چگونگی تغییرات مصرف و تولید مواد غذایی خانگی بازنشستگی: درسهایی از داده مدت زمان استفاده آلمان

In order to test whether a retirement-consumption puzzle does exist, we examine how food-related time use alters within the 50+ generation in Germany due to retirement. Based on the German Time-Use Survey, time-use patterns of retired and non-retired persons are compared statistically and determinants of time-use are elaborated by the use of double-hurdle and multiple regression models. With retirement, major changes take place in the food-related time use. Work-related food-away-from-home consumption is substituted by food production and consumption at home and associated shopping activities. Leisure-related away-from-home consumption gains importance for a portion of pensioners. These impacts are strong and highly significant for German households. By and large, there is no indication of a retirement-consumption puzzle but of a planned behavioral change in a new phase of life. Econometric analysis shows that other personal and sociodemographic variables explain food-related time use patterns in the 50+ generation, in particular gender, age, perceived health and the educational level attained

Constant or negative population growth rates in industrialized countries and a higher life expectancy have induced a demographic change in which the share of elderly persons has increased substantially. Thus, the economic behavior of the elderly has gained a rising interest over the last decade in macroeconomics, consumer economics, and agricultural and food economics. One particularly lively debate has been related to whether a retirement-consumption puzzle exists and how the empirical evidence can be explained. Although broad empirical evidence has emerged from country studies (for surveys see Hurst, 2008 and Attanasio and Weber, 2010), a comprehensive study of the implications of retirement for the consumption and household production of food is lacking. We intend to provide such an analysis with time-use data for Germany. The retirement-consumption puzzle is closely related to the life-cycle theory of income and consumption (Modigliani and Brumberg, 1954 and Ando and Modigliani, 1963) and the permanent-income theory of consumption (Friedman, 1957). According to basic versions of these theories, rational forward-looking consumers are expected to save at younger ages and to dissave at older ages in order to keep a constant utility level over the life-cycle. Doubts have been raised in theoretical and empirical analyses against this consumption-smoothing hypothesis. Firstly, Hamermesh (1984) argued that savings of consumers are too low to keep the level of consumption constant after retirement. Secondly, it was shown that income and consumption expenditures dropped with retirement in the USA (Bernheim et al., 2001), in the United Kingdom (Banks et al., 1998 and Smith, 2006), in Italy (Battistin et al., 2009), Japan (Wakabayashi, 2008), and Germany (Lührmann, 2010). The existence of a retirement-consumption puzzle has been stated in the literature when the joint decline of consumption and income was not consistent with rational, forward-looking behavior in extended life-cycle models. Banks et al. (1998) argue along these lines and conclude that retirement cannot fully explain the fall in expenditures observed in the British Family Expenditure Survey. Bernheim et al. (2001) cannot reject a retirement-consumption puzzle, too, for their analysis of U.S. panel data. According to their analysis, households do not smooth the effects of predicted income changes in a “framework of rational, farsighted optimization” (ibid., 2001, p. 855) but rather follow heuristic rules of thumb according to which they adjust their saving behavior at retirement. Other authors challenge the existence of a retirement-consumption puzzle. Some take up the question whether retirees face anticipated or unanticipated declines of income and consumption. When contrasting U.S. information on expected changes in consumption at retirement with actual changes, Hurd and Rohwedder (2003, p. 17) find out that “consumption changes come as no surprise to most people” and spending changes are anticipated on average. The authors see “no reason to interpret the changes to a lack of forward-looking behavior” (ibid., p. 17). Wakabayashi (2008) for Japan and Haider and Stephens (2007) for the U.S. reach similar conclusions. Haider and Stephens look at workers’ subjective beliefs about when they will retire and find that these expectations are “strong predictors” of actual retirement decisions. Survey results by Ameriks et al. (2007, p. 265) confirm “that many working households do expect a considerable fall in consumption when they retire”. The empirical evidence further reveals that a disaggregation of consumption expenditures is crucial for a comprehensive assessment of linkages between ageing and consumption (Aguiar and Hurst, 2013). The influence of ageing differs markedly between foods and other nondurables such as entertainment. According to Aguiar and Hurst, 2013 and Aguila et al., 2011, food expenditures are one major expenditure category that declines after retirement whereas other expenditure categories for nondurable goods remain constant or even rise. One reason might be that expenditures for food away from home fall when leaving the labor force. Secondly, household-production theory suggests that pensioners will substitute market-related consumption of foods by home production as more time is available after retirement. Most likely, they will also spend more time on shopping and may thus be able to save on the purchase of commodities for home production of food. It may well be possible to keep a given level of utility from food consumption with lower food expenditures after retirement. Precisely this pattern has been found as a rational explanation of declining expenditures for food after retirement, for example, in the USA (Aguiar and Hurst, 2005), in Germany (Lührmann, 2010) and recently in Spain (Luengo-Prado and Sanz, 2013). For U.S. households, Aguiar and Hurst (2005) confirm a fall of food expenditures by 17% at retirement, but it is matched by a 53% increase in time spent on food production at home. Moreover, the time spent on shopping rises strongly and leads to a certain decline of unit values of grocery items after retirement. According to the authors, “neither the quality nor the quantity of food intake deteriorates with retirement status” (Aguiar and Hurst, 2005, p. 919). Thus, it is crucial to distinguish carefully between expenditures and consumption. We are aware of four studies which deal with retirement-consumption linkages for Germany. Schwerdt, 2005 and Lührmann, 2010 address the retirement-consumption puzzle directly and deal with expenditures and home production around retirement for German households. Both authors agree that consumption expenditures fall and home production rises. Based on the German Income and Consumption Survey (Einkommens- und Verbrauchsstichprobe – EVS) between 1991/92 and 2001/02, Lührmann (2010, p. 241) finds “an expenditure drop of 17% of pre-retirement expenses on non-durable consumption in Germany” which is associated with an increase of time used for home production – by 89 min per day. She concludes that “households flexibly adapt to the change in time and money resources in retirement” (Lührmann, 2010, pp. 241-242). Two other studies on Germany, by Burzig and Herrmann (2012) and Drescher and Roosen (2013), suggest that it matters to distinguish between the situations before and after retirement and those with and without retirement. In the before-and-after comparison, other variables like income, age and health may alter as well. The with-and-without comparison captures the pure retirement effect. Burzig and Herrmann (2012) explore how food-at-home and food-away-from-home expenditures of German households in the generation 50+ are affected by income and socio-demographic factors, including age and retirement. Analyses with the ‘Survey of Health, Age and Retirement’ (SHARE) database reveal that per-capita expenditures for food consumption at home increase with retirement whereas the macroeconomic studies cited above measure a drop in consumption or food expenditures. The authors derive a pure retirement effect from the SHARE data by controlling for age and income. Rising food-at-home expenditures are consistent with a substitution of work-related food expenditures by home production, which will lead to more grocery purchases as inputs for higher food production at home. Many other sociodemographic variables including age, income and perceived health affect the pattern of food expenditures significantly, too. In applying cohort analysis, Drescher and Roosen (2013) analyze the determinants of German food expenditures per capita econometrically1. The authors argue that period and cohort effects may be as important as the age effects stressed in the life-cycle theory of consumption. For six years of the German Income and Consumption Survey, between 1978 and 2003, Drescher and Roosen (2013) elaborate significant age, period and cohort effects on food-at-home and food-away-from-home expenditures. Additionally, they identify significant impacts of the pensioner variable, gender, occupation, household composition and regions on the expenditure categories. Drescher and Roosen also derive a significantly positive impact of retirement on food-at-home expenditures. In summary, the available literature on Germany shows that implications of retirement on consumption expenditures, food expenditures, and time use for home production have been analyzed. What is lacking is a detailed analysis of changes in all food-related time-use patterns due to retirement and a comprehensive analysis of determinants affecting food-related time use. We will provide such an analysis in the following sections: (i) We elaborate the time-use pattern of the German generation 50+ with regard to food production in the households, food consumption at home and away from home as well as shopping. Firstly, time-use patterns of retired and non-retired persons and households are compared statistically. Secondly, the implications of retirement on time-use are elaborated under ceteris-paribus conditions within multivariate analyses. (ii) Like Lührmann (2010), we utilize data from the German Time Use Surveys but with a different focus. Whereas Lührmann concentrated on aggregate time use for home production, we disaggregate time use for food production and consumption in the household, for food consumption away from home and shopping. (iii) Apart from the influence of retirement, we intend to explain food-related time use within the generation 50+ econometrically. We derive how other household characteristics and sociodemographic variables affect time use in order to draw general lessons on food consumption behavior of the German elderly. The article is organized as follows. After this introduction and literature review, we explain the data in Section 2 and provide statistical evidence on how food-related time-use alters with retirement. In Section 3, the methodology of a multivariate analysis of the determinants of time use for food production, food consumption at home and away from home, and for shopping is explained and quantitative results from double-hurdle and multiple regression models are presented. In Section 4, we discuss these results in the context of the literature on the retirement-consumption puzzle, and draw conclusions in Section 5. Data and descriptive statistics on consumption and household production of food The data used in this paper are drawn from the cross-sectional German 2001/02 Time-Use Survey (TUS). It is the second and most recent TUS provided by the German Federal Statistical Office; the first dates back to 1991/92. Both are regarded as “one of the most comprehensive time use studies in Germany” (Destatis, 2012). The German 2001/02 TUS was conducted as a representative quota sample of private households throughout Germany2. The German microcensus formed the basis for the quotation and projection. The data were collected in written form via time-use diaries, personal questionnaires and household questionnaires. Time-use diaries were kept by all household members aged ten years or older. For a three-day period (including one weekend day), the entire 72 h were documented in ten-minute intervals. In the diaries, the main activity, the side activity, the individuals present while performing the indicated activity, and the location or means of transportation used were recorded. Sociodemographic and socioeconomic characteristics of diary-keeping individuals and surveyed households were collected through personal and household questionnaires3, respectively. Altogether, approximately 5400 private households, 12600 individuals and 37700 diary days were covered (Statistisches Bundesamt (Ed.), 2005, pp. 1 et seq.)4. Subject to the present investigation are household heads aged 50–80 years and, if applicable, the respective spouse or life partner (regardless of age), hereinafter collectively referred to as household. However, only households for which a complete set of information is available are included in the analytical sample. That is, households with lacking information on either of the considered household members or on certain household characteristics or with incomplete time use diaries were dropped altogether from the sample5. After imposing these restrictions, an analytical sample of 2020 private households, 3304 male and female individuals and 9902 diary days was obtained. Basic information in the database refers to whether or not households participate in the activities Food Production at Home (DFPAH), Food Consumption At Home (DFCAH), Food Consumption Away From Home (DFCAFH) and Shopping (DSHOP). At the individual level, participation or non-participation is captured by a dummy variable being unity if an individual in the household participates in the activity and zero otherwise. Moreover, the amount of time used for the activities Food Production At Home (FPAH), Food Consumption At Home (FCAH), Food Consumption Away From Home (FCAFH) and Shopping (SHOP) is measured. Later, in Section 3, these variables will be used as dependent variables in multivariate analyses. The effect of retirement on all these activities will then be analyzed. Hence, a dummy variable, i.e. RETIRED, is used which is equal to one if the individual is retired and zero otherwise. The definition of retirement is based on an individual’s social status. It is a rather strict classification, since any pattern of labor market participation during retirement is completely ruled out. This, however, enables the identification of the maximum household production potential due to retirement. For comparison, the time use of unemployed (UNEMPLOYED) and non-working (NON_WORKING) 6 individuals is also taken into account, captured through corresponding dummy variables which are defined according to the reported social status. An empirical overview of the nature of the relationship between the labor market status and the activities DFPAH, DFCAH, DFCAFH and DSHOP as well as FPAH, FCAH, FCAFH and SHOP is provided in Fig. 1 and Fig. 2, respectively.Fig. 1 plots the sample’s degree of participation in the activities. It reveals that participation in home-food-production-related activities (i.e. DFPAH and DSHOP) is expanded, whereas participation in activities related to out-of-home food purchases (i.e. DFCAFH) is reduced, as the labor market is left (i.e. for unemployed, retired and non-working individuals). On the contrary, no major difference is evident in DFCAH between the four types of labor market status. It is as high as 95% or more for all groups. Fig. 2 illustrates FPAH, FCAH, FCAFH and SHOP, i.e. the average amount of time devoted by working, unemployed, retired and non-working individuals to the four activities. One key message can be drawn from Fig. 2: All types of non-labor market participation (i.e. unemployed, retired and non-working individuals) are associated with, on average, more time being spent on all four activities considered, compared to working individuals. Bringing together the findings of Fig. 1 and Fig. 2, it can be concluded that, overall, the importance of the activities Food Production At Home and Shopping grows substantially as individuals exit the labor market, given that both the participation rates and the actual time allocated to these activities increase strongly. The absolute amount of time devoted to the activity Food Consumption At Home is also higher when no labor market participation takes place, but the difference in time use is smaller than found for FPAH and SHOP, given that all four categories of labor market status have nearly identical participation rates. For FCAFH, it might very well be possible that the total amount of time dedicated to this activity falls, given that its frequency drops strongly. However, when the activity occurs, the actual amount of time spent on FCAFH increases starkly. Consequently, these findings support the importance of home production in explaining the retirement-consumption puzzle. However, the robustness of these preliminary results to the inclusion of additional personal, household and diary-day characteristics with a conceivable effect on the activities of interest still needs to be verified. A description of the dependent and independent variables considered and the corresponding summary statistics are presented in Table 1.The descriptive statistics for the overall sample in Table 1 might hide structural differences in time use between different social groups. Therefore, we present disaggregate information on food-related time use for various socioeconomic groups in the Appendix. Cross-tabulating DFPAH, DFCAH, DFCAFH, DSHOP, FPAH, FCAH, FCAFH, SHOP and the social status with other personal and household characteristics, on the one hand confirms some of the aggregate results in Fig. 1 and Fig. 2. Considering exemplarily the gender (MALE), the Appendix shows that both male and female retired individuals engage more in Food Production at Home (DFPAH) and Shopping (DSHOP) and less in Food Consumption Away From Home (DFCAFH) than their working counterparts, despite the traditional gender-based division of labor usual among older generations. Moreover, both male and female retired individuals spent, on average, more time on all four activities compared to male and female workers. On the other hand, the Appendix reveals some major differences in food-related time use with regard to the working status and other characteristics of the households. The post-retirement substitution of food consumption away from home by home production of food is visible for males as well as for females but the magnitude of the substitution differs by gender. In the aggregate, participation in and intensity of food production at home increases by 14 percentage points and 24 min per day, respectively, after retirement (see Fig. 1 and Fig. 2). However, the Appendix shows that it is nearly exclusively male participation which drives the rising DFPAH. DFPAH increases by 17 (5) percentage points for males (females) after retirement. On the other hand, FPAH rises by more than 20 min per day for females as well as males after retirement. With a change from 37.5 to 58.7 min per day, however, this is equivalent to a much higher relative increase of men’s time use for food production at home. Households’ DFCAFH declines after retirement, on average, from 31% to 18% (see Fig. 1). The Appendix reveals that the fall is stronger for males (by 17 percentage points) than for females (by 9 percentage points). If retirees participate in food-away-from home consumption, however, FCAFH strongly rises at retirement and more so for men (by 26.5 min per day) than for women (by 17.9 min per day). How the shift from the status “working” to the status “retired” affects food-related time use is also dependent of some other sociodemographic and personal characteristics. Retirement compared to working raises FPAH more if the following values of dummy variables hold: if persons assess their health status negatively (POOR_HEALTH = 1), if they have a lower degree of school education (HIGH_SCHOOL = 0), if they are not at risk of monetary poverty (MONETARY_POOR = 0) and if they live in Eastern Germany (WEST = 0). Moreover, under retirement, FCAFH rises more if the following values of dummy variables hold: if persons assess their health status negatively (POOR_HEALTH = 1), attained a higher degree of school education (HIGH_SCHOOL = 1), are not at risk of monetary poverty (MONETARY_POOR = 0), and if they live in Western Germany (WEST = 1).

The empirical findings allow some important conclusions on how retirement affects (i) home production of food and food consumption in general and (ii) the validity of a retirement-consumption puzzle in Germany. In general, the results indicate that elderly people alter their food-related behavior substantially after retirement. Changes appear consistent with a household-production-function approach in which basic decision parameters change. In their new life situation, as they gain flexibility and considerably more time for leisure-related activities, a strongly revised time-use pattern occurs. All empirical findings can be rationalized and, thus, we see no support for a retirement-consumption puzzle on the basis of time-use data. It is very clearly shown by the results in Table 2 that a substitution of market-related activities by home-oriented activities emerges in the context of food consumption. After retirement, the probability of consuming food away from home declines whereas the probability of engaging in food production at home rises as does the average time of doing so. These impacts are further strengthened by the significant influence of other sociodemographic characteristics of households that may – to a certain extent – be associated with retirement. An increasing AGE and the variable POOR_HEALTH additionally raise the dependent variable FPAH and diminish the likelihood that participation in food-away-from-home consumption occurs, i.e. that DFCAFH is unity. The important role of home production after retirement is confirmed by the fact that the probability of shopping and the time used for shopping is significantly raised by the variable RETIRED. Although SHOP measures the aggregate shopping time of individuals, it is very likely that the same effect would hold if a more disaggregate measure of time use for grocery shopping was available. In a methodological sense, it is the strength of the data basis and the model used that the pure effect of retirement can be distinguished from the effects of age and the perceived health status. In earlier studies, such as Aguiar and Hurst (2005), retirement had to be approximated by the age of individuals since no retirement variable was available. Given the separate impacts of the individual determinants in Table 2, it seems important to model the time use for different food-related activities each with the retirement, age and health variables. Three additional conclusions can be drawn when putting our findings into the context of recent research. Firstly, although our results seem very consistent with those of other recent studies, this is more the case for the signs than the magnitude of impacts by retirement on time use. Differences between the results by Luengo-Prado and Sanz (2013) for Spain and our results for Germany may illustrate it. Luengo-Prado and Sanz, who mainly deal with the effects of retirement on expenditure patterns, analyze time-use data, too. They show that “a substitution of market goods for home-produced goods” (ibid., p. 766) occurs and they conclude, like we do for Germany, that there is no indication of a retirement-consumption puzzle. However, the strength of the substitution processes differs widely between the two countries. The Spanish results reveal that retirement raises the time to cook per person and week by 38.6 min, for shopping by 44.9 min and eating at home by 38.7 min (Luengo-Prado and Sanz, 2013, Table 7). In our results from Table 2, retirement boosts food production (food consumption) at home by 46.1 (17.0) and shopping by 40.2 min per person and day. Whereas the impact on eating at restaurants is not significant in the Spanish study, the effect on consumption away from home is significant and large in our study (+33.8 min per day for those individuals who participate in food consumption away from home). There are methodological reasons for the larger impacts of retirement in our study. We defined all activities broader than Luengo-Prado and Sanz (2013) such as “food production at home” rather than “cooking”. Moreover, zero observations were excluded in our second-stage results of Table 2. There will also be country differences: Luengo-Prado and Sanz (2013, p. 765) report that pensions are high enough in Spain and, hence, “household income does not fall at retirement for a large fraction of households”. Moreover, as joint meals are more common in Spain than in Germany before and after retirement, there is less substitution due to retirement in Spain. Substantially different coefficients do still remain and further research is needed to clarify whether these differential findings are rather driven by methodology or by different food consumption patterns. Secondly, the strong substitution effects after retirement in our analysis suggest to reconsider the impact of retirement on food expenditures. As the time use for home production and consumption of food rises with retirement as does the time use for shopping, this might theoretically be associated with either increasing or decreasing food-at-home expenditures. When work-related out-of-home consumption declines and is substituted at least partially by at-home production of food, more foods have to be purchased as inputs for cooking and household production. Ceteris paribus, this will raise food expenditures for at-home consumption. However, time use grows not only for household production, but also for shopping. It is plausible that more use can be made of bargains and that food expenditures fall with a lower unit value. For Germany, recent studies by Burzig and Herrmann (2012) and Drescher and Roosen (2013) showed the uniform result that the pure retirement effect – apart from income and age effects – is an increase of per-capita food expenditures for at-home consumption. This seems very consistent with our findings suggesting that growing food expenditures are driven by the higher demand for food inputs which clearly overcompensate a potential, opposite effect of cheaper purchases. Thirdly, interesting questions arise from our analysis of time use for food consumption away from home. Within FCAFH, there seems to be a strong substitution of work-related by leisure-related away-from-home consumption. Apparently, DFCAFH falls with retirement. This is very plausible as many occasions of work-related food consumption away from home disappear. On the other hand, it is striking that the time per visit rises with retirement in the context of leisure-related FCAFH. This effect is statistically significant, but only at the 10%-level of statistical significance. There is a large standard deviation of time use for FCAFH. The empirical findings suggest that more detailed analyses are needed as a very positive impact of retirement on FCAFH seems to exist for some segments of retirees. It is likely that very active segments of retirees may play a crucial role for the food service and hotel sectors, particularly in some touristic regions.