Transcript Slide 1

Micro Data For Macro Models
Topic 2:
Lifecycle Consumption
Part A:
Overview of Lifecycle Expenditures
Why Do We Care About Lifecycle Expenditure?
•
Why is it important?
-
Learn about household preferences broadly
C.E.S. vs. log vs. other / Habits? / Status?
-
Estimate preference parameters
intertemporal elasticity of substitution/ risk aversion/ discount rate
-
Learn about income process
permanent vs. transitory shocks / expected vs. unexpected
-
Learn about financial markets/constraints
liquidity constraints / risk sharing arrangements
-
Learn about policy responses
spending after tax rebates, fiscal multipliers, etc.
Why Do We Care (continued)?
•
The big picture with consumption:
-
Use estimated parameters to calibrate models
-
Understand business cycle volatility
-
Conduct policy experiments (social security reform, health care
reform, tax reform, etc.)
-
Estimate responsiveness to fiscal or monetary policy
-
Broadly understand household behavior
How We Will Proceed
•
The outline of the next part of the lecture:
-
Understand lifecycle consumption movements
o
Illustrative of how one fact can spawn multiple theories.
o
Show how a little more data can refine the theories
o
Illustrate the empirical importance of the Beckerian consumption
model (i.e, incorporating home production and leisure).
Fact 1: Lifecycle Expenditures
0.30
0.25
Log Difference From Age 25
0.20
0.15
0.10
0.05
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.05
-0.10
Plot: Adjusted for cohort and family size fixed effects
Define Non-Durable Consumption (70% of outlays)
• Use a measure of non-durable consumption + housing services
• Non-durable consumption includes:
Food (food away + food at home)
Alcohol and Tobacco
Non-Durable Transportation
Clothing and Personal Care
Domestic Services
Entertainment Services
Utilities
Charitable Giving
Net Gambling Receipts
Airfare
• Housing services are computed as:
Actual Rent (for renters)
Imputed Rent (for home owners) – Impute rent two ways
• Exclude: Education (2%) , Health (6%), Non Housing Durables (16%),
and Other (5%) <<where % is out of total household expenditures>>
Empirical Strategy: Lifecycle Profile of Expenditure
• Estimate:
ln(Citk )  0  age Ageit  cCohortit  t Dt   fs Familyit  itk
k
where Cit is real expenditure on category k by household i in year t.
Note:
All expenditures deflated by corresponding product-level
NIPA deflators.
Cohortit = year-of-birth (5 year range – i.e., 1926-1930)
Dt = Vector of normalized year dummies (See Hall (1968))
Family Composition Controls:
Household size dummies, Number of Children Dummies
Marital status dummies , Detailed Age of Children Dummies
(1)
Fact 2: Hump Shaped Profile – By Education
From Attanasio and Weber (2009)
Fact 3: Retirement Consumption Dynamics
From Bernheim, Skinner and Weinberg (AER 2001)
The Puzzle? (Friedman, Modigliani, Hall, etc.)
max
Ct
u(Ct , t )  Et
T

s t 1
1
1 
( s t )
u(Cs , s )
X t 1  (1  rt 1 )( X t  Ct )  Yt 1
Yt  PV
t t
Pt  g Pt 1 Nt
{Nt, Vt} are permanent and transitory mean zero shocks to income with underlying
variances equal to σ2N and σ2V
Preferences
Ct1 
u (Ct , t ) 
exp(Θt ),   1
1 
(1/  )  intertemporal elasticity of consumption
r  real interest rate
  time discount rate
  vector of taste shifters
Euler Equation
 ln(1   ) ln(1  rt 1 ) (t 1  t ) *
C t 1 


  t 1





where C t 1  ln Ct 1  ln Ct
if r   (in all periods) or if they are constant and
if the forecast error of future consumption (embedded in  * ) is constant
then consumption growth only depends on changes in tastes ()
or changes in the real interest rate.
What Are Potential Taste Shifters Over Life Cycle
1.
Family Size
o
o
o
2.
Makes some difference
Hump shaped pattern still persists
See Facts 1 and 3 (above) – these were estimated taking out detailed
family size controls.
Other Taste Shifters (that change over the lifecycle – for a given
individual)?
Fact 4: Deaton and Paxson (1994)
“Intertemporal Choice and Inequality” (JPE)
Hypotheses:
PIH implies that for any cohort of people born at the same
time, inequality in both consumption and income should
grow with age.
How much consumption inequality grows informs
researchers about:
o
o
Data:
Lifecycle shocks to permanent income
Insurance mechanisms available to households.
U.S., Great Britain, and Taiwan
15
Deaton and Paxson Methodology (U.S. Application)
•
Variance of Residual Variation
k
k
ln Citk  0  age
Ageit  cohort
Cohortit  tk Dt   fsk Familyit  itk
•
Compute variance of εkit at each age and cohort
•
Regress variance of εkit on age and cohort dummies
•
Plot age coefficients (deviation from 25 year olds)
Note: This is my application of the Deaton/Paxson Methodology
(very similar in spirit to theirs).
16
Fact 4: Deaton-Paxson Cross Sectional Dispersion:
With and With Out Housing Services
0.24
Log Deviation From Age 25
0.20
0.16
0.12
0.08
0.04
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
Fact 4: Deaton-Paxson Cross Sectional Dispersion:
With and With Out Housing Services
0.24
Cross Sectional Variance of Total Nondurables for 25 Year Olds = 0.16
0.20
Log Deviation From Age 25
0.16
0.12
0.08
0.04
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.04
Fact 4: Deaton-Paxson Cross Sectional Dispersion:
With and With Out Housing Services
0.24
Cross Sectional Variance of Total Nondurables for 25 Year Olds = 0.16
0.20
Log Deviation From Age 25
0.16
0.12
0.08
0.04
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.04
Questions:
1. What Else Drives the Hump Shaped
Expenditure Profile?
2. Why Does Expenditures (on food)
Fall Sharply At Retirement?
3. Why Does Cross Sectional Consumption
Inequality Increase Over the Lifecycle?
Explanations for Questions (1) and/or (2)
•
Liquidity Constraints and Impatience - Gourinchas and Parker (2002)
•
Myopia - Keynes (and others)
•
Time Inconsistent Preferences (with liquidity constraints) - Angeletos et
al (2001)
•
Habits and Impatience
•
Non-Separable Preferences Between Consumption and Leisure Heckman (1974)
•
Home Production/Work Related Expenses - Aguiar and Hurst (2005,
2008)
Part B:
Gourinchas and Parker (2002)
Gourinchas and Parker (2002)
“Consumption Over the Lifecycle” (Econometrica)
You should read this paper.
Estimates lifecycle consumption profiles in the presence of realistic labor
income uncertainty (via calibration).
Use CEX data on consumption (synthetic cohorts).
Estimates the riskiness of income profiles (from the Panel Study of Income
Dynamics) and feeds those into the model.
Use the model and the observed pattern of lifecycle profiles of expenditure to
estimate preference parameters (risk aversion and the discount rate).
Gourinchas and Parker Structure
 N

t
N 1
max E    u (Ct , )   VN 1 (WN 1 ) 
 t 1

Wt 1  (1  r )(Wt  Yt  Ct )
C 1 
u (C , Z )  v()
1 
Yt  PV
t t
Pt  Gt Pt 1 N t
Impose some liquidity constraints on model: Wt > some exogenous level
Goal of Gourinchas-Parker: Estimate Utility Parameters
•
Intertemporal elasticity of substitution (I.E.S.) (1/ρ)
•
Risk Aversion (ρ)
•
Time Discount Factor (β = 1/(1+ δ))
Note:
Risk aversion = (1/I.E.S.) with CES preferences
Why is the I.E.S. (1/ρ) important?
•
The intertemporal elasticity of substitution determines how levels of
consumption respond over time to changes in the price of consumption
over time (which is the real interest rate – or more broadly – the real return
on assets).
•
This parameter is important for many macro applications.
•
Economics:
Raising interest rates lowers consumption today (substitution effect)
Raising interest rates raises consumption today (income effect – if net
saver)
Consumption tomorrow unambiguously rises
Graphical Illustration – No Substitution Effect
C
High interest rate
ΔC2 = X
ΔC1 = X
1
Low interest rate
2
period
With only an income effect – consumption growth rate will not respond to interest
rate changes. Estimate of (1/ρ) = 0.
Graphical Illustration – With Substitution Effect
C
High interest rate
ΔC2 > X
Low interest rate
ΔC1 < X
1
2
period
As the substitution effect gets stronger, the growth rate of consumption
increases more as interest rates increase. Estimate of (1/ρ) > 0.
One Way to Estimate I.E.S.
T t
 1 
max Et  

j 0  1   
t j
1 
 (Ct  j )

 1 



 C   

Et  t 1   (1  rt 1 )   1
 Ct 

 ln Ct 1   t 1 
1

ln(1  rt 1 )   t 1
Issues With Estimating I.E.S.
 ln Ct 1   t 1 
1

ln(1  rt 1 )   t 1
•
Use of data source (micro or aggregate)
•
Forecast of future interest rates?
•
Correlation of forecast of interest rate with error term (things that make
interest rates go up could be news about permanent income – which affect
consumption).
•
•
Hall (1988) “Intertemporal Substitution in Consumption” (JPE; 1/ρ = 0)
Attanasio and Weber (1993) “Consumption Growth, the Interest Rate and
Aggregation” (ReStud; 1/ρ = 0.60-0.75).
Vissing-Jorgensen (2002) “Limited Asset Market Participation and the
Elasticity of Intertemporal Substitution” (JPE; 1/ρ = 0.3 (stockholders) and
1/ρ = 0.8 (bondholder).
•
Gournichas-Parker Methodology: Calibration
Choose preference parameters that match the lifecycle profiles of consumption
given the mean and variance of income process.
Use synthetic individuals (based on education and occupation)
Using PSID
•
Computed “G” from the data (mean growth rate of income over the
lifecycle).
•
Estimated the variances from the data.
Using CEX
•
Compute lifecycle profiles of consumption
•
Compute lifecycle profile of wealth/income (at beginning of life)
Intuition
No Uncertainty:
No “Buffer Stock Behavior” (uncertainty coupled with liquidity constraints)
Consumption growth determined by Rβ
(where β = 1/(1+δ))
With Income Uncertainty
Buffer stock behavior takes place (household reduce consumption and increase
saving to insure against future income shocks).
Consumption will track income if households are sufficiently “impatient”
Sufficiently Impatient with Uncertainty:
RβE[(GN)-ρ] < 1
Results
Estimates (Base Specification):
δ
= 4.2% - 4.7%
(higher than chosen r = 3.6%)
ρ
= 0.5 – 1.4
(1/ρ = 0.6 – 2.0)
Interpretation
Early in the lifecycle, households act as “buffer stock households”. As income
growth is “high”, consumption tracks income (do not want to accumulate too
much debt to smooth consumption because of income risk)
In the later part of the lifecycle, consumption falls because households are
sufficiently impatient such that δ > r.
Gourinchas-Parker Conclusions
•
Optimizing model of household behavior with income risk can match the
lifecycle profile of household consumption
•
Liquidity constraints can explain early life patterns.
•
Impatience explains the late lifecycle patterns.
•
Households face significant labor earnings risk (holding assets early in
lifecycle even though they are impatient).
Take Away:
Households are sufficiently impatient
Households face non-trivial income risk (even in middle age).
Part C:
The Beckerian Model
of Consumption
Ghez and Becker (1975);
Aguiar, Hurst and Karabarbounis (2011)
V (a, , t )  max U(Ci ,..., CN )   E V (a ', ', t  1)
t
subject to:
Ci  Fi ( H i , X i ), i  1,..., N
H
i
(assume C.E.S., CRS)
 L 1
i
a '  (1  r )a  (1   ) wL  T   pi X i
i
L  0, a '  a.
Let μ, λ, θ, and κ be the respective multipliers on the time budget constraint, the
money budget constraint, the positive hours constraint and the positive
assets constraint.
Assume U(.) is additively separable across time and across goods.
ψ= is vector of wages, commodity prices (p), taxes and transfers
First Order Conditions
U Fi
Xi :
  pi , i
Ci X i
U Fi
Hi :
    , i
Ci H i
L : w    
a ' :  E tVa ' ( a ', ', t  1)     .
If θ = 0 (L > 0), price of time (in permanent income units) (μ/λ = w)
More generally (given L often = 0), μ/λ = ω
First Order Conditions
Intra-period tradeoff between time and goods:
Fi
H i
Fi  w
 
(if L > 0)
X i pi pi
(1)
Marginal rate of transformation between time and goods in
production of n is equated to the relative price of time.
First Order Conditions
A few assumptions:
o
o
Some algebra
Fi is constant elasticity of substitution
pi’s are constant over time
 Xi 
d ln  
 Hi 
 Fi
d ln 
 H i
X
d ln  i 
 Hi   
i
d ln  
Note:
To get (3), sub (2) into (1)
Fi 
 i

X i 
(2)
(3)
Static First Order Condition
The static F.O.C. pins down expenditure relative to time inputs.
If we know σ and the change in the opportunity cost of time, we
should be able to pin down the relative movement in expenditures
relative to time.
%ΔXi -%ΔHi =σi %Δω
Notice, this equation does not require us to make any
assumptions about borrowing or lending, perfect foresight,
etc.
More Intuition (Assume separability in cn’s)
Differentiate FOC for xn with respect to ω holding λ constant. Get:
d ln X i
d ln 
s
H
i
d  0
 i   i 
U Ci
; i  
Ci ( 2U Ci2 )
Fi
Hi
H i
H
si 
Ci
This is just Ghez and Becker (1975)
Need to compare the intra-elasticity of substitution between time and goods
(σ) to the elasticity of substitution in utility across consumption goods (γ).
Note:
Complicates mapping of expenditures into permanent income in
general and the estimation of Engel curves in particular.
Different Than Standard Predictions
Differentiate FOC for xn with respect to ω holding λ constant. Get:
n
d ln c
d ln 
 i
d  0
Spending should fall the most (with declines in the marginal value of wealth) for
goods that have high elasticities of substitution (high income elasticities).
Implications
• For given resources (λ):
– As the price of time increases, consumers substitute market goods for
time (Xi increases) – depends on σi
– As the price of time increases, consumers substitute to goods (periods)
in which consumption is “cheaper” (Xi falls) – depends on γi
• What goods have high/low σ:
-
High σ: goods for which home production is an available margin of
substitution (e.g., food)
Low σ: goods for which time and spending are complements (e.g.,
entertainment goods)
• What goods have high/low γ:
-
High γ: goods which have a high income elasticity (luxuries)
Low γ: goods which have a low income elasticity (necessities)
Predictions: Lifecycle Movements
Gourinchas and Parker model (and most other models)
o
o
Luxuries (entertainment) should decline more late in life relative to
necessities (food)
No importance of changing opportunity cost of time over lifecycle
Beckerian Model
o
o
Goods for which home production is important can move over the
lifecycle in ways that are different than goods for which expenditure
and time are complements.
If opportunity cost of time declines after middle age, food may
decline more than entertainment later in life.
Part D:
Tests for Beckerian Model of
Consumption
Test 1: Aguiar and Hurst
“Consumption vs. Expenditure” (JPE 2005)
Question
• What causes the decline in spending for households at the time of
retirement?
• Bernheim, Skinner, and Weinberg (AER 2001) “What Accounts for the
Variation in Retirement Wealth Among U.S. Households”
o
People do not plan for retirement (myopic)
• Banks, Blundell, and Tanner (AER 1998) “Is There a Retirement Savings
Puzzle”
o
People get bad news (on average) at retirement (shock to λ)
• Hundreds of other papers documenting similar patterns for different
countries.
• Do not think about the cost of time changing with retirement.
47
Fact 3: Retirement Consumption Dynamics
From Bernheim, Skinner and Weinberg (AER 2001)
Our Approach: Measuring Consumption Directly
•
Main Data Set: Continuing Survey of Food Intake of Individuals (CSFII)
–
–
–
–
–
–
–
Conducted by Department of Agriculture
Cross Sectional / Household Level Survey
Two recent waves: Wave 1 (1989 -1991) ; Wave 2 (1994-1996)
Nationally Representative
Multi Day Interview
All individuals within the household are interviewed (C at individual level)
Tracks final food intake (not intermediate goods --- think about a cake)
•
Detailed food expenditure, demographic, earnings, employment, and health
measures
•
Large sample sizes:
– 6,700 households in CSFII-91
– 8,100 households in CSFII-96
• Focus on intake NOT expenditure!
49
Actual Consumption Data (CSFII)
• The key to the data:
24 hour food intake diaries (asked for all days in the survey)
• Diaries are detailed:
–
–
–
–
Amount of food item consumed (detailed 8 digit food codes)
Brand of food item (often unusable by researchers)
Cooking method
Condiments added
• Dept of Agriculture converts the total day’s food intake into several
nutritional measures (calories, protein, saturated fat, total fat, vitamin C,
riboflavin, etc.).
– The conversion is made using all food diary data (i.e., brand, whether
cooked with butter).
50
8 digit food codes: Cheese
• Example 18 of the 100 8-digit codes for cheese.
14101010
14102010
14102110
14103020
14104010
14104020
14104200
14104250
14105010
14105200
14106010
14106200
14106500
14107010
14107020
14107030
14107040
14107060
CHEESE, BLUE OR ROQUEFORT
CHEESE, BRICK
CHEESE, BRICK, W/ SALAMI
CHEESE, BRIE
CHEESE, NATURAL, CHEDDAR OR AMERICAN TYPE
CHEESE, CHEDDAR OR AMERICAN TYPE, DRY, GRATED
CHEESE, COLBY
CHEESE, COLBY JACK
CHEESE, GOUDA OR EDAM
CHEESE, GRUYERE
CHEESE, LIMBURGER
CHEESE, MONTEREY
CHEESE, MONTEREY, LOWFAT
CHEESE, MOZZARELLA, NFS (INCLUDE PIZZA CHEESE)
CHEESE, MOZZARELLA, WHOLE MILK
CHEESE, MOZZARELLA, PART SKIM (INCL ""LOWFAT"")
CHEESE, MOZZARELLA, LOW SODIUM
CHEESE, MOZZARELLA, NONFAT OR FAT FREE
51
Changes in “Spending” At Retirement
Run: ln(xi) = γ0 + γ1 Retiredi + γ2 Zi + errori
• Retiredi is a dummy variable equal to 1 if the household head is retired.
• Instrument Retiredi status with age dummies (potential endogeneity)
• Z includes: race, sex, health, region, time, family structure controls
• Sample: Relatively “young” older households: Heads aged 57-71
• Total food expenditure (x) falls by 17% for retired households (γ1), p-value
< 0.01
• Other results:
– Food expenditure at home falls by 15%
– Food expenditure away from home falls by 31%
52
Changes in “Consumption” at Retirement
•
How do we turn these food diaries into meaningful measures of consumption?
•
Our approach:
1. Examine Nutritional Quality of Diet (vitamins, cholesterol, fat, calories, etc.)
2.
Examine individual goods with strong income elasticities (hotdogs, fruit,
yogurt, shellfish, wine)
3.
Luxury/Quality goods (e.g. brands vs generics, lean vs. fatty meat)
4.
Use structural model to aggregate food consumption data and perform
formal PIH test.
53
Nutritional Measures
•
•
Regress: ln(ci) = α0 + α1 ln(yperm) + demographics <<sample: heads 25-55>>
Regress: ln(ci) = β0 + β1 Retired + demographics <<sample: heads 57-71>>
Consumption Measure (in logs)
Calories
Protein *
Estimated Elasticity (α1)
-4%
(2%)
-1%
(1%)
Vitamin A *
Vitamin C *
Vitamin E *
Calcium *
44%
34%
18%
10%
(5%)
(5%)
(3%)
(2%)
36% (9%)
33% (9%)
11% (4%)
13% (4%)
- 26%
- 9%
(3%)
(2%)
-9% (5%)
-7% (3%)
Cholesterol *
Saturated Fat *
•
•
•
Retirement Effect (β1)
-2% (4%)
-3% (2%)
* Includes log calories as an additional control ; Include supplements as an additional control.
Instrument for retirement status with age; Examined non-linear specifications (not reported)
No evidence of any deterioration in diet quality
54
Some Specific Consumption Measures
•
•
Regress: ci = α0 + α1 ln(yperm) + demographics <<sample: heads 25-55>>
Regress: ci = β0 + β1 Retired + demographics <<sample: heads 57-71>>
Consumption Measure (Dummy)
Eat Fruit
Eat Yogurt
Eat Shellfish
Drink Wine
Eat Oat/Rye/Multigrain Bread
Estimated Semi-Elasticity
0.25 (0.03) <<59%>
0.14 (0.02) <<8%>>
0.05 (0.01) <<6%>>
0.15 (0.02) <<8%>>
0.10 (0.02) <<9%>>
Retirement Effect
0.14 (0.04)
0.01 (0.03)
-0.02 (0.02)
-0.03 (0.03)
0.06 (0.04)
Eat Hotdog/Sausage
Eat Ground beef
-0.16 (0.03) <<51%>>
-0.10 (0.03) <<22%>>
-0.06 (0.05)
-0.01 (0.04)
•
•
Sample means in << >>
Instrument for retirement status with age
•
•
Drawback: Tastes could differ across income types
Drawback: Categories are broad and do not allow for differences in quality
55
Luxury Goods/Quality: My Favorite….
• Examine some dimensions of quality:
– Eating at restaurants with Table Service
– Eating Branded vs. Generic Goods
– Eating Lean vs. Fattier Cuts of Meat
• Restaurants, Brands, and Eating Lean Meat have very STRONG income
elasticities in the cross section of working households.
• If households are unprepared for retirement, we should see them switching
away from such consumption goods.
• No evidence of that in the data.
56
Creating a Food Intake Aggregate
ln( y perm,i )  0  1c1,i t  ....   J cJi ,t   X ln( X t )  ti  ti
• Where
c1, ….. cJ are quantities of individual consumption categories consumed
X is monthly expenditure on food
θ is a vector of demographic and health controls (including education, sex,
race, family composition, ect.)
yperm is the household’s predicted permanent income
• Estimated on a sample of 40 – 55 year old household heads where the head is
working full time.
57
Thought Experiment
• Permanent income is our numeraire – one unit increase in our consumption
index maps into a one percent increase in permanent income.
– What are we doing: We project permanent income of household i onto
household i’s consumption (controlling for taste shifters).
• Basically, in a statistical sense, if you tell me what you eat, I can predict your
permanent income. Our consumption index is in permanent income
dollars!
• We also did this for households aged 25-55 who are working fulltime (results
did not change).
• We want to ask if households act like their permanent income has
changed once they become retired.
58
Is Our Permanent Income Measure Predictive?
• Projection of income on consumption and expenditure patterns
• How well does consumption forecast income?
– Split sample into odd and even years (again focusing only on prime age
household heads working full time).
– Focus only on odd years of our sample (in sample):
• In sample R-square 0.53
• Food consumption on its own explain 21% of variation in income
• Incremental R-square is 0.12
– Focus on even years (test out of sample):
• Out of sample R-square: 0.42
• Food consumption and expenditure a fairly good predictor of income
59
60
A Note on the Unemployed
• Unemployed, on average, should experience some decline in expenditure.
• Labor studies find that the unemployed (from exogenous plant closings)
have earnings that are 5-10 percent lower during the subsequent decade.
• Can our methodology detect a decline in expenditures for the unemployed?
• Our study is imperfect – we only have cross sectional data.
• Using the panel dimension of the PSID, the unemployed experience a
reduction in expenditures of about 8 percent (Stephens, 2002). We find a
decline of about 15 percent (in expenditures) using our data.
• In terms of actual consumption intake, we find the unemployed reduce their
intake by about 6 percent.
61
Conclusions
• No “Retirement Consumption Puzzle”
• Technically, preferences between “consumption” and leisure are not
substitutes.
– Leisure goes up dramatically in retirement (we will show this in a few
weeks).
– Food consumption (as measured by intake) remains roughly constant (if
anything it increases slightly).
• However, “expenditures” and leisure could still be non-separable.
– Non-separability enters through “home production”
62
Test 2: Aguiar and Hurst (2009)
“Deconstructing Life Cycle Expenditure”
Question
• What about the lifecycle patterns of consumption more broadly?
o
Can a Beckerian model explain the declining expenditures post
middle age with relying on either:
-
really impatient consumers?
myopia (or time inconsistent preferences)?
• Use the disaggregated consumption data by category?
• Estimate a model on the disaggregated data.
-
estimate time preference rate
estimate the amount of risk households face
Entertainment Spending
1.0
Log Deviation From Age 25
0.8
0.6
0.4
0.2
0.0
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.2
All Non Decreasing Categories
3.00
Log Deviation from Age 25
2.50
2.00
1.50
1.00
0.50
0.00
-0.50
25
30
Entertainment
35
Utilities
40
45
50
Housing Services
55
60
Other ND
65
70
Domestic Svcs
75
Decreasing Categories
0.40
Log Deviation from Age 25
0.20
0.00
-0.20
-0.40
-0.60
-0.80
-1.00
-1.20
25
30
35
40
45
50
55
60
65
70
Age
Clothing
Transportation
Food at Home
Food Away
75
Summary (in Log Differences)
Log Change Log Change Log Change
Between
Between
Between
25 and 44
45 and 59
60 and 68
Consumption Category
Share
Decreasing Categories
Food at Home
Transportation
Clothing/Personal Care
Food Away from Home
Alcohol and Tobacco
0.17
0.13
0.08
0.06
0.03
0.24
0.25
0.04
0.13
-1.35
-0.07
-0.20
-0.36
-0.55
-1.69
-0.04
-0.17
-0.20
-0.29
-1.22
Non-Decreasing Categories
Housing Services
Utilities
Entertainment
Other Non-Durable
Domestic Services
0.33
0.11
0.04
0.03
0.02
0.73
0.72
0.80
1.44
1.52
0.23
0.28
0.07
0.16
0.30
0.14
0.11
0.17
0.17
0.32
What About Deaton-Paxson Fact?
•
Examine lifecycle profile of cross sectional inequality by
category
•
Goods which have expenditures that increase with market
work (due to home production or complementarity) should
experience increasing dispersion when the dispersion of work
increases.
•
Portion of lifecycle profile of cross sectional inequality due to
these goods does NOT inform researchers about:
o
o
Lifecycle profile of shocks to permanent income
Insurance mechanisms available to households
69
Dispersion of Propensity to Work Over Life Cycle
0.60
0.50
0.40
0.30
0.20
0.10
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
Age
0.50
Cross Sectional Dispersion Over Lifecycle
Percentage Point Deviation From Age 25
0.40
0.30
0.20
0.10
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.10
-0.20
Core Nondurable
0.50
Cross Sectional Dispersion Over Lifecycle
Percentage Point Deviation From Age 25
0.40
0.30
0.20
0.10
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.10
-0.20
Core Nondurable
Work Related
Food At Home
Cross Sectional Dispersion Over Lifecycle: Figure 6b
0.50
Percentage Point Deviation From Age 25
0.40
0.30
Total
0.20
0.10
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
Core
-0.10
-0.20
Cross Sectional Dispersion Over Lifecycle: Figure 6b
0.50
Percentage Point Deviation From Age 25
0.40
0.30
Total
0.20
0.10
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
Core
-0.10
-0.20
Food, Transportation and Clothing
• Food is amenable to “Beckerian” home production (see
Aguiar and Hurst 2005, 2007)
No evidence of any decline in food intake over the lifecycle
despite declining food expenditures.
As opportunity cost of time declines later in life, households
substitute towards home production of food (including more
intense shopping for bargains).
Data (and calibrated model) actual show food intake increases
over the back half of the lifecycle
Work Related Expenses
• Transportation, Clothing and Food Away From Home are
work related expenses:
Lazear and Michael (1980) – Net out work related expenses
(clothing and transportation) when making welfare calculations
across people
Banks et al (1998) and Battistin et al (2008) when measuring
consumption changes of retirees
Nelson (1989) and DeWeese and Norton (1991) comprising
models of “clothing demand”
Level of Work Hours Over the Lifecycle
1
45
0.9
40
Fraction Working
Fraction Working
0.7
35
Hours Worked Per Week
0.8
30
0.6
25
0.5
20
0.4
15
0.3
0.2
Hours Per Week
Worked
10
0.1
5
0
0
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
Age
New Facts About Food, Clothing, and Transport
• Look at food away patterns at different types of establishments
• Look at changes in different amounts of transportation patterns
using time use data
• Estimate “simple” demand systems and control directly for
work status
Propensity To Eat Away At Home
0.05
Percentage Point Deviation From Age 25
0.00
-0.05
-0.10
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
-0.15
Any Eating
Establishment
-0.20
-0.25
-0.30
-0.35
Age
Propensity To Eat Away At Home
0.05
Restaurants at Lunch
Percentage Point Deviation From Age 25
-3E-16
Restaurants at Dinner
-0.05
-0.1
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
-0.15
Any Eating
Establishment
-0.2
-0.25
-0.3
-0.35
Age
Propensity To Eat Away At Home
0.05
Restaurants at Lunch
Percentage Point Deviation From Age 25
0
Restaurants at Dinner
-0.05
-0.1
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
-0.15
Any Eating
Establishment
-0.2
-0.25
Fast Food and
Cafeteria
-0.3
-0.35
Age
Travel Times and Employment Status
Hours Per Week Deviation From 25-29 Year Olds
1.00
0.50
0.00
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
-0.50
All Travel Time
-1.00
-1.50
-2.00
-2.50
Travel Times and Employment Status
1.00
Hours Per Week Deviation From 25-29 Year Olds
Non Work Travel Time
0.50
0.00
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
-0.50
-1.00
-1.50
-2.00
-2.50
All Travel Time
Travel Times and Employment Status
1.00
Hours Per Week Deviation From 25-29 Year Olds
Non Work Travel Time
0.50
0.00
25-29
30-34
35-39
40-44
45-49
50-54
55-59
60-64
65-69
70-74
-0.50
-1.00
All Travel Time
-1.50
-2.00
-2.50
Work Travel Time
Control Directly For Work Status
• Estimate a demand system
• Control for labor supply (conditional on total expenditures)
• Estimate:
1)
what consumption categories where spending is positively
associated with market work
2)
to what extent is the decline in spending on clothing,
transportation and food away from home attributable to
employment status.
Estimate Simple Demand System
sitk  0  age Ageit  cCohortit  t Dt   fs Familyit   pk ln Pt k  p ln Pt
k
 X ln X it  L Lit   itk ,
Xit
is total nondurable expenditures (less alcohol and tobacco, plus
housing) for household i in year t.
sitk
is the share of expenditures in consumption category k out of Xit
Ptk
is the price index for consumption category k in year t
Lit
is a vector of work status controls for household i in year t.
Note:
Instrument lnXit with household total income and education controls
1. Simple Demand System Results
• Restrict sample to married households between age 25 and 50
• Use two work status controls: Husband working? Wife working?
Simple Demand System Results
• Restrict sample to married households between age 25 and 50
• Use two work status controls: Husband working? Wife working?
Consumption Category
Husband Work?
Wife Work?
Transportation (0.13)
0.014 (0.002)
0.014 (0.002)
Clothing/P. Care (0.08)
0.003 (0.001)
0.001 (0.001)
Food Away From Home (0.06)
0.008 (0.001)
0.005 (0.001)
Simple Demand System Results
• Restrict sample to married households between age 25 and 50
• Use two work status controls: Husband working? Wife working?
Consumption Category
Transportation
(0.13)
Husband Work?
Wife Work?
0.014 (0.002)
0.014 (0.002)
Clothing/P. Care (0.08)
0.003 (0.001)
0.001 (0.001)
Food Away From Home (0.06)
0.008 (0.001)
0.005 (0.001)
Housing Services (0.34)
-0.009 (0.003)
-0.012 (0.002)
Utilities
(0.12)
-0.005 (0.001)
-0.003 (0.001)
Food At Home (0.18)
-0.016 (0.002)
-0.013 (0.001)
Entertainment (0.04)
0.000 (0.001)
0.000 (0.001)
2. Adding Work Controls To the Lifecycle Profile
• Married Sample, 25 – 75
• Work Status Controls:
7 Dummies for Husband Weeks Worked
7 Dummies for Wife Weeks Worked
9 Dummies for Hours per week Husband Worked
9 Dummies for Hours per week Wife Worked
• Three Categories:
Food (food at home and food away)
Work Related Expenses (transportation and clothing)
Core Non Durables (everything else)
• Ask: “How do work status controls effect lifecycle profiles?”
Share of Expenditure:
Difference from Age 25
Demand Estimates, Transportation
0.010
0.005
0.000
-0.005
-0.010
-0.015
-0.020
-0.025
-0.030
-0.035
-0.040
-0.045
25
30
35
40
45
50
Age
55
60
65
70
75
Demand Estimates, Food Away
Share of Expenditure:
Difference from Age 25
0.005
0.000
-0.005
-0.010
-0.015
-0.020
25
30
35
40
45
50
Age
55
60
65
70
75
Share of Expenditure:
Difference from Age 25
Demand Estimates, Clothing
0.005
0.000
-0.005
-0.010
-0.015
-0.020
-0.025
-0.030
-0.035
-0.040
25
30
35
40
45
50
Age
55
60
65
70
75
Level of Lifecycle Expenditure
1.2
1.0
Log Deviation from Age 25
0.8
0.6
0.4
0.2
0.0
-0.2
-0.4
-0.6
-0.8
25
30
35
Work Related
40
45
50
Core Nondurables
55
60
65
Food at Home
70
75
Level of Lifecycle Expenditure (Older Version)
1.20
1.00
Log Deviation From Age 25
0.80
0.60
0.40
0.20
0.00
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75
-0.20
-0.40
-0.60
-0.80
Core Nondurable
Total Nondurable
What Does it Mean?
• Write down a model where households maximize utility with three
consumption goods and leisure with the following constraints:
one good (food) is amenable to home production
one good (transport, clothes) are complements to market work
there is a time budget constraint
Assumptions:
o
o
o
conditional on work, income process is uncertain
take the lifecycle process of work as exogenous
assume that individual receives no utility for the lifecycle component
of work related expenses.
• Other from the disaggregated consumption data (and home production
functions), very similar procedure to Gourinchas and Parker.
Model: Household
Income Risk While Working:
Retirement/Disability Shock (Rt)
Conditional on Rt = 0, there is an age dependent hazard that next period Rt+1 = 1.
Model: Household
•
Close the model with a standard representative competitive firm.
•
Calibrate the model to match: real interest rate of 4%, aggregate wealth to income
ratio of 3.1, average labor supply of prime age workers (of 1/3 time endowment),
lifecycle profile of spending on “core” and “home-production”/ “work-related” goods
the variance of spending on those goods and the co-variance between the
two goods.
Findings
Findings: Lifecycle Profiles
Findings: Lifecycle Profiles
Findings: Lifecycle Profiles
Findings: Lifecycle Profiles
Home Production vs. Non-Separable Preferences
A Question:
• Does one need to model the home production sector formally?
• There is always a mapping between home production (non-separability
between X and N through home production technology) and preferences
(non-separability between X and N through preferences).
o
o
X = expenditures
N = labor
• However, to match the data, may need to have preference parameters
change over time (or states).
• We will talk more about this in Topic 4.
Heckman (1974):
Non-Separable Consumption and Leisure
max
Ct , Nt
u (Ct , Nt )  Et
T

s t 1
1
1 
( s t )
u (Cs , N s )
1
u (Ct , Nt ) 
(Ct (1  Nt )1 )1
1


C t 1   0  1 ln(1  rt 1 )  2 (1  Nt 1 )  t*1
Big Picture Wrap Up: Non Separabilities
My belief:
U(C,N) can be written as u(C) + v(N)
However – we do not measure C directly:
C = f(x,h) where h is directly related to N (through time budget
constraint).
We measure X and N in the data.
X = f-1(C,h(N))
Implication:
U(X,N) cannot be written as U(X) + V(N).
A Short Summary
Non Separabilities between X and N (expenditure and labor supply) are
important.
When is it important to implicitly model the home production sector?
When changes to home production technology are important!
When care about cross good predictions.
When have actual consumption (intake) measures.
For most applications, a reduced form assumption that X and N are nonseparable can be important.
Show a situation (with labor supply) where it may be useful to separate the
home production sector separately from preferences.