Resistant to the Recession: Low-Income Adults’ Maintenance of Cooking and Away-From-Home Eating Behaviors During Times of Economic Turbulence

Abstract

Objectives. We examined the effects of state-level unemployment rates during the recession of 2008 on patterns of home food preparation and away-from-home (AFH) eating among low-income and minority populations.

Methods. We analyzed pooled cross-sectional data on 118 635 adults aged 18 years or older who took part in the American Time Use Study. Multinomial logistic regression models stratified by gender were used to evaluate the associations between state-level unemployment, poverty, race/ethnicity, and time spent cooking, and log binomial regression was used to assess respondents’ AFH consumption patterns.

Results. High state-level unemployment was associated with only trivial increases in respondents’ cooking patterns and virtually no change in their AFH eating patterns. Low-income and racial/ethnic minority groups were not disproportionately affected by the recession.

Conclusions. Even during a major economic downturn, US adults are resistant to food-related behavior change. More work is needed to understand whether this reluctance to change is attributable to time limits, lack of knowledge or skill related to food preparation, or lack of access to fresh produce and raw ingredients.

The obesity epidemic in the United States exhibits a steep social gradient according to socioeconomic status (SES) and race,^1–4 with the poor, Hispanics, and non-Hispanic Blacks bearing a disproportionate burden of overweight and obesity.^5,6 This disparity stems in part from substantial barriers to achieving a healthful diet, including food insufficiency, food deserts, and the preponderance of cheap, high-energy foods.^7–10 However, a growing body of evidence indicates that time scarcity, as opposed to money, is the most significant barrier to achieving the nutritional targets set by nutrition allotment programs such as the Thrifty Food Plan, which typically necessitate cooking from scratch to meet financial constraints.^11–14

For many resource-scarce households, the struggle to manage the competing demands of work, transportation, social services, and child care limits the time available to prepare healthy meals^15–19 and prompts the purchase of quick, convenient foods^20–23 such as away-from-home (AFH) foods or processed foods, which tend to be energy dense and nutrient poor.^9,10,24–26 In fact, the Institute of Medicine recently recognized the importance of such time limitations, calling for the US Department of Agriculture to incorporate the value of time in calculating adequate nutrition program allotments.²⁷

Notwithstanding the growing recognition of time as a major limitation to home food preparation, it is less clear how food preparation and consumption patterns have been influenced by recent economic downswings. The Great Recession of 2008 was characterized by an increase in the national unemployment rate from 5.0% in December 2007 to 9.5% in June 2009.²⁸ Historically, low-SES adults have been more likely to suffer recession-related effects, including layoffs, earnings reductions, and human capital losses,^29,30 and this was also the case during the Great Recession, with younger, less educated, and minority workers experiencing steeper increases in unemployment.^31,32

Previous economic downturns have been linked to shifts in food preparation patterns and diets. For example, during Russia’s economic collapse in 1998, low-income households increased the amount of food they prepared at home with basic ingredients as a strategy to decrease the cost per calorie of food and preserve a diet comparable to their precrisis diet.³³ In fact, recent studies have shown that, in developed countries, recessionary periods are associated with better health behaviors, including increases in physical activity, increased fruit and vegetable intake, and decreases in obesity.^34–39 These effects occur outside of individual employment status, with increases in aggregate unemployment rates changing the way people in the affected area make decisions about money and time.

For example, wage rate increases are dampened during economic downturns, affecting household income and consequently food shopping and preparation. Similarly, volatility in the job market⁴⁰ decreases the opportunity cost of time, making it less costly to undertake health-promoting activities such as exercise and cooking. Persistent elevated stress associated with the increased likelihood of unemployment can also affect the amount and healthfulness of food that people buy and consume.^41,42

However, previous work on recessions and health has neglected to consider the impact of a recession on the sociodemographic groups most affected, including low-SES and minority populations. One possibility is that these groups might save money by reverting to cooking with basic ingredients, similar to the situation in Russia in 1998. Alternately, households struggling to find and maintain resources in an economically precarious environment might choose to save time and spend money by purchasing more convenience or AFH foods.

We assessed whether US households increased the amount of cooking done at home and decreased AFH eating in association with the Great Recession and whether low-income and racial/ethnic minority groups were disproportionately affected by the economic downturn. Using data from a nationally representative time use study and state-level information on differences in employment rates (to parallel recession effects), we examined patterns of food preparation and consumption before and after the recession. Our goals were to determine how temporal and recession-related effects were associated with changes in home cooking practices and whether these effects varied across key demographic groups disproportionately affected by changes in the labor market (including low-income and minority populations). We also sought to understand how temporal trends and the recession were associated with patterns of AFH eating to address the issue of how people chose to maximize the time–cost tradeoff: by cooking more and spending less or by spending more and cooking less.

METHODS

The methodological details of the American Time Use Survey (ATUS) have been published elsewhere.⁴³ Briefly, in 2003 ATUS began developing nationally representative estimates of time use in the United States. ATUS includes free-living residents of households in all 50 states and the District of Columbia who are aged 15 years or older; active military personnel are excluded. One individual from each selected household is randomly chosen to participate in ATUS. Computer-assisted telephone interviews were used to ask respondents about their time use (including activity and location) during a specific 24-hour period. We included in our analyses pooled data from 2003 to 2011 for all adults aged 18 years or older; the final sample consisted of 118 635 respondents.

Outcome Measures

Time spent cooking.

The time spent cooking variable included any time spent in food preparation or meal-related cleaning. In this sample, cooking participation and the distribution of time spent cooking per day varied significantly between men and women, with 68% of women but only 40% of men reporting that they took part in cooking meals. To reflect the different gender distributions and capture the inherent differences between not cooking at all, cooking for a short time, and cooking for an extensive amount of time, we defined time spent cooking as a 3-level categorical variable with separate cut points for male and female respondents. Categories for male respondents were not cooking (60%), cooking between 1 and 39 minutes each day (23%), and cooking 40 or more minutes each day (17%). For female respondents, the corresponding categories were not cooking (32%), cooking 1 to 59 minutes per day (35%), and cooking 60 or more minutes per day (33%; Table A, available as a supplement to the online version of this article at http://www.ajph.org).

Away-from-home eating.

A respondent was designated as having eaten away from home if he or she reported any time spent eating at a bar or restaurant (including full-service and fast food establishments) at any point during the 24-hour period assessed.

Explanatory Measures

Age, race/ethnicity, educational level, type of household, individual employment status, time period (before or after the recession), high unemployment level, and poverty were used as explanatory variables. Respondents were assigned into one of 4 age groups (18–29 years, 30–44 years, 45–64 years, and ≥ 65 years) to reflect progression through early and middle adulthood and retirement age. Self-reported race/ethnicity was categorized as non-Hispanic White, non-Hispanic Black, Hispanic, or non-Hispanic other.

Education was grouped as follows: less than high school, high school or equivalent, some college, bachelor’s degree, or graduate degree. Household type reflected household composition and was defined as living alone, living with a spouse, living with another nonspousal adult, living with a child or children (single parent), and living with a child or children and another adult.

We initially categorized individual employment status as not in the labor force, unemployed, employed part time (< 35 hours/week employed in all jobs combined), or employed full time (≥ 35 hours/week). However, unemployed respondents accounted for only 6.4% of the sample, and they did not differ significantly from those who were not members of the labor force on any of the measures assessed. Thus, we subsequently recoded employment status as not in the labor force (including those who were unemployed), employed part time, and employed full time.

Time period was classified as prerecessionary (2003–2007) or postrecessionary (2008–2011) to reflect the beginning of the economic downswing in December 2007. Respondents’ state of residence and survey year were linked to state-level unemployment rates from the Bureau of Labor Statistics. High unemployment, a binary variable, was defined as an unemployment rate in the top quintile (i.e., ≥ 8.2%). Respondents were categorized as below the poverty line if their reported family income was less than the weighted average federal poverty threshold for the relevant year.

Statistical Analysis

We used Stata version 12 (StataCorp LP, College Station, TX) in conducting all of our statistical analyses. We used multiple imputation to account for missing data on household income (13.7% of the respondents had missing data on this variable).⁴⁴ The imputation model incorporated all covariates used in our analysis along with additional variables associated with family income, including type of housing, industry and occupation, number of children younger than 18 years, household size, and spouse’s employment status and education.

Because a power correlation revealed that family income level was associated with whether or not data on income were missing, we conducted a sensitivity analysis comparing the findings with and without imputed incomes. The results of this analysis showed very little change in magnitude of effect and no reversal of direction for any covariate. Only one covariate, high unemployment level among men cooking 0 to 39 minutes per day, shifted from nonsignificance to statistical significance (P < .05) when the model included imputed income, suggesting that these findings are generally robust to any bias induced by the nonrandom structure of missing income data.

We conducted multinomial logistic regression analyses to evaluate the associations between below poverty level status, high state-level unemployment, time period (prerecessionary or postrecessionary), age group, education, race/ethnicity, household structure, respondent employment status, time spent cooking, and whether the participant ate away from home. All models controlled for whether the diary day was a holiday. The results of a Wald “chunk” test assessing interactions between gender and all of the covariates were significant, justifying the stratification of subsequent models by gender.

To assess whether the recession affected low-income and racial/ethnic minority groups differently than high-income and racial/ethnic majority groups, we examined interactions between time period and poverty, high unemployment and poverty, and high unemployment and race/ethnicity. Log binomial regression was used to examine the associations between the aforementioned covariates and the likelihood of AFH eating. Results are presented as average marginal effects to show how distributions of home food preparation and AFH eating changed in relation to the recession and sociodemographic covariates.

We used the ATUS final probability weight to account for the distribution of the sample across days of the week, to account for differences in response rates across demographic groups, and to adjust the sample so that it was nationally representative. We did not use replicate weights to calculate standard errors because successive difference weights cannot be applied to multiple imputation models in Stata. However, a sensitivity analysis conducted without multiple imputation and with and without the replicate weights showed very little difference in standard errors and did not alter effect sizes or statistical significance for any covariates.

RESULTS

Descriptive characteristics for the sample according to the percentages of respondents who participated in household cooking are presented in Table A. Relative to their counterparts, those who cooked were more likely to be female, older, and better educated; not to be members of the labor force; and to have a household income below the poverty threshold.

Among men, household income below the poverty line was not associated with likelihood of cooking. However, there was a significant interaction between recessionary period and income, with low-income men increasing their amount of cooking more than higher income men in the postrecessionary period. The number of men with household incomes below the poverty line who reported some or an extensive amount of cooking increased by 6%, whereas the corresponding increase among higher income men was 2% (P < .01; Figure 1 ). High unemployment rates increased the likelihood of cooking but only slightly, and there were no significant findings with respect to interactions between high unemployment and income or race/ethnicity (Table B, available as a supplement to the online version of this article at http://www.ajph.org).

FIGURE 1—

Predicted probability of cooking among men, by income and time period: American Time Use Survey, 2003–2011.

Note. Values are multinomial logistic regression results weighted to be nationally representative and adjusted for holidays, age, education, race/ethnicity, household type, individual employment status, and state-level unemployment rate. The sample size was n = 51 139.

^aWithin the poverty group, the percentage of respondents in the cooking category is significantly different between the prerecessionary and postrecessionary periods (P < .01).

Among women, those who were below the poverty line were more likely to cook than higher income women (Table B). Women living in states with high unemployment rates were more likely to cook for a longer duration each day than women living in states with low unemployment rates, but the differences were minor (Figure 2 ). In addition, there was no effect of time period or interaction between time period and income (Table B).

FIGURE 2—

Predicted probability of cooking among women, by income and state-level unemployment: American Time Use Survey, 2003–2011.

Note. Values are multinomial logistic regression results weighted to be nationally representative and adjusted for holidays, age, education, race/ethnicity, household type, individual employment status, and time period. Within income groups, there were no significant differences with respect to low vs high state-level unemployment in the percentage of respondents cooking at any level (P < .01). The sample size was n = 67 496.

White non-Hispanic men (58%) were significantly less likely than non-Hispanic Black men (62%), Hispanic men (66%), or men in the “other” category (62%) to report that they did not cook (Figure 3 ). Black women (39%) were significantly more likely than White women (32%), Hispanic women (28%), and women in the “other” category (30%) to report that they did not cook for any amount of time; Hispanic women and women in the “other” category were more likely to report extensive cooking. There were no significant interactions between race/ethnicity and state-level unemployment for either men or women (Table B).

FIGURE 3—

Predicted probability of cooking, by race/ethnicity: American Time Use Survey, 2003–2011.

Note. NH = non-Hispanic. Values are multinomial logistic regression results weighted to be nationally representative and adjusted for holidays, age, education, household type, individual employment status, time period, and state-level unemployment rate. Cooking duration categories were as follows: no cooking, 0 minutes/day for both men and women; moderate cooking, 0–39 minutes/day for men and 0–59 minutes/day for women; and extensive cooking, ≥ 40 minutes/day for men and ≥ 60 minutes/day for women. The sample size was n= 1 18 635.

^aWithin gender groups, the percentage of respondents in the cooking category is significantly different than that in the non-Hispanic White group (P < .01).

^bWithin gender groups, the percentage of respondents in the cooking category is significantly different than that in the non-Hispanic Black group (P < .01).

^cWithin gender groups, the percentage of respondents in the cooking category is significantly different than that in the Hispanic group (P < .01).

^dWithin gender groups, the percentage of respondents in the cooking category is significantly different than that in the non-Hispanic “other” group (P < .01).

Both men and women were much less likely to eat away from home if they were below the poverty line or older and much more likely to do so if they were well educated (Table C, available as a supplement to the online version of this article at http://www.ajph.org). Among men, eating away from home did not appear to be affected by time period or high state-level unemployment. High state-level unemployment was associated with a 2% decrease in the likelihood of AFH eating among women below the poverty line and no change among women above the poverty line, but the interaction between high state-level unemployment and poverty was not statistically significant (Figure 4 ). Although women appeared to be less likely to eat away from home in the postrecessionary period than in the prerecessionary period (P < .01), this effect resulted in decreases of only 2% and 1% in AFH eating among women below and above the poverty line, respectively, and the interaction between time period and income was not significant (Figure A, available as a supplement to the online version of this article at http://www.ajph.org).

FIGURE 4—

Predicted probability of eating away from home, by state-level unemployment and income: American Time Use Survey, 2003–2011.

Note. Values are log binomial regression results weighted to be nationally representative and adjusted for holidays, age, education, household type, race/ethnicity, individual employment status, and time period. Within income groups, there were no significant differences with respect to low versus high state-level unemployment in the percentage of respondents eating away from home (P < .01). The sample size was n = 118 635.

Among both men and women, non-Hispanic White respondents were most likely to eat away from home (21% of men and 19% of women), and non-Hispanic Blacks were least likely to do so (13% of men and 10% of women; Figure B, available as a supplement to the online version of this article at http://www.ajph.org). However, there was no interaction between state unemployment rate and race/ethnicity (Table C).

DISCUSSION

The key finding of this study is that the Great Recession had little effect on food preparation and AFH consumption patterns among US adults. Although high state-level unemployment was linked to a shift toward an increased likelihood of cooking among both men and women, this association produced only minor increases in time spent cooking and had virtually no effect on the likelihood of AFH eating.

This lack of an effect is somewhat surprising considering that global food prices skyrocketed in 2007 and 2008⁴⁵ and overall food expenditures decreased.⁴⁶ We expected to see consumers moving away from processed items such as ready-to-heat and ready-to-eat meals, which are typically more expensive (by volume) and less time consuming, to the use of raw, unprocessed ingredients to assemble meals from scratch.

Similarly, US adults did not change their AFH eating patterns. This finding is consistent with food expenditure data showing that although the recession shifted the relative share of food dollars from AFH to at-home foods by approximately 3% to 5%, the decline in AFH spending represented a switch from higher-end, sit-down restaurants to cheaper options such as fast food and casual dining restaurants rather than increased home food preparation.⁴⁶

Our results are consistent with time use studies showing that, during economic downturns, people spend increased time on leisure and personal care activities (e.g., television watching and sleeping) but exhibit only relatively small increases in domestic production activities such as cooking.^47,48 In addition, although fluctuating unemployment rates appear to increase domestic activities, long-term elevated unemployment similar to what the United States experienced during the Great Recession has little effect.⁴⁸ In fact, recent research shows that the link between higher state-level unemployment and improved health status and health behaviors all but disappeared during the Great Recession, suggesting that the countercyclical relationship between economic downturns and health has diminished.⁴⁹

Finally, the propensity to cook is influenced by whether a person has the skill or knowledge to do so. The recent decreases in home cooking practices have had a negative effect on the intergenerational transmission of cooking knowledge and skill.⁵⁰ Without personal experience in cooking, many of today’s working-age adults who may have wanted to increase their frequency of cooking to save money might not have been able to do so.

Income and Race/Ethnicity

Although household income below the federal poverty threshold was not associated with an increased likelihood of cooking among either men or women, it was associated with a strong and consistent increase in the likelihood of women cooking for longer durations. Education showed a similar effect in women; those who were more educated were more likely than those who were less educated to cook at all, but less educated women were more likely to cook extensively. These results are consistent with expectations: adults in low-wage jobs have lower opportunity costs with respect to time and a higher financial incentive to save money by increasing the time they spend in activities such as food preparation. However, our results also underscore the need for nutrition assistance programs to address the limited time spent on cooking; nearly a third of low-income women reported no cooking at all and an additional 37% reported cooking fewer than 60 minutes per day, translating to roughly 20 minutes or less per meal.

Despite their economic vulnerability, the recession did not disproportionately affect food preparation or consumption patterns in low-income households. The reason for this lack of effect could be that individuals with low incomes are more likely to be unemployed or more likely to work outside of the traditional labor market during nonrecessionary periods, and thus they are less likely to be affected by increases in overall unemployment during a recession. Because these individuals were already operating within a low-wage market, a weaker labor market was less likely to reduce their opportunity costs in terms of time or increase the time they spent on home cooking activities, especially considering that other demands on time remained high or even higher. In addition, social welfare programs such as the Special Supplemental Nutrition Program for Women, Infants, and Children can provide a buffer against the effects of economic fluctuations on food-based resources.^35,51

Men who were below the poverty line exhibited larger increases in time spent cooking from the prerecessionary period to the postrecessionary period than higher income men. This trend also is apparent in food expenditure patterns, with low-income households showing the greatest increases in at-home food expenditures.⁴⁶ These results may reflect the higher proportion of low-income men working in lower-paid, physically demanding industries such as construction and agriculture, which experienced the greatest increases in unemployment during the recession.³⁹

Although food preparation and AFH food consumption patterns varied significantly according to race/ethnicity, the recession did not appear to affect these patterns. Non-Hispanic Black households were less likely to cook and less likely to eat away from home, suggesting a high level of reliance on preprepared convenience foods. These results make sense in the context of research showing that predominantly Black neighborhoods tend to have fewer supermarkets^52,53and fewer full-service and fast food restaurants than White neighborhoods.⁵⁴ Given the persistence of these patterns throughout the recession, programs promoting home cooking must consider whether the groups they are targeting have the necessary skill, available time, and access to supermarkets to prepare foods from scratch.

Limitations

We did not consider the stability or duration of state-level unemployment rates or community-level unemployment rates, which could have had an impact on the degree to which the recession affected cooking behaviors. In addition, because our measure of AFH food consumption was based on respondents’ reports of eating at a restaurant, it did not include situations in which respondents consumed takeout or delivery meals or instances in which they consumed AFH food while engaging in other activities. As a result, we may have underestimated AFH eating in our sample. Similarly, respondents’ perceptions of cooking and multitasking limited the precision of the home cooking measure, in that those who simply heated up a frozen meal or cooked while doing other household activities might not have reported this as cooking, most likely resulting in an inflated number of respondents who reported that they did not cook.

More importantly, because of the cross-sectional nature of ATUS, we were unable to examine the causal role of the recession in changes in food preparation patterns or explore how these recessional effects were linked to changes in diet patterns or health outcomes. Recent research suggests a link between maternal food-related time use and childhood obesity,⁵⁵ but longitudinal methods are needed to understand whether food preparation patterns are linked to improved diet quality and obesity and, if so, the mechanisms that mediate this effect.

Conclusions

Overall, our results show that the Great Recession did not have an overwhelming effect on the cooking patterns or AFH food consumption patterns of US adults, nor were low-income groups disproportionately affected. Our findings suggest that, even during a major economic downturn, US adults are resistant to dietary change and are willing to preserve their precrisis diets despite rising costs and decreased employment. Although low-income women were more likely to cook than their well-off counterparts, many did not cook at all or cooked for small amounts of time. Minority men and Black women were less likely to cook than their counterparts, and these patterns persisted throughout the recessionary period.

Our study provides further evidence that nutrition assistance programs promoting home cooking with basic ingredients must consider whether the groups they are targeting are able to meet nutritional recommendations for the preparation of healthy foods. However, more research is needed to determine whether the resistance of households to changing their food preparation behaviors is attributable to time limits or to other barriers such as lack of cooking knowledge and skills^56,57 or lack of access to fresh produce and raw ingredients.