbia University.
Milkman, Ruth. 1976. Women's work and economic crisis: some lessons of the Great Depression. Review of Radical Political Economics 8 (Spring): 73-97.
Morgan, Winona L. 1939. The family meets the Depression. Minneapolis: The University of Minnesota Press.
Reid, Margaret. 1934. Economics of household production. New York: John Wiley and Sons, Inc.-------. 1943. Food for people. New York: John Wiley and Sons, Inc.
Weir, David R. 1992. A century of U.S.unemployment, 1890-1990: revised estimates and evidence for stabilization. Research in Economic History 14: 301-346.

Table 1. Tobit Regression Results for Household Production Indicators