How Do Hours Worked Vary with Income? Cross-Country Evidence and Implications Alexander Bick, Arizona State University Nicola Fuchs-Sch¨ undeln, Goethe University Frankfurt, CFS and CEPR David Lagakos, UC San Diego and NBER
EEA Annual Congress August 24, 2017
Motivation • How do average hours worked vary with aggregate income?
⇒ We do not know: lack of data for low-income countries
1
Motivation • How do average hours worked vary with aggregate income?
⇒ We do not know: lack of data for low-income countries
• Why should we care?
⇒ Important for cross-country productivity differences
⇒ Important for cross-country welfare differences
1
Motivation • How do average hours worked vary with aggregate income?
⇒ We do not know: lack of data for low-income countries
• Why should we care?
⇒ Important for cross-country productivity differences
⇒ Important for cross-country welfare differences
• How do individual hours worked vary with individual income?
1
Measuring Hours Worked
2
A New Data Set of Internationally Comparable Hours Worked • Household surveys for 80 countries from 2005 or closest avail. year
,→ Nationally representative and have 5,000+ individuals aged 15+
* European Labor Force Surveys (26)
* IPUMS (7)
* Individual surveys (47)
⇒ Large efforts to harmonize surveys
3
Why all that Effort? Low Quality of Available Data! • Maddison Data (25 countries) - 1870, 1913, 1950, 1973, 1990 and younger - Yet, lot of arbitrary assumptions: * up to 1913: hours per worker taken from UK for all countries * 1950: mostly inter- or extrapolated (e.g. hours for Australia taken from US, Peru is average of 6 other Latin American countries, etc.)
4
Why all that Effort? Low Quality of Available Data! • Maddison Data (25 countries) - 1870, 1913, 1950, 1973, 1990 and younger - Yet, lot of arbitrary assumptions: * up to 1913: hours per worker taken from UK for all countries * 1950: mostly inter- or extrapolated (e.g. hours for Australia taken from US, Peru is average of 6 other Latin American countries, etc.)
• Total Economy Database/PWT (67 countries) - Annual time-series back to 1950 - Most data points before 1990 are inter- or extrapolated - Asia: Data source for hours per worker mostly missing
4
Why all that Effort? Low Quality of Available Data! • Maddison Data (25 countries) - 1870, 1913, 1950, 1973, 1990 and younger - Yet, lot of arbitrary assumptions: * up to 1913: hours per worker taken from UK for all countries * 1950: mostly inter- or extrapolated (e.g. hours for Australia taken from US, Peru is average of 6 other Latin American countries, etc.)
• Total Economy Database/PWT (67 countries) - Annual time-series back to 1950 - Most data points before 1990 are inter- or extrapolated - Asia: Data source for hours per worker mostly missing
⇒ None of these datasets has any individual level information 4
49 “Core Countries” with Most Comparable Data
1
Hours Information (a) Producing output counted in NIPA
⇒ includes self-employment and unpaid family work
(b) Actual (not usual) hours worked at all jobs (not just primary job)
(c) In the last/recent reference week
5
49 “Core Countries” with Most Comparable Data
1
Hours Information (a) Producing output counted in NIPA
⇒ includes self-employment and unpaid family work
(b) Actual (not usual) hours worked at all jobs (not just primary job)
(c) In the last/recent reference week
2
Survey covers a full year
5
Sample Countries
Country Income Group Core Non-Core N/A
Sample countries cover close to 40% of world population.
6
How Do Hours Worked Vary with Aggregate Income?
7
Key Fact: Avg. Weekly Hours per Adult Decrease with GDPpc 50 45 40
KHM
Hours per Week
35
TZA VNM KEN LAO
30 25
MWI
TLS RWAUGA
ECU
GHA
MNG PAK
20 15
IRQ
PER COL
USA BWA CZE CYP LVA CHE EST PRT TUR MUS ALB SVN ROM SVK GRC DNK NAM ZAF LTU AUT GBR POL HUN SWE DEU IRL ESPFIN FRANLD BGR ITA BEL
10 5 0 7
7.5
8
8.5 9 9.5 ln(GDP per Capita)
8
10
10.5
11
Key Fact is Broad-Based
• Hours per adult are higher in low-income countries - Full sample of countries - By demographic group (gender, age and education) - In manufacturing and services, but not for agriculture
9
Key Fact is Broad-Based
• Hours per adult are higher in low-income countries - Full sample of countries - By demographic group (gender, age and education) - In manufacturing and services, but not for agriculture
• Hours in home prod. of services also higher in low-inc. countries - Lower data quality for non-market hours - Bridgman et al (’17): 136 time-use surveys since 1960 (43 countries)
9
Cross-Country Evidence in Line with U.S. Time-Series 50 45
Hours per Week
40
KHM
35
TZA KEN
VNM
PER
LAO
30
1900 GHA
25
ECU
MNG
1944 COL
TLS MWI
RWA
PAK
UGA
20
USA
BWA
1933 NAM
ALB ROM ZAF
LVA EST MUS TUR LTU SVK HUN POL
BGR
15
CYP CZE PRT SVN GRC
2005 CHE
DNK AUT GBR FIN SWE IRL ESP DEU NLD FRA ITA BEL
IRQ
10 5 0 7
7.5
8
8.5 9 9.5 ln(GDP per Capita)
Core Countries
10
10.5
U.S. Time-Series (1900-2005)
Source U.S. time-series: Ramey and Francis (2009)
10
11
Extensive and Intensive Margin Country Income Group Hours Per Adult Employment Rate Hours Per Worker
Low 28.5 75.3 38.4
High 19.0 54.5 35.0
• Hours per adult are decreasing in GDP per Capita Low- vs. high-inc. countries: 29 hrs/week vs. 19 hrs/week Decrease of empl. rate accounts for 3/4 of decrease of hours per adult
11
Extensive and Intensive Margin
Hours Per Adult Employment Rate Hours Per Worker
Low 28.5 75.3 38.4
Country Income Group Middle High 21.7 19.0 52.7 54.5 41.1 35.0
• Hours per adult are decreasing in GDP per Capita Low- vs. high-inc. countries: 29 hrs/week vs. 19 hrs/week Decrease of empl. rate accounts for 3/4 of decrease of hours per adult • Differences between both margins Employment rate: decreases only b/w low- & middle-inc. countries Hours per worker: hump-shaped pattern (driven by agriculture) 11
Aggregate Implications of Higher Hours in Poor Countries 1
Development Accounting - Accounting for differences in GDP per worker across countries * Caselli (2005), Klenow and Rodriguez Clare (1997), Hall and Jones (1999), Hsieh and Klenow (2010)
⇒ Accounting for hours per worker: 15% larger productivity differences
12
Aggregate Implications of Higher Hours in Poor Countries 1
Development Accounting - Accounting for differences in GDP per worker across countries * Caselli (2005), Klenow and Rodriguez Clare (1997), Hall and Jones (1999), Hsieh and Klenow (2010)
⇒ Accounting for hours per worker: 15% larger productivity differences
2
Welfare Differences - CEV of flow utility over cons. & leisure
(Jones and Klenow, 2016)
- Neo-classical growth model w/ subsistence cons.
(Ohanian et al., 2008)
⇒ Accounting for hours per worker: welfare diffs. 60% larger 12
How Do Hours Worked Vary with Individual Income?
13
Measuring Individual Income: A Challenge for Poor Countries
• Two monthly earning variables: 1
Earnings from paid employment * Highly selected sample in low-income countries: ⇒ Avg. share of wage workers: 23.4% ⇒ Avg. hours of wage workers: 48.2 (vs. 38.4)
14
Measuring Individual Income: A Challenge for Poor Countries
• Two monthly earning variables: 1
Earnings from paid employment * Highly selected sample in low-income countries: ⇒ Avg. share of wage workers: 23.4% ⇒ Avg. hours of wage workers: 48.2 (vs. 38.4)
2
Total earnings (incl. income from self-employment) → Paper
14
Measuring Individual Income: A Challenge for Poor Countries
• Two monthly earning variables: 1
Earnings from paid employment * Highly selected sample in low-income countries: ⇒ Avg. share of wage workers: 23.4% ⇒ Avg. hours of wage workers: 48.2 (vs. 38.4)
2
Total earnings (incl. income from self-employment) → Paper
• Hourly wage =
monthly earnings actual hours in last week∗4.3
14
90
Hours per Worker by Hourly Wage Decile for Wage Earners
70
80
TZA
Hours per Week 30 40 50 60
KEN UGA GHA TLS RWA
KHM PAK VNM
0
10
20
MWI
-6
-5
-4
-3
-2 -1 0 ln(Hourly Wage) Low-Income
15
1
2
3
4
90
Hours per Worker by Hourly Wage Decile for Wage Earners
70
80
TZA
Hours per Week 30 40 50 60
KEN UGA GHA
BWA ECU MNG COL KHM PAK IRQ VNM ROM ALB PERMUS POL LTU LVA BGR
TLS RWA
0
10
20
MWI
-6
-5
-4
-3
-2 -1 0 ln(Hourly Wage)
Low-Income
15
1
2
Middle-Income
3
4
90
Hours per Worker by Hourly Wage Decile for Wage Earners
70
80
TZA
Hours per Week 30 40 50 60
KEN UGA GHA
BWA ECU MNG COL KHM PAK IRQ VNM ROM ALB PER
TLS RWA
LVA
EST CZE HUN GBR LTU MUS TUR SVN DEU POLPRT AUT CYP FIN FRA ESP DNKIRL BEL GRC SWE ITA NLD
BGR
0
10
20
MWI
-6
-5
-4
-3
Low-Income
-2 -1 0 ln(Hourly Wage) Middle-Income
15
1
2
3
4
High-Income
Is Aggregate or Individual Income More Important?
• Aggregate Income: also captures institutional features
• Individual wage: income vs. substitution effect
16
Is Aggregate or Individual Income More Important?
• Aggregate Income: also captures institutional features
• Individual wage: income vs. substitution effect
⇒ Run following regression on pooled data from all countries:
log (hic ) = α + βlog (wic ) + γlog (GDPphc ) + δ1 ageic + δ2 ageic 2 + ic
16
Is Aggregate or Individual Income More Important?
• Aggregate Income: also captures institutional features
• Individual wage: income vs. substitution effect
⇒ Run following regression on pooled data from all countries:
log (hic ) = α + βlog (wic ) + γlog (GDPphc ) + δ1 ageic + δ2 ageic 2 + ic
• Here focus on men – results for women qualitatively similar
16
Elasticities of Hours to Agg. and Ind. Income for Men ln Hours
ln (Hourly Wage)
−0.096 (0.035) –
Country Fixed Effects R2 Obs.
No 0.075 405,431
ln (GDP per Hour)
ln Hours –
ln Hours
ln Hours
−0.094 (0.032)
0.018 (0.060) −0.105 (0.052)
– −0.116 (0.039)
No 0.118 405,431
No 0.119 405,431
Yes 0.245 405,431
• Both country and individual income negatively related to hours • Coefficient on country income insignificant in joint regression
⇒ Individual income more important than country income
17
.1
.2
Country-Specific Hours-Wage Elasticities Increase w/ GDPpc
BGR ALB MWI
RWA
ROM
KHM
-.1
βw
0
USA
TZA
PAK
FIN ITA ESP
CYP GRC
MNG ECU IRQ
-.3
-.2
KEN GHA UGA
SVK PRT
POL PER BWA COL MUS
VNM TLS
GBR BEL FRA LVA IRL CHE AUT EST CZE DEU DNK SWE LTU SVN HUN NLD
TUR
7
7.5
8
8.5 9 9.5 ln(GDP per Capita)
10
10.5
• Elasticity negative for most countries, positive for richest 18
11
.1
.2
Cross-Country Evidence in Line with U.S. Time-Series
USA
GBR BEL FRA IRLCHE AUT DEU DNK SWE NLD
LVA BGR EST CZE LTU HUN SVN ALB MWI
RWA
ROM
KHM
-.1
βw
0
USA-1991
USA-1895
VNM TLS
TZA
PAK
FIN SVK PRTUSA-1973 ITA ESP
POL
CYP GRC
MUS
MNG ECU IRQ
-.3
-.2
KEN GHA UGA
PER BWA COL
TUR
7
7.5
8
8.5 9 9.5 ln(GDP per Capita)
10
10.5
11
• U.S. estimates from Costa (2000) for 1890s, 1973, and 1991 19
Variation of Employment Rates by Education
• For non-working, education as proxy for permanent income
• How do employment rates vary by education?
⇒ Show evidence for prime-aged men, same patterns for older individuals and for women
20
0
10
20
Employment Rate (in %) 30 40 50 60 70 80
90 100
Education Gradient of Employment Rate
Low-Inc. Countries
Middle-Inc. Countries
< Secondary
High-Inc. Countries
Sec. Completed
> Sec.
Sample: Men Aged 25-54
• Low-income countries:
empl. rates flat
21
by education
0
10
20
Employment Rate (in %) 30 40 50 60 70 80
90 100
Education Gradient of Employment Rate
Low-Inc. Countries
Middle-Inc. Countries
< Secondary
High-Inc. Countries
Sec. Completed
> Sec.
Sample: Men Aged 25-54
• Low-income countries:
empl. rates flat
by education
• Middle- & high-inc. countries: empl. rates increase in education
21
0
10
20
Employment Rate (in %) 30 40 50 60 70 80
90 100
Education Gradient of Employment Rate
Low-Inc. Countries
Middle-Inc. Countries
< Secondary
High-Inc. Countries
Sec. Completed
> Sec.
Sample: Men Aged 25-54
• Low-income countries:
empl. rates flat
by education
• Middle- & high-inc. countries: empl. rates increase in education
⇒ As for hours-wage elasticity, slope turns more positive 21
(Aguiar and Hurst, 2007)
Conclusion
22
How Do Hours Worked Vary with Income? 1
New data set of internationally comparable hours worked measures
2
Aggregate level evidence: - 29 hrs/week in low-income vs. 19 hrs/week in high-income countries * 15% larger productivity differences (development accounting) * 60% larger welfare differences (growth model with subsistence cons.)
- Key fact is broad based 3
Individual level evidence: - Individual wage more important than country income for decrease - Within country patterns differ b/w low- and high-income countries 23
Looking Ahead: A Basic Labor-Leisure Model
max {ln(c − c¯) + αln(1 − h) − qIh>0 } h
s.t. c =wh(1 − τ ) + T
• Without c¯ and T , inc. and substitution effect cancel out
24
Looking Ahead: A Basic Labor-Leisure Model
max {ln(c − c¯) + αln(1 − h) − qIh>0 } h
s.t. c =wh(1 − τ ) + T
• Without c¯ and T , inc. and substitution effect cancel out
• Low-inc. countries: w are low, T = 0 → income effect dominates
24
Looking Ahead: A Basic Labor-Leisure Model
max {ln(c − c¯) + αln(1 − h) − qIh>0 } h
s.t. c =wh(1 − τ ) + T
• Without c¯ and T , inc. and substitution effect cancel out
• Low-inc. countries: w are low, T = 0 → income effect dominates
• High-inc. countries: w are high, T > 0 → substit. effect dominates
24
Social Benefits Increase with GDPpc 30
DEU
Social Benefits as % of GDP
25
FRA BEL AUT
20 SRB
15
POLSVK HUN TUR
MNG
10
BGR BRA ROM
TUN ARM COL EGY PRYZAF PER SLV ALB IDN
5
BOL
0 7
7.5
8
8.5 9 9.5 ln(GDP per Capita)
ITA NLD DNK SWE SVN FIN IRL PRT GRC CZE GBR ESP
RUS LTU EST LVA
AUS CAN
USA CHE
MUS CHL
10
Source: Government Finance Statistics, IMF
25
CYP
10.5
11
A Quantitative Exploration • Quantitative importance of subsistence cons. and the welfare state - across countries on the aggregate level - within countries on the individual level
26
A Quantitative Exploration • Quantitative importance of subsistence cons. and the welfare state - across countries on the aggregate level - within countries on the individual level
• Qualitative challenge - Two margins - Agricultural sector
26
A Quantitative Exploration • Quantitative importance of subsistence cons. and the welfare state - across countries on the aggregate level - within countries on the individual level
• Qualitative challenge - Two margins - Agricultural sector
• Complementary mechanism: leisure goods
26
(e.g., Vandenbroucke ’09, Aguiar et al. ’17)