Lecture 4: Home Production and Labor Supply Part 1: Labor Supply and Home Production Simple Labor Supply Example: No Home Production Look at static model: C 1 N 1 U (C , N ) d 1 1 s.t. C wN F .O.C. UC : C UN : dN w 3 Simple Labor Supply Example: No Home Production UN : dN w Take Logs: ln( N ) 1 ln( w) 1 ln( ) 1 ln( d ) Estimate : ln( N ) A ln( w) ln( ) = labor supply elasticity with respect to wages (holding constant) = labor supply elasticity with respect to marginal utility of wealth (holding w constant) 4 How Do Things Change With Home Production? C 1 ( N H )1 U (C , N ) d 1 1 s.t . X wN C ( H H X X )1/ F .O.C . UC : UH : UN : X C 1 X 1 H C 1 H 1 d ( N H ) d ( N H ) w How Do Things Change With Home Production? Similar in spirit to no-home production model ln( N H ) 1 ln( w) 1 ln( ) 1 ln( d ) But, the elasticity of market work changes: H 1 ln( H ) ln( N ) ln( w) ln (1 ) (1 ) X Interpretation • Home production makes work hours more elastic to changes in wages (holding the marginal utility of wealth constant). • Implications: Women’s labor supply more elastic than men (if they do most of the home production) (Mincer 1962) Labor supply is more elastic during temporary wage changes (recessions) with home production. Expenditure (X) is more elastic during temporary wage changes (recessions) with home production. • Last two implications are flushed out in Benhabib, Rogerson, and Wright’s “Home Work in Macroeconomics: Household Production and Aggregate Fluctuations” (JPE, 1991) Part 2: Review of Trends in Leisure and Leisure Inequality Based on Aguiar and Hurst (2007, 2009) Measures of Economic Inequality Have Increased Recently • Wages (See Katz and Autor 1999) • Consumption (Attanasio and Davis 1996 ; Attanasio et al. 2007) • Consensus: “High Educated” standard of living has increased dramatically relative to “Low Educated” standard of living since early 1980s Implications for Changing Inequality in Well Being • Are expenditure differences sufficient to make welfare comparisons? • Usual assumptions: Individuals gain utility from both market expenditures (i.e., consumption) and “leisure” time: U(c,l) Consider the following trade-off Job 1: Job 2: Earn $120,000 a year working 50 hours per week Earn $100,000 a year working 30 hours per week • Some people would choose job 2, even though the earnings (and corresponding market expenditures) would be lower than job 1. • People “value” time. What We Do in this Paper • Explore the changing nature of the allocation of time over the last 40 years. – Focus on the aggregate trends. – Examine the changing nature of “leisure inequality”. • Ask a related question: Can changing educational differences in employment status explain changing leisure inequality? • Why is that interesting? In terms of welfare implications, it is important to know whether low education individuals are taking more leisure because they are unable to find employment at their reservation wage. (Individuals will be off their labor supply curve). Key Finding • Leisure inequality has increased dramatically since 1985 – • Relative to high educated men, low educated men have gained an additional 7 hours per week of leisure! At most, only 40% of the increase in leisure inequality can be explained by the potential of involuntary unemployment (including disability) of low educated men. – This is an upper bound. Some low educated men are out of the labor force by choice. Some high educated men are involuntarily unemployed. • Take Away: Need to think about how to measure changing relative well being between high and low educated men since 1985. Why do we care about distinguishing between non-market uses of time? • To start, we care a lot about labor supply elasticities (or the structural parameters that underlie the elasticities)? • Standard story (with no home production): – Labor supply elasticities are governed by parameters of the utility function Substitution Effect: Wages increase, work more (as wages increase, substitute away from leisure towards working) Income Effect: • Wages increase, work less (effectively richer, so can afford to take more leisure). Constant aggregate market work hours during the last forty years as wages have increased has been interpreted as income and substitution effects (determined by these preference parameters) canceling. 13 Labor Supply Elasticities and Home Production • Interpretation Allowing for Home Production – Labor supply elasticities are governed by both parameters of the utility function and parameters of the commodity production technologies: c = f(h,x) – Additional substitution and income effects via home production technologies. Substitution Effect: Wages increase, home production time is expensive, switch away from time in home production (towards expenditure). Income Effect: – Wages increase, home produce more (effectively richer, so want more consumption good, need more home production time). 14 Note, by definition, Leisure goods have little (if any) substitution effects Implications • To the extent that time spent in “home production” changes over time or differs across people, estimated labor supply elasticities will: – – Change over time (even if preferences are constant) Differ across people (even if they have the same preferences) • Constant aggregate market work hours need not imply that income and substitution effects (via preferences) cancel. • For those doing home production, rising market work hours could still imply that income effects (via preferences) dominate substitution effects (via preferences). <<Mincer (1962), Becker (1965), Gronau (1977)>> • Can learn about preferences, the home production technologies and the leisure technologies by jointly examining the behavior of home production, leisure, and market work. 15 The Data (Table 1) • 1965-1966: Americans’ Use of Time 2,001 individuals Aged 19-65 One household member must be working in last year Only one person per household is surveyed 24 hour recall of previous day/ Lots of additional demographic information • 1975-1976 Time Use in Economic and Social Accounts 2,406 adults (1519 households) Interviews both husbands and wives (same household) Interviews them four times (once per quarter) Designed to be nationally representative 24 hour recall of previous day/ Lots of demographic and earnings data Note: We only use first interview (fall 1975) 16 The Data (Table 1) • 1985 Americans’ Use of Time 4,939 adults (over the age of 18) One adult per household Designed to be nationally representative 24 hour recall of previous day Limited demographics • 1992-1994 National Human Activity Pattern Survey (sponsored by the EPA) 9,386 individuals (7,514 adults over the age of 18) One person per household Designed to be nationally representative 24 hour recall of previous day Limited demographics 17 The Data (Table 1) • 2003 American Time Use Survey (BLS) Over 20,000 individuals One person per household Designed to be nationally representative 24 hour recall of previous day Very detailed demographics Sample is drawing from exiting CPS main sample (after survey month 8) Only have time use linked to actual wages in 2003 Note: 2004 data is not available from BLS (discuss results throughout the talk) Two problems? Much finer time use categories One of goals is to create better measures of time spent with children. Some comfort: 1993 data and 2003 data are very similar along many dimensions 18 Some Existing Work on Time Use • Juster and Stafford (1985, 1991) and Robinson and Godbey (1997) – – Analyze 1965, 1975, and 1985 time diaries Present unconditional means (mostly) – * Robinson and Godbey also analyze a small 1995 pilot time use survey in their last chapter of second edition of their 1997 book 1995 sample does not match well with either 85 or 03 survey. Focus on 65 – 85 trends – – • What we do is: – – – – Extend through 03 Harmonize the data in consistent manner Adjust for differences in sample composition between surveys Also show conditional means. 19 Creating consistent measures of Time Use • • For the 1965, 1975, 1985, and 1993 data, it was relatively easy Classifying activities in 2003 was a bit harder Some codes for 1985 (time spent in): Act10 Act11 Act12 Act14 Meal preparation, cooking, and serving food Meal cleanup, doing dishes Cleaning house (dusting, vacuuming, cleaning bathrooms, etc.) Laundry, Ironing, Clothes Care (sewing, mending, etc.) Some codes for 1993 (time spent in): Act10 Act11 Act12 Act14 Meal preparation, cooking, and serving food Meal cleanup, doing dishes Cleaning house (dusting, vacuuming, cleaning bathrooms, etc.) Laundry, Ironing, Clothes Care (sewing, mending, etc.) 20 Sample • All non-retired individuals between the age of 21 and 65 (inclusive) – – 1965 time use survey excludes retired households. 1965 survey only includes individuals up until the age of 65 • Restrict individuals to have a “full” time use report (1440 minutes/day) • Throughout the talk: – – • All individuals By sex, education, marital status, and employment status All results are presented in units of “Hours per Week” 21 Are Time Use Samples Representative (Table A1)? • Compare males in time use data to males in PSID (weighting both data sets). Restrict sample: Age 21 – 65, non-retired 1965 1975 1985 1993 2003 Time PSID Time PSID Time PSID Time PSID Time PSID Age 20s 0.25 0.21 0.27 0.30 0.27 0.23 0.25 0.18 0.20 0.16 Age 30s 0.23 0.25 0.28 0.24 0.32 0.33 0.31 0.33 0.26 0.27 Age 40s 0.26 0.27 0.20 0.24 0.20 0.20 0.25 0.30 0.28 0.31 Age 50s 0.19 0.19 0.19 0.18 0.16 0.18 0.15 0.15 0.20 0.21 Age 60s 0.07 0.08 0.06 0.05 0.05 0.05 0.04 0.05 0.06 0.05 Ed > 12 0.30 0.28 0.30 0.39 0.46 0.49 0.58 0.54 0.55 0.59 Married 0.87 0.89 0.85 0.85 0.69 0.76 ---- 0.71 0.69 0.70 Have Kid 0.65 0.65 0.55 0.60 0.42 0.51 0.32 0.46 0.42 0.45 # of Kids Employed 1.57 0.97 1.66 0.96 1.24 0.93 1.30 0.93 0.76 0.88 0.96 0.90 ---0.89 0.89 0.91 0.80 0.88 0.86 0.91 • • Note: 30/40 year olds have increased 1965 to 2003 Note: Population is becoming more educated between 1965 and 2003 22 Are Time Use Samples Representative? Allocation of women with children by day of week 1965 1975 1985 1993 2003 Monday .115 Tuesday .169 .139 .164 .159 .128 .126 .133 .151 .140 .139 .143 .137 .156 .154 .140 .149 .147 .130 .144 .135 .188 .129 .132 .123 .097 .152 .179 .140 .136 .151 .140 .142 .143 .148 Wednesday Thursday Friday Saturday Sunday • • Data weighted using survey “weights” to make the sample representative by day of the week! If random, each cell should have a value equal to 0.142 23 Definitions: Time Spent in Market Production (Table A2) 1. “Core Market Work” – Time spent working for pay on all jobs (Main job, other jobs, overtime) Analogous to measure of hours worked in PSID 2. “Total Market Work” - Direct market work, plus commuting to work, plus ancillary work activities Ancillary work activities includes time at work “off the clock” (mandatory breaks, meals at work) 24 Figure 1: Comparison of Weekly Core Market Work Hours in PSID and Time Use Surveys: Sample: All Non-Retired Men Between Ages of 21 and 65 46.00 42.00 40.00 38.00 36.00 20 03 20 01 19 99 19 97 19 95 19 93 19 91 19 89 19 87 19 85 19 83 19 81 19 79 19 77 19 75 19 73 19 71 19 69 19 67 34.00 19 65 Hours Per Week 44.00 Year PSID Work Hours Time Use Work Hours 25 Time Use Categories (Table A1) • Market Work: Paid work in formal sector Paid work in informal sector Job search • Non-Market Work: Home and vehicle maintenance Shopping/Obtaining goods and services All other home production (cooking, cleaning, laundry, house work) • Child Care • Gardening, Lawn Care, Pet Care Note: All associated travel time is embedded in the time use category Time Use Categories (continued) • Leisure TV watching Socializing Exercise/Sport Reading Hobbies/Other Entertainment Eating Sleeping Personal Care • • • • • Other Medical Care Care of Other Adults Religious/Civic Activities Education Other Trends in the Allocation of Time (Men): Table 1 Changes Over Time (Adjusted for Demographics) 05-65 85-65 05-85 Total Market Work -11.7 -7.7 -4.0 Non Market Work 3.5 4.3 -0.8 Child Care 1.8 0.0 1.8 Leisure 4.7 4.3 0.4 Trends in the Allocation of Time (Men): Table 1 Changes Over Time (Adjusted for Demographics) 05-65 85-65 05-85 Total Market Work -11.7 -7.7 -4.0 Non Market Work 3.5 4.3 -0.8 Child Care 1.8 0.0 1.8 Leisure 4.7 4.3 0.4 Trends in the Allocation of Time (Women): Table 1 Changes (Adjusted for Demographics) 05-65 85-65 05-85 Total Market Work 3.4 1.2 2.1 Non Market Work -10.4 -6.1 -4.3 Child Care 1.8 -0.8 2.6 Leisure 3.3 6.4 -3.1 Trends in Leisure by Sub-Aggregate: ALL 10 8 tv 6 leisure 2 Hours per Week 4 sleeping + personal care 2 gardening sports entertainment 0 1965 1975 1985 1993 eating 2003 hobbies -2 reading -4 -6 socializing Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 2.1 1.5 13.3 8.2 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 2.1 1.5 13.3 8.2 (1) 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 2.1 1.5 13.3 8.2 (2) 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 (3) <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 2.1 1.5 13.3 8.2 Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 (4) <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 2.1 1.5 13.3 8.2 Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 (5) <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 2.1 1.5 13.3 8.2 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 Time Allocation By Education (Leisure Dispersion): Men Changes adjusted for demographics 65 85 03-05 05-65 104.9 107.3 104.1 105.8 113.0 107.9 104.4 99.7 8.7 6.7 5.8 -2.2 2.1 1.5 13.3 8.2 85-65 05-85 0.5 6.1 5.5 3.9 8.1 0.6 0.3 -6.1 < 12 12 13-15 16+ 104.3 101.2 98.6 101.9 <12 vs. 16+ 12 vs. 16+ 2.4 -0.7 Question: Is the dispersion driven by the changing pool of individuals within each educational category? General Increase in Leisure Dispersion 20 15 Hours per Week 90th Percentile 10 75th Percentile 5 50th Percentile 25th Percentile 0 1965 -5 1975 1985 1993 2003 10th Percentile Summary of Trends • Leisure increased dramatically since 1965 for average individual • Most of the average increase occurred prior to the 1990s • There is a large increase in leisure dispersion that also occurred during this period. Most of that occurred post 1985 (particularly for men). • Note: The timing of the increase in leisure inequality matches the timing of the well documented increase in consumption inequality and wage inequality. Remaining Questions • Can the increase in leisure for low educated men be interpreted as an increase in well being? Set out to answer four new questions: 1. Conditional on working full time, is there an educational gap in leisure in either 1985 or 2003? 2. How do men who do not work, regardless of education, allocate their foregone market work hours? 3. Is there an educational gap in leisure for the unemployed? the disabled? other non-employed? 4. How much of the increased leisure dispersion across education groups can be explained by changes in employment status by education? Employment Status By Education Low Ed High Ed Conditional Difference 1985 Share Employed 1985 Share Non-Employed Unemployed Other Non-employed 0.89 0.11 0.04 0.07 0.94 0.06 0.02 0.04 -0.04 0.04 0.02 0.02 03-05 Share Employed 03-05 Share Non-Employed Unemployed Disabled Other Non-employed 0.83 0.17 0.05 0.08 0.04 0.92 0.08 0.04 0.02 0.03 -0.09 0.09 0.02 0.05 0.02 Note: From now on, we only focus on two education groups (because of small sample sizes in some cells). Employment Status By Education Low Ed High Ed Conditional Difference 1985 Share Employed 1985 Share Non-Employed Unemployed Other Non-employed 0.89 0.11 0.04 0.07 0.94 0.06 0.02 0.04 -0.04 0.04 0.02 0.02 03-05 Share Employed 03-05 Share Non-Employed Unemployed Disabled Other Non-employed 0.83 0.17 0.05 0.08 0.04 0.92 0.08 0.04 0.02 0.03 -0.09 0.09 0.02 0.05 0.02 Note: From now on, we only focus on two education groups (because of small sample sizes in some cells). Time Allocation By Education: All Men 2003-2005 Low Ed 36.9 High Ed 41.9 Total Non-Market Work 10.9 11.7 -0.7 Child Care Gardening, Lawn Care, Pet Care 2.7 2.2 3.4 2.1 -0.7 0.2 Total Leisure T.V. 109.8 21.6 102.3 15.3 7.1 6.0 0.8 1.7 1.5 0.7 1.4 1.9 0.1 0.2 -0.4 Total Market Work Own Medical Care Care of Other Adults Religious/Civic Activities Difference* -4.6 Time Allocation By Education: Employed Men 2003-2005 Low Ed 44.5 High Ed 45.5 Total Non-Market Work 10.0 11.1 -1.0 Child Care Gardening, Lawn Care, Pet Care 2.6 2.2 3.4 1.9 -0.7 0.2 Total Leisure T.V. 104.1 18.4 100.1 14.3 3.9 4.0 0.5 1.6 1.3 0.6 1.3 1.8 -0.1 0.3 -0.5 Total Market Work Own Medical Care Care of Other Adults Religious/Civic Activities * Conditional on Demographics Difference* -0.9 Time Allocation By Education: Employed Men 2003-2005 Low Ed 44.5 High Ed 45.5 Total Non-Market Work 10.0 11.1 -1.0 Child Care Gardening, Lawn Care, Pet Care 2.6 2.2 3.4 1.9 -0.7 0.2 Total Leisure T.V. 104.1 18.4 100.1 14.3 3.9 4.0 0.5 1.6 1.3 0.6 1.3 1.8 -0.1 0.3 -0.5 Total Market Work Own Medical Care Care of Other Adults Religious/Civic Activities * Conditional on Demographics Difference* -0.9 Time Allocation By Education: Unemployed Men 2003-2005 Low Ed High Ed Difference* Total Market Work Job Search Education 3.0 2.4 0.9 3.8 5.5 2.1 -0.5 -2.9 -1.2 Total Non-Market Work 18.7 19.2 -0.1 Child Care Gardening, Lawn Care, Pet Care 4.4 2.3 4.2 4.5 -0.5 -2.2 127.9 29.7 121.5 22.2 5.5 7.5 0.6 3.0 2.4 0.5 2.4 2.6 0.2 0.8 0.1 Total Leisure T.V. Own Medical Care Care of Other Adults Religious/Civic Activities Time Allocation By Education: Unemployed Men 2003-2005 Low Ed High Ed Difference* Total Market Work Job Search Education 3.0 2.4 0.9 3.8 5.5 2.1 -0.5 -2.9 -4.6 -1.2 Total Non-Market Work 18.7 19.2 -0.1 Child Care Gardening, Lawn Care, Pet Care 4.4 2.3 4.2 4.5 -0.5 -2.2 127.9 29.7 121.5 22.2 5.5 7.5 0.6 3.0 2.4 0.5 2.4 2.6 0.2 0.8 0.1 Total Leisure T.V. Own Medical Care Care of Other Adults Religious/Civic Activities Where Did the Foregone Work Hours Go (in percent)? Low Ed 6.7 5.2 0.0 High Ed 8.4 11.9 4.0 Total Non-Market Work 19.6 17.8 Child Care Gardening, Lawn Care, Pet Care 4.0 0.2 1.8 5.7 53.5 25.4 12.6 12.6 -0.7 47.0 17.4 8.4 10.1 8.6 Total Market Work Job Search Education Total Leisure T.V. Socialization Sleeping Other Entertainment/Hobbies Where Did the Foregone Work Hours Go (in percent)? Low Ed 6.7 5.2 0.0 High Ed 8.4 11.9 4.0 Total Non-Market Work 19.6 17.8 Child Care Gardening, Lawn Care, Pet Care 4.0 0.2 1.8 5.7 53.5 25.4 12.6 12.6 -0.7 47.0 17.4 8.4 10.1 8.6 Total Market Work Job Search Education Total Leisure T.V. Socialization Sleeping Other Entertainment/Hobbies Where Did the Foregone Work Hours Go (in percent)? Low Ed 6.7 5.2 0.0 High Ed 8.4 11.9 4.0 Total Non-Market Work 19.6 17.8 Child Care Gardening, Lawn Care, Pet Care 4.0 0.2 1.8 5.7 53.5 25.4 12.6 12.6 -0.7 47.0 17.4 8.4 10.1 8.6 Total Market Work Job Search Education Total Leisure T.V. Socialization Sleeping Other Entertainment/Hobbies 24% Time Allocation By Education: Disabled Men 2003-2005 Low Ed High Ed Difference* Total Market Work Job Search Education 0.0 0.0 0.2 0.7 0.2 1.6 -0.7 -0.2 -1.7 Total Non-Market Work 10.6 12.8 -1.8 Child Care Gardening, Lawn Care, Pet Care 2.5 2.2 2.0 1.3 0.2 1.0 144.1 43.2 138.7 36.0 5.7 7.5 4.3 1.5 2.2 4.6 2.5 2.1 -0.5 -1.4 0.1 Total Leisure T.V. Own Medical Care Care of Other Adults Religious/Civic Activities Where Did the Foregone Work Hours Go (in percent)? Total Market Work Education Low Ed 0.0 -1.6 High Ed 1.5 0.2 Total Non-Market Work -1.6 2.9 Child Care Gardening, Lawn Care, Pet Care 1.3 -0.2 3.7 -3.1 Total Leisure T.V. Socialization Sleeping Other Entertainment/Hobbies 89.9 55.7 7.9 19.1 5.6 84.8 47.7 6.6 24.8 4.2 Own Medical Care 8.5 8.8 Time Allocation By Education: Other Men 2003-2005 Low Ed High Ed Difference* Total Market Work Job Search Education 0.8 0.0 0.8 2.0 0.3 0.9 -1.0 -0.3 -0.1 Total Non-Market Work 17.5 20.1 -3.4 Child Care Gardening, Lawn Care, Pet Care 4.0 3.0 4.5 5.0 -0.4 -1.4 135.2 32.9 124.6 24.6 9.8 8.5 2.3 2.5 3.6 -1.0 0.0 -0.8 Total Leisure T.V. Own Medical Care Care of Other Adults Religious/Civic Activities 1.4 2.2 2.5 Where Did the Foregone Work Hours Go (in percent)? Total Market Work Job Search Education Low Ed 1.8 -0.2 -0.2 High Ed 4.4 0.4 1.3 Total Non-Market Work 16.9 19.8 Child Care Gardening, Lawn Care, Pet Care 3.1 1.8 2.4 6.8 Total Leisure T.V. Socialization Sleeping Other Entertainment/Hobbies 69.9 32.6 8.5 18.7 5.8 53.8 22.6 9.2 14.7 2.9 2003-2005 Cross Sectional Decomposition • How much of the difference in leisure between high and low educated men in 2003-2005 is due to differences in job status? Perform a Blinder-Oaxaca Decomposition: Define Wjk = probability of being in job status k for educational attainment j Xjk = hours per week of leisure for individual in job status k and educational attainment j. Conditional Difference: 7.5 Hours Per Week (WL – WH) XH WL(XL – XH) 2.4 Hours Per Week 5.1 Hours Per Week • (vectors): (vectors): Roughly 30% of difference in leisure in 2003-2005 between low and high educated men can be attributed to employment status differences. Perform Same Analysis for 1985 Leisure Unconditional Difference Low Ed High Ed All 107.4 105.1 2.2 Employed Men Non-Employed Men 103.9 134.6 103.5 130.0 0.4 4.6 Perform a similar Blinder-Oaxaca decomposition • Roughly 60% of difference in leisure in 1985 between low and high educated men can be attributed to employment status differences. Perform Same Analysis for 1985 Leisure Unconditional Difference Low Ed High Ed All 107.4 105.1 2.2 Employed Men Non-Employed Men 103.9 134.6 103.5 130.0 0.4 4.6 Perform a similar Blinder-Oaxaca decomposition • Roughly 60% of difference in leisure in 1985 between low and high educated men can be attributed to employment status differences. Perform Same Analysis for 1985 Leisure Unconditional Difference Low Ed High Ed All 107.4 105.1 2.2 Employed Men Non-Employed Men 103.9 134.6 103.5 130.0 0.4 4.6 Perform a similar Blinder-Oaxaca decomposition • Roughly 60% of difference in leisure in 1985 between low and high educated men can be attributed to employment status differences. Time Series Decomposition (85-05) Change Less Educated More Educated • 2.5 -2.8 (W05-W85)X05 2.0 0.6 W85(X05-X85) 0.4 -3.4 Percent Explained 0.82 <0.00 How much of the overall dispersion (combining cross section and time series) can be explained by changing employment status? Answer: ~ 40% • Conclusion: If all non-employment is involuntary for low educated men, 60% of the documented leisure dispersion remains. • Low educated men are still “choosing” to take more leisure than high educated men over last 25 years. Implications for Changing Inequality #1 • How does one value the additional leisure time? If individuals are on their labor supply curve, we can use their wage to value their increased leisure time. • Back of the envelop calculation: Approximately 4 to 7 hour increase in leisure per week for low educated men relative to high educated men since the mid 1980s. After tax low educated wage ~14 hours per hour. Value of the additional leisure time: $3,000 - $5,000 a year. • Is this large? Implications for Changing Inequality #2 • Provides a caution for interpreting measures of consumption inequality. Time can be allocated to “home production” which can cause expenditure to diverge from true consumption. Examples: Shopping intensity Take advantage of time dependent discounts Cooking meals Do their own home production • The unemployed do allocate more time to home production/shopping than their employed counterparts. • Changes in employment propensities over time can be expected to change the mix of market expenditures and time that enter the commodity production function. (Aguiar and Hurst 2005, 2007a, 2007b) Broader Implications • Why do low educated men choose higher leisure relative to higher educated men? 1) Do wages differences cause the leisure differences? – Substitution effects are important? 2) Or are preference differences driving the leisure differences? There are stark differences in behavior among the non-employed. - Perhaps those with a taste for leisure are sorting are the ones sorting into the low educated category. One Last Point: Within Education Dispersion 25 <12 12 13-15 16+ 20 Change 1965 - 2003 Hours per Week 15 10 5 0 5 15 25 35 45 55 -5 -10 Percentile of Distribution 65 75 85 95 Conclusions (Update) • The allocation of time has changed dramatically over the last 40 years. • The allocation differed dramatically by educational attainment with low educated individuals experiencing larger “leisure” increases than high educated individuals. • Only about 40% of the dispersion can be explained by involuntary nonemployment. Part 3: Other Data I am Thinking About Do Income Effects Dominate Substitution Effects? • I am not sure 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 50 Men, Unconditional (U.S. CPS) 45 40 35 30 25 20 Series2 Series3 Series4 Series5 Series6 Growth in Hours vs Growth in Per Capita Income .2 Men 25-65 Annual Hours Unadjusted WV .1 PA 0 NH AL AR MS -.2 -.1 DE RI CT OH MA WA NJ IL MI ME IN UT KY CA CO VA WY MD NY TN OK KS NC OR NV MOAZ TX LA NM MT SC GA ID MN NE ND SD WI IA FL VT .5 1 1.5 Growth in Per Capita Income Fitted values gr_annual_hours_m_2_40_80 2 Growth in Participation vs Growth in Per Capita Income 0 Men 25-65 Participation Unadjusted 1940 - 1980 CO NH WY NE -.05 DE UT CT IL IAMN MA WI WA IN NJ CA MD NV OH RI ID MEOR MO NY PAMTVT AZ MI KS SD VATX ND OK NM LA NC GA TN SC -.1 WV FL KY AL MS -.15 AR .5 1 1.5 Growth in Per Capita Income Fitted values gr_working_m_2_40_80 2 Growth in Hours vs Growth in Per Capita Income Men 25-65 Annual Hours Unadjusted 1980-2000 -.05 0 FL VA MD MI ARPA GA TN VT NJ NH CO IL AZ OH CT NC WI IN WA RI MNNY DE MO KYNE CA OR IA AL ME ID KS UT SC ND TX NV -.1 OK LA MT MS SD NM WV -.15 WY .1 .2 .3 .4 Growth in Per Capita Income Fitted values gr_annual_hours_m_2_80_00 .5 MA Growth in Participation vs Growth in Per Capita Income Men 25-65 Participation Unadjusted 1980-2000 -.04 -.02 VT IA FL -.06 WI UT WA OR AR PA KS ID MO OH DE IN -.08 AZ MN NE ND RI MD VA MT OK ME IL -.1 NV -.12 WY TN NC CT GA NJ KY CA TX MA SD CO MI NH AL NY SC MS LA NM WV .1 .2 .3 .4 Growth in Per Capita Income Fitted values gr_working_m_2_80_00 .5 Do Income or Substitution Effects Dominate on Labor Supply Decisions? Part 4: Estimating Home Production Functions Aguiar and Hurst (AER 2007) Some Preliminaries • Within most economic models, individual well being is usually measured as some function of consumption (c) and leisure (l) (i.e., U(c,l) ) • Empirically: c is always measured as market expenditures (in dollars) l is usually measured as time spent away from market work • To the extent that non market activities are important (i.e., shopping and home production), the empirical measurements of c and l may not map directly into their theoretical counterparts. 85 An Example: Shopping • Expenditure is price (p) * quantity (q) <<our measure of consumption>> • Shopping is time intensive but it may affect prices paid (holding quantities constant) • Given that time is an input into shopping, the opportunity cost of one’s time should determine how much an individual shops. – Those whose time is less valuable should shop more and, all else equal, pay lower prices (holding quantities constant) • A similar story could be told for home production • By focusing on expenditure as sole measure of consumption, researchers will make false conclusions about individual well being. – For example, declines in expenditure at the time of retirement 86 What We Do in This Paper • Use new scanner data (on household grocery packaged goods) to document: – Prices paid differs across individuals for the same good – Price paid varies with proxies for cost of time. • Use this micro data to actually estimate household shopping functions which relate prices paid to shopping intensity. – This shopping function will give us the implied opportunity cost of time for the shopper • Given margin conditions, we can use the shopping function and time use data on home production to estimate the home production technology. • Show empirically that the ratio of consumption to expenditure varies over the lifecycle 87 Scanner Data on Prices • Note: In this data part of the paper, we will only be talking directly about food consumptions and expenditures (in model, we will extend the implications) • Data is from AC Nielson HomeScan – Panel of households – Random sample within the MSA of households • – The survey is designed to be representative of the Denver metropolitan statistical area and summary demographics line up well with the 1994 PSID Coverage at several types of retail outlets 88 Scanner Data (continued) • • • Each household is equipped with an electronic home scanning unit Each household member records every UPC-coded food purchase they make by scanning in the UPC code After each shopping trip, household records: – What was purchased (i.e. scan in UPC code) – Where purchase was made (specifically) – Date of purchase – Discounts/coupons (entered manually) • AC Nielson collects the price data from all local shopping outlets. • Data has decent demographics (income categories, household composition, employment status, sex, race, age of members, etc.). Collected annually. 89 Sample • We have access to the Denver data for the years 1993-1995. – Short panel • Sample: – 2,100 households (focus on age of shopper between 24 and 75) – 950,000 transactions – 40,000 household/month observations. 90 How am I going to Use the Data • Derive a price index using the scanner data • Show some unconditional means of how this price index varies across differing income and demographic groups • Think about measurement issues relating to our estimate of the price index • Goal is to get estimate shopping and home production functions that I could import into our model 91 Potential Measurement Issue 1: Underreporting • Average monthly expenditure in the data set: $176/month (1993 dollars) • Average total food “at home” in the PSID for similarly defined sample (1993 dollars) is $320 (55% coverage rate in the HomeScan Data) • Differences between the coverage due to: – Omission of certain grocery expenditures due to lack of UPC code (some meat, diary, fresh fruit and vegetables). – Omission of expenditures due to household self-scanning. • Explore underreporting by different age/education/year cells (forming a ratio by comparing homescan data to PSID). The gap does not vary with age – however, it does vary with education levels (only 42% of expenditures for high educated vs 55% for low educated). • Underreporting not a problem for our analysis if random. 92 Potential Measurement Issue 2: Attrition • Cannot observe on the extensive margin (homescan only releases data for households who participated consistently over the sample) • Can observe attrition on intensive margin – Compare average expenditures in Homescan between 1993, 1994, and 1995 – first quarter of 1994 had 1% less expenditures than first quarter of 1993 – first quarter of 1995 had 5% less expenditures than first quarter of 1993 • No difference in expenditure declines by age or education • For completeness, we redid our whole analysis only including 1993 – no differences found 93 Potential Measurement Issue 3: Store Effects • Price of a good may be associated with better (unmeasured) services – – – – – – 83.6% of purchases made at grocery stores 4.1% at discount stores 3.1% at price clubs 1.7% at convenient stores 1.5% at drug stores remainder from vending machines, liquor stores, gas stations, pet stores, etc. – Of the grocery stores, essentially all came from Albertsons, King Sooper, Safeway or Cubs Food • For robustness, we computed everything with store chain fixed effects (identify off of price differences at a given chain during a given period of time) 94 Aggregation over Prices • We want a summary of the price a household pays – Relate to cost of time • Households buy many goods and basket varies over time – Look at one popular good (milk) – Define an index that answers: For its particular basket of goods, does this household pay more or less than other households? 95 Definition: Price Index Household j, good i, month m, day t • Expenditure for household j • Average price for good i • Average quantity of good i • “Real” basket of goods (at average price) 96 Price Index 97 Notes on Price Index • Controls for quality. Same UPC code. – Low price does not mean low quality – Does not reflect “bulk” purchases (those are a different UPC code) • “Brand Switching” may occur – robust to inclusion of control for brand switching. • Like a traditional price index – hold quantities constant and vary prices. • Unlike a traditional price index – not prices over time, but prices in the same market at the same time. 98 Simple “Hypothesis Tests” • • Households with high value of time will pay higher prices than households with low value of time. We would expect (all else equal – particularly amounts): – Higher income households to pay higher prices than lower income households – Households with larger families/children to pay higher prices than households with smaller families or no children – Middle aged households (with high wages and lots of child commitments) to pay higher prices than both younger and older households. <<Lifecycle prediction>> Predictions consistent with data 99 Price and Income (Table 1) 1.04 1.03 1.02 Price Index 1.01 1.00 0.99 0.98 0.97 0.96 0.95 Less than $30,000 $30,000 to $50,000 $50,000 to $70,000 Greater than $70,000 p-value of difference < 0.01 p-value of difference < 0.01 100 Price and Household Size (Table 1) 1.06 1.04 Price Index 1.02 1.00 0.98 0.96 0.94 1 2 3 4 >4 Hosuehold Size 101 Price and Household Composition (Table 1) 1.06 1.04 Price Index 1.02 1.00 0.98 0.96 0.94 Married with children Unmarried female Unmarried male with children with children Married w/o children Unmarried female Unmarried male w/o children w/o children 102 103 104 105 106 Cost Minimization on Part of Household subject to Q = market expenditures h = home production time s = shopping time N = some measure of size of shopping basket 107 First Order Condition From Cost Minimization Need to estimate shopping function: p(s,N) Use Homescan data to estimate above equation 108 109 110 Figure A1: Wage Rates Over the Life Cycle, Married Males 25.000 Conditional 20.000 Conditional/Fixed Effect Wage 15.000 10.000 Unconditional 5.000 69 66 63 60 57 54 51 48 45 42 39 36 33 30 27 24 0.000 Age Note: PSID data 111 Estimation of Home Production Function • Cost minimization: MRT between time and goods in shopping = MRT between time and goods in home production • Independent of preferences and dynamic considerations. • Caveat = assuming that the shopper is the home producer • Note: We are allowing shopping functions to differ from home production functions 112 min p( s, Q )Q s h { s ,Q ,h } s.t. f ( h, Q ) C •First-order conditions: p Q s p f Q p C Q Q f C h f p Q h s f p Q p Q Q 113 • Home Production Function – Functional Form: f (h, Q) hh QQ • MRT condition: p p h ln h / Q ln ln s Q Q Q p • σ= 1/(1-ρ) = elasticity of substitution between time and goods 114 in home production • RHS variable can be constructed from shopping data. • No measure of h in scanner data set – Merge in from ATUS using cells based on – 92 separate cells represented in data • Run “between effects” regression over cells 115 116 117 • We estimate an elasticity of substitution between time and goods in home production between 1.5 and 2.1. – Less aggregation leads to lower estimates • With estimated home production parameters, can estimate actual consumption given observed inputs. – Consumption/Expenditure varies over lifecycle – Even if consumption and leisure are separable in utility, need to be careful in interpreting lifecycle expenditure. 118 119 Conclusions (Need To Update) • Fairly large elasticities between time and money due to shopping and home production. • We find that households can and do alter the relationship between expenditures and consumption by varying time inputs. • Household time use, prices, and expenditures vary in a way that is consistent with standard economic principles and the lifecycle profile of the relative price of time. • Supports growing emphasis on importance of non-market sector in understanding household’s interaction in market – Expenditures: “Hump” in household expenditures – particularly decline after middle age – consistent with forward-looking, patient agents. – Labor supply: Stable market hours over last 40 years masks dramatic changes in time spent in home production and leisure. 120