How Do Average Hours Worked Vary with Development? 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
April 18, 2016
How do Average Hours Vary Across World Income Distribution?
• No clear conclusion thus far; limited by lack of cross-country data • This paper: answer using new data set
- Harmonize 81 countries of all income levels - Draw on nationally representative household surveys - Challenge: surveys not already standardized - Large efforts to maximize international comparability
How do Average Hours Vary Across World Income Distribution?
• No clear conclusion thus far; limited by lack of cross-country data • This paper: answer using new data set
- Harmonize 81 countries of all income levels - Draw on nationally representative household surveys - Challenge: surveys not already standardized - Large efforts to maximize international comparability
Main Empirical Findings
• Average hours worked per adult are higher in poor countries
• True for both sexes, by broad age group, education and sector
• Magnitudes substantial: 29 hours/week in poorest countries,
compared to 19 hours per week in richest
• Extensive and intensive margins both at play; extensive more for
poorest vs middle income, intensive more for middle vs rich
Relevant for Development Accounting and Welfare
• Relevant for “development accounting” & theories of development
- Cross-country differences in output/hour larger than output/worker - Evidence against theories of income differences based on work effort / work ethic (e.g. Landes, 1999; Leamer, 1999) • Relevant for “welfare” differences across countries
- Example: Comparing North Americans and Africans - Penn World Tables: c roughly ∼ 5% as high for Africans - We show h ∼ 50 % higher too; welfare differences larger than suggested by c differences alone
Relevant for Development Accounting and Welfare
• Relevant for “development accounting” & theories of development
- Cross-country differences in output/hour larger than output/worker - Evidence against theories of income differences based on work effort / work ethic (e.g. Landes, 1999; Leamer, 1999) • Relevant for “welfare” differences across countries
- Example: Comparing North Americans and Africans - Penn World Tables: c roughly ∼ 5% as high for Africans - We show h ∼ 50 % higher too; welfare differences larger than suggested by c differences alone
This Talk
1
Measuring average hours worked across countries
2
Empirical findings
3
Implications for development accounting
4
Model + implications for welfare
Measuring Hours Worked
Constructing Our Data Set • Use surveys from 2005 or closest available year
• Use only nationally representative surveys of households
the European Union Labor Force Surveys (27), IPUMS (7), other individual surveys (47)
• Challenge: surveys not standardized; required large efforts to
harmonize
• Full data set: 81 countries; focus on 43 “core countries” with most
comparable data
Core Countries We define core countries as those that meet the following criteria 1
Survey covers whole calendar year, 5,000+ individuals
2
Actual hours worked (not usual) at all jobs (not just primary job)
3
In the last week, or recent reference week
4
Producing output counted in NIPA (not e.g. home child care)
Measuring Hours Worked
• Adult ≡ age 16+ • Employed ≡ working in prior week or self reports having regular
employment relationship • Three main measures of hours worked 1
Average hours per adult (unconditional)
2
Employment rate
3
Average hours per worker (conditional)
Empirical Findings
Average Weekly Hours per Adult 50 45 KHM
40 Hours per Week
35
TZA
LAOVNM
PER
30 GHA
25
PAK
UGA RWA
MNG IDN
COL USA
BWA
LVA EST CYP CZE PRT CHE SVN GRC GBR AUT NOR LTUSVK DNK POL HUN SWE FIN DEU ESP NLD BGR FRA ITA BEL
TUR MUS ROM
20 15
IRQ
10 5 0 6.5
7
7.5
8
8.5 9 9.5 ln(GDP per Capita)
10
10.5
11
Income Group Average
Males
Females
Average Hours per Prime (Aged 25-55) 50 45
KHM
40
PER
LAO GHA
35 Hours per Week
VNM
TZA
MNG
UGA
30
COL
IDN PAK
RWA
25 20
LVA USA CZE SVN BWA LTUEST CYP PRT GRC ROM SVK AUT CHE POLHUN MUS DNK GBR FIN BGR DEU TUR SWE FRA NOR ESP ITABEL NLD
IRQ
15 10 5 0 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita) Income Group Average
9.5
10
10.5
11
Average Hours per Old (Aged 55+) 50 45 40 KHM
Hours per Week
35 30
BWA
TZA
25
PER
GHA LAO
20
RWA UGA
15
PAK VNM
IDN COL
MNG
10
IRQ
5
NOR USA CYP EST PRT CHE MUS SWE LVA DNK ROM CZE GBR TUR LTU GRCFIN NLD DEU SVK HUN SVNESP BGR POL FRA AUT ITA BEL
0 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita) Income Group Average
9.5
10
10.5
11
Average Hours per Adult: Differences in Means
Age Group All Young Prime Old
Country Income Group Low-High Low-Middle Middle-High 10.2∗∗∗ 7.3∗∗∗ 2.9∗∗ 9.4∗∗∗ 7.5∗∗∗ 1.9∗ ∗∗∗ ∗∗ 7.7 6.7 1.0 ∗∗∗ ∗∗ 13.3 8.2 5.1∗∗∗
Average Weekly Hours Per Adult: Broader Sets of Countries
Core Countries Core + Partial-Year All Countries
Country Income Group Low-High Low-Middle Middle-High 10.2∗∗∗ 7.3∗∗∗ 2.9∗∗ ∗∗∗ ∗∗ 6.0 4.0 2.0∗∗ 5.8∗∗∗ 3.5∗∗ 2.3∗∗
Comparison to US Time-Series (Ramey & Francis, 2009) 50 45
Hours per Week
40 35 30 25 20 15 10 5 0 6
7
8 9 ln(GDP per Capita) 2005 Core Countries
10
US Time−Series
11
Account for Findings by Disaggregating Further?
• Are aggregate hours differences due to composition differences, by
e.g. education or sector? • We look at hours per adult by: less than secondary, secondary, more
than secondary • Also hours per worker at those employed in agriculture, services,
manufacturing
Account for Findings by Disaggregating Further?
• Are aggregate hours differences due to composition differences, by
e.g. education or sector? • We look at hours per adult by: less than secondary, secondary, more
than secondary • Also hours per worker at those employed in agriculture, services,
manufacturing
Account for Findings by Disaggregating Further?
• Are aggregate hours differences due to composition differences, by
e.g. education or sector? • We look at hours per adult by: less than secondary, secondary, more
than secondary • Also hours per worker at those employed in agriculture, services,
manufacturing
Differences in Average Hours by Education
Average Hours by Age and Country Income Education All All (Ages 25+, Non-missing Educ.) Ages 25+ Less than Secondary Secondary Completed More than Secondary
Country Income Low Middle 29.3 22.0 33.9 25.3 32.6 38.1 39.5
20.9 29.6 30.9
Group High 19.1 21.0 11.9 23.6 26.9
Differences in Average Hours by Education
Education All All (Ages 25+, Non-miss Educ.) Ages 25-55 Less than Secondary Secondary Completed More than Secondary
Country Income Group Low-High Low-Middle Middle-High 10.2∗∗∗ 7.3∗∗∗ 2.9∗∗ 12.9∗∗∗ 8.6∗∗ 4.3∗∗∗ 13.3∗∗∗ 10.5∗∗∗ 10.4∗∗∗
9.2∗∗∗ 7.4∗∗∗ 7.6∗∗
4.1∗∗ 3.1∗ 2.8∗
Differences in Hours per Worker by Sector
All All (Non-missing Sec.) Agriculture Manufacturing Services
Country Income Group Low-High Low-Middle Middle-High -1.4 6.6∗∗∗ 5.2∗∗∗ 5.2∗∗∗ -0.9 6.1∗∗∗ -2.6 0.1 -2.7 8.1∗∗∗ 1.9 6.2∗∗∗ 13.5∗∗∗ 5.1∗∗ 8.4∗∗∗
Hours per Worker in Agriculture
50 45
COL
TZA KHM
40 LAO GHA
Hours per Week
35
PAK
MNG
IRQ ROM
ESP
LTU
PER
30
GBR EST DEU HUN CZESVN SVK BEL GRC FIN ITA SWE CYP DNK POL
USA
LVA
VNM
IDN
AUT FRA
BWA
MUS
NOR NLD
PRT
BGR RWA
25
UGA
20 15 10 5 0 6.5
7
7.5
8
8.5 9 ln(GDP per Capita)
9.5
10
10.5
11
Hours per Worker in Manufacturing
KHM
50 45
GHA
RWA TZA
40
PAK VNM
MNG
LAO
COL PER
IDN IRQ
UGA
BWA ROM MUS BGR
Hours per Week
35 30
GRC LVAPOL USA EST HUN CZE LTU SVK PRTSVN CYP GBR DEUAUT ESP ITA FIN BEL FRA DNK NOR SWE NLD
25 20 15 10 5 0 6.5
7
7.5
8
8.5 9 ln(GDP per Capita)
9.5
10
10.5
11
Hours per Worker in Services
RWA
50
KHM
UGA TZA
45
PAK GHA LAO VNM
BWA
PER ROM
40
IRQ
35 Hours per Week
COL
IDN MNG
MUS BGR
30 25
LVA GRC POLSVK HUN CZE USA EST PRT LTU CYP SVN ITA AUT DEU ESP FRA GBR FIN BEL DNK SWE NOR NLD
20 15 10 5 0 6.5
7
7.5
8
8.5 9 ln(GDP per Capita)
9.5
10
10.5
11
Account for Findings by Extensive-Intensive Margin?
• Higher employment rates in poor countries?
• Greater hours worked per worker in poor countries?
• We’ll look at the aggregates (males and females separately in paper)
Employment Rates 100 90 UGA RWA TZA
Employment Rate (in %)
80
KHM LAO VNM
PER
70 GHA
IDN MNG
60 50
PAK
NOR CHE USA DNK NLD GBR CYP SWE PRT COL AUT SVN FIN CZE MUS BWALVA LTUEST ESP DEU FRA ROM SVK BEL GRC ITA TUR POLHUN BGR
40 IRQ
30 20 10 0 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita) Income Group Average
9.5
10
10.5
11
Average Weekly Hours Per Worker 50 KHM PAK
45
TZA
40 Hours per Week
35 30
COL
MNG GHAVNM IDN IRQ LAO
PER
BWA TUR
ROM MUS BGR
RWA UGA
LVA GRC POL SVK USA EST HUNCZE LTU CYP PRT SVN AUT ITA DEU ESP CHE GBR FIN FRA BEL DNK SWE NOR NLD
25 20 15 10 5 0 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita) Income Group Average
9.5
10
10.5
11
Implications for Development Accounting
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)
• Every study has ignored differences in average hours per worker
Development Accounting
Low GDP per Adult GDP per Worker GDP per Hour
7.7 6.2 5.2
Country Income Group Middle High 30.5 33.3 27.1
100.0 100.0 100.0
High/Low 12.9 16.1 19.2
⇒ Hours don’t help account for income differences; make puzzle worse
⇒ Evidence against theories based on higher labor input in richer countries
Implications for Welfare
Quantifying Welfare Differences • Welfare metrics
(Jones and Klenow, 2015)
,→ Consumption equivalent variation of flow utility over cons. and leisure ⇒ We need to take a stand on preferences • Standard Neo-Classical Growth Model
(Prescott, 2004)
1
Add subsistence consumption
2
Calibrated version predicts cross-country hours per adult reasonably well
(Ohanian, Raffo and Rogerson, 2008)
* Benchmark exercise abstracts from taxes & government consumption 3
Use calibrated preferences for welfare analysis
Standard Neo-Classical Growth Model • Flow utility
u(c, h) = ln(c − c¯) − αv (h) (1) Marginal rate of substitution b/w leisure & cons. = the price ratio αv 0 (h) =w 1/(c − c¯)
(2) Profit-maximization: wage = marginal product of labor w = (1 − θ)Ak θ h−θ = (1 − θ)y /h
(1) & (2): hv 0 (h) =
(1−θ)
( yc − yc¯ )α
Calibration hv 0 (h) =
1
2
3
(1 − θ) (1 − θ) ⇔ α= c¯ c c¯ 0 −y α − y y hv (h)
c y
Given average of yc , and yc¯ in high-income countries, set α to match avg. hours per adult in high-income countries Given α, yc , and yc¯ in each country, predict hours per adult in each country Specification v (h) (a) −ln(1 − h) 1 h1+1/φ , φ ∈ {1, 2, ∞} (b) 1+1/φ
θ
c¯
c, y
0.3224
$1 per day
PWT
Predictions
Data Model Log φ=∞ φ=2 φ=1
Mean Hours
¯ low inc. h ¯ high inc. h
¯ middle inc. h ¯ high inc. h
21.8
1.53
1.15
21.0 21.9 20.7 20.3
1.35 1.53 1.31 1.22
1.09 1.15 1.08 1.06
• Frisch φ = ∞ gives best fit of the slope
• The higher the labor supply elasticity, the better the fit of the slope
Predictions for Frisch φ = ∞ 50
Hours per Week
40
30
20
10
0 6.5
7
7.5
8
8.5 9 9.5 ln(GDP per Capita) Data
10
10.5
11
Model Inputs
Welfare Analysis • Consumption equivalent variation
(Jones & Klenow, 2015)
Details
- Welfare of country i is λi s.t. u(ci , hi ) = u(λi · cˆHI , hˆHI ) - Use empirical ci , hi and calibrated parameters to pin down λi
Welfare Analysis • Consumption equivalent variation
(Jones & Klenow, 2015)
Details
- Welfare of country i is λi s.t. u(ci , hi ) = u(λi · cˆHI , hˆHI ) - Use empirical ci , hi and calibrated parameters to pin down λi
Low
Country Income Group Middle High
High/Low
Consumption
6.4
26.3
100
15.7
+ Hours Per Adult Log φ=∞ φ=2 φ=1
4.5 4.6 3.4 2.6
23.8 23.9 22.2 21.7
100 100 100 100
22.1 21.7 29.7 38.4
Average Hours per Week on Home Production Country Income Group
Cooking Cleaning Childcare Shopping Collecting Water Total Hours
Low
Middle
High
8.9
8.1
6.1
(5)
(6)
(9)
6.0
7.1
5.7
(5)
(6)
(9)
6.0
6.4
2.6
(7)
(6)
(9)
2.0
2.2
3.7
(5)
(6)
(9)
3.5
2.0
0.0
(8)
(2)
(0)
26.4
25.8
18.1
Conclusions
• Adults work 50% more hours on average in poorest countries
• Largely consistent with model of subsistence consumption needs
(though still trying to reconcile within-country patterns)
• Welfare differences across countries larger than suggested by GDP
• Individuals in low-income countries are “leisure poor” in addition to
being “consumption poor”
Extra Slides
Average Weekly Hours per Adult Males 50 45
KHM
40
PAK
Hours per Week
35
PER COL
LAOVNM TZA
30
GHA
IRQ
UGA RWA
25
IDN MNG
20
TUR USA BWA MUS CYP CZEGRC CHE LVA AUT PRT NOR EST SVN GBR SVK ROM DNK NLD ESP LTU DEU POLHUN ITASWE FIN FRA BEL BGR
15 10 5 0 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita)
9.5
10
10.5
11
Income Group Average
Return
Average Weekly Hours per Adult Females 50 45 40 TZA
Hours per Week
35
KHM LAOVNM
30 25
PER
GHA
MNG
UGA RWA
20
IDN
15 10
BWA USA LVA ESTPRT COL ROM CYP SVN NOR LTUSVK CZE DNK FIN CHE BGR POLHUN SWE AUT GBR GRCFRA MUS DEU ESP NLD BEL ITA TUR
PAK
5 IRQ
0 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita)
9.5
10
10.5
11
Income Group Average
Return
Welfare Analysis for Frisch φ = ∞
1.4
Ratio of Welfare to Consumption
BGR
1.2
POL LTU HUN
ROM UGA
1
RWA
TUR
IRQ
MUS
GHA PAK
LVA
FRA ITA GBR BEL DEU ESP GRC SWE
SVK PRT EST CYP SVN CZE
.8
TZA
IDN
NLD FIN AUT DNK CHE USA
COL NOR
LAO MNG KHM VNM
.6
PER
BWA
.4 6
6.5
7
7.5
8 8.5 9 ln(GDP per Capita)
9.5
10
10.5
11
Average Hours Per Adult by Education
Education All All (Ages 25+, Non-missing Educ.) Ages 25+ Less than Secondary Secondary Completed More than Secondary
Country Income Low Middle 29.3 22.0 33.9 25.3 32.6 38.1 39.5
20.9 29.6 30.9
Group High 19.1 21.0 11.9 23.6 26.9
Average Hours Per Worker by Sector
Sector All All (Non-missing Sec.) Agriculture Manufacturing Services
Country Income Low Middle 40.2 41.6 40.3 41.2 36.8 36.7 44.8 42.9 47.8 42.7
Group High 35.0 35.1 39.4 36.7 34.3