Transcript Título presentación - Stone Center for Latin American Studies

Permanent effects of economic
crises on household welfare:
Evidence and projections from
Argentina’s downturns
Guillermo Cruces
Pablo Gluzmann
CEDLAS – Universidad Nacional de La Plata, CONICET
Luis Felipe López-Calva
UNDP-RBLAC
NIP Network Meeting
New Orleans, Tulane University, April 9 2010
One slide presentation



Objective: estimate the impact of the economic
crisis on household welfare.
Extension: document permanent effects.
Motivation:






© Hugo Ñopo, 2006
Transient or permanent?
Not easy to find (beyond poverty & employment)
Ambiguous impacts for some outcomes
Use data from Argentina’s up and downs to
estimate crisis effects, and extrapolate.
Significant effects on maternal and infant
outcomes, more ambiguous for education.
What is not included: flawless id strategy, fancy
IV estimates.
Motivation and objective
Motivation




Part of a regional UNDP-RBLAC project
on the impact of the recent financial crisis
on household welfare in LAC.
Case study for Argentina.
Objective: estimate the impact of the
economic crisis on household welfare.
Feasible short term estimates:

No ideal data (in fact no data in ARG)
 No flawless identification strategy
Objective



No contemporary data for Argentina… access to
surveys, quality, prices, and GDP.
Exploit Argentina’s up and downs to estimate
past crisis effects and extrapolate.
Some relatively obvious (& easy) impacts:
 Poverty
 Unemployment


Extension (and focus): permanent effects.
Motivation:
 Some
seem obvious, but not easy to document (e.g.:
Ernesto’s paper later today)
 Ambiguous impacts for some outcomes
GDP and changes in GDP
400,000
7%
6%
380,000
5%
360,000
4%
340,000
3%
2%
320,000
1%
0%
300,000
-1%
280,000
-2%
260,000
-3%
240,000
-4%
-5%
220,000
-6%
-7%
II
93
I9
4
IV
94
III
95
II
96
I9
7
IV
97
III
98
II
99
I0
0
IV
00
III
01
II
02
I0
3
IV
03
III
04
II
05
I0
6
IV
06
III
07
II
08
I0
9
200,000
GDP
% of change in GDP
Discussion:
Potential effects of crises
Ambiguous effects of crises:
Ferreira and Schady’s 2009 framework
Aggregate income shocks can have
ambiguous impacts on investments in
child education and health.
 Not necessarily pro-cyclical: income and
substitution effects.
 In F&S: income effect dominates for
poorer countries, with less developed
credit markets, and less protected public
spending.
 Ambiguous effects in middle-income
countries.

Previous evidence

Crises result in higher school enrollment:
 Brazil (Duryea and Arends-Kuenning
 Mexico (McKenzie 2003)
 Argentina (Lopez-Boo
 Peru (Schady 2004)

2008)
Negative shocks to crop prices increase
enrollment:
 Nicaragua (Maluccio
 Brazil (Kruger 2007)

2003)
2005)
Crises increase infant mortality:
 Mexico
(Cutler et al. 2002)
 Peru (Paxson and Schady 2005)

+ This project: Aguero & Valdivia, …
Source: Ferreira and Schady, Aggregate Economic Shocks,
Child Schooling and Child Health, WBRO and PEGNET Presentation, 2009.
Protected?
Social expenditure as percentage of GDP
5.5
8.5
5.0
8.0
4.5
7.5
4.0
7.0
3.5
6.5
3.0
6.0
1993
1994
1995
Education
1996
1997
Health
1998
1999
2000
2001
2002
Employment, assistance, housing & infraestructure
2003
2004
Social Security
2005
2006
Estimation strategies
Estimation strategy


1.
2.

Objective: estimate the impact of the current
crisis on socioeconomic outcomes.
Methodology: projections based on past
experience.
Methodology 1: growth elasticities of relevant
outcomes over 15 years (provincial panels – id
through regional variation).
Methodology 2: estimates of the impact of the
2001 crisis (difference in differences - ad
hoc/upper bound).
Outcomes: schooling, child and overall
poverty, maternal and infant mortality, low
weight at birth.
GDP elasticities (regional data)
Compute growth elasticities of the
outcomes.
 Limitation at the national level: Short time
span – little variation.
 Solution: Exploit time and regional
variability from a panel of per capita GDP
and outcomes at the province (25) level to
identify the relevant elasticities.

Y j t  Y j t1     log GDPpc j t  log GDPpc j t1    ' X jt   ' F j   jt





Yjt denotes the outcome of interest for province j in
time t.
Xjt are a series of covariates (e.g. gender
composition and mean age) for each province
Fj captures regional fixed effects.
 is the semi-elasticity of the selected outcome
with respect to per capita GDP.
Regression with fixed effects at the province level,
weighted by province population, robust standard
errors.
Fixed Effect Model
Strongly significant effects on poverty (expected)
Significant effects on health outcomes
% change in GDPpc
Mean age
Male rate
Constant
Observations
R-squared
Dependent Variable: Change for each socioeconomic indicator
Poverty
Poverty
Poverty
Poverty 1.25 Poverty 2.5
Poverty 4
Maternal
children 1.25 children 2.5
children 4
USD
USD
USD
mortality
USD
USD
USD
-0.4192
-0.6610
-0.5859
-0.2389
-0.4675
-0.5398
-3.3174
(7.28)***
(7.96)***
(6.82)***
(6.59)***
(7.11)***
(6.61)***
(1.96)*
-0.0001
-0.0065
-0.0155
-0.0019
-0.0071
-0.0115
-0.0891
(0.01)
(0.77)
(1.78)*
(0.49)
(1.23)
(1.65)
(0.38)
0.2877
-0.1766
-0.4027
0.0581
-0.1035
-0.2031
-5.4329
(0.39)
(0.19)
(0.40)
(0.13)
(0.15)
(0.24)
(0.25)
-0.1255
0.3044
0.6986
0.0382
0.2849
0.4731
5.3942
(0.24)
(0.49)
(1.05)
(0.13)
(0.64)
(0.84)
(0.35)
235
235
235
235
235
235
244
0.389
0.498
0.405
0.384
0.488
0.454
0.027
Children
mortality
-4.0781
(1.79)*
-0.0340
(0.14)
-39.2225
(1.71)*
19.1320
(1.37)
244
0.070
and no significant effects on educational indicators
No significant negative effects on education
(an ambiguous impact in years of education)
Dependent Variable: Change for each socioeconomic indicator
School
School
School gap 6 School gap
attendance 6 attendance
to 12 years
13 to 17
to 12 years
13 to 17
old
years old
old
years old
% change in GDPpc
(Positives)
% change in GDPpc
(Negatives)
Mean age
Male rate
Constant
Observations
R-squared
0.0180
(0.72)
-0.0344
(0.91)
-0.0011
(0.87)
0.0455
(0.40)
0.0127
(0.17)
235
0.047
-0.0530
(0.49)
0.0552
(0.80)
-0.0065
(1.49)
0.6714
(1.07)
-0.1026
(0.31)
235
0.095
-1.5171
(1.83)*
0.1931
(0.73)
0.0287
(0.95)
5.2069
(1.68)*
-3.3094
(1.74)*
235
0.101
-0.7750
(1.16)
0.2473
(0.88)
0.0100
(0.28)
-0.4839
(0.14)
-0.0579
(0.02)
235
0.038
Years of
Years of
education 6 education 13
to 12 years to 17 years
old
old
0.6840
(1.07)
-0.3157
(0.51)
0.0449
(1.37)
-2.9136
(0.96)
-0.0716
(0.03)
235
0.062
0.4745
(0.57)
-0.1849
(0.67)
0.0355
(0.64)
1.4048
(0.31)
-1.8028
(0.50)
235
0.040
Also: asymmetrical relationship for growth and recession periods
Years of
education
Not working
nor attending
school 13 to
17 years old
1.6455
(2.37)**
-0.8465
(2.70)***
0.0009
(0.03)
-4.7483
(1.50)
2.2137
(1.02)
235
0.100
-0.0152
(0.16)
-0.0461
(0.68)
0.0038
(1.01)
-0.8212
(1.39)
0.2694
(0.88)
235
0.086
Asymmetrical relationship
for growth and crisis periods
15.0%
15%
10.0%
10%
5.0%
5%
0.0%
0%
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
-5.0%
-5%
-10.0%
-10%
-15.0%
-15%
Change in PP of poverty 2.5 USD (inverted)
GDPpc Growth
Larger poverty & child mortality effects
during crises, no effect on maternal mortality
% change in GDPpc
(Possitives)
% change in GDPpc
(Negatives)
Mean age
Male rate
Constant
Observations
R-squared
Dependent Variable: Change for each socioeconomic indicator
Poverty
Poverty
Poverty
Poverty 1.25 Poverty 2.5
Poverty 4
Maternal
children 1.25 children 2.5
children 4
USD
USD
USD
mortality
USD
USD
USD
-0.3993
-0.5389
-0.4470
-0.2417
-0.3499
-0.3980
-7.5769
(2.83)***
(2.86)***
(2.15)**
(2.94)***
(2.64)***
(2.22)**
(2.19)**
-0.4358
-0.7624
-0.7013
-0.2365
-0.5651
-0.6576
0.2378
(6.69)***
(6.37)***
(6.38)***
(4.39)***
(4.75)***
(4.79)***
(0.07)
-0.0003
-0.0080
-0.0173
-0.0019
-0.0087
-0.0133
-0.0339
(0.05)
(0.86)
(1.76)*
(0.44)
(1.34)
(1.69)*
(0.14)
0.2754
-0.2523
-0.4888
0.0598
-0.1764
-0.2910
-2.7856
(0.35)
(0.25)
(0.45)
(0.13)
(0.24)
(0.32)
(0.13)
-0.1126
0.3836
0.7888
0.0363
0.3612
0.5651
2.6402
(0.20)
(0.56)
(1.08)
(0.11)
(0.75)
(0.92)
(0.17)
235
235
235
235
235
235
244
0.390
0.503
0.410
0.384
0.496
0.461
0.033
Children
mortality
3.9059
(0.92)
-10.7419
(4.77)***
-0.1374
(0.60)
-44.1845
(1.93)*
24.2940
(1.75)*
244
0.098
Beyond simple elasticities…
School attendance 6 to 12
Maternal Mortality
1
9,000
6
9,000
8,500
8,500
5
0.995
8,000
8,000
4
7,500
7,500
0.99
7,000
3
7,000
0.985
6,500
6,500
2
6,000
6,000
0.98
1
5,500
5,500
0.975
5,000
1993
1994
1995
1996
1997
1998
1999
2000
School attendance 6 to 12 years old
2001
2002
2003
2004
2005
0
5,000
1993
2006
1994
1995
1996
1997
1998
1999
2000
Maternal mortality
Real GDPpc
School attendance 13 to 17
2001
2002
2003
2004
2005
2006
Real GDPpc
Child Mortality
95%
9,000
25
9,000
8,500
90%
8,500
20
8,000
7,500
85%
8,000
7,500
15
7,000
7,000
10
80%
6,500
6,500
6,000
6,000
5
75%
5,500
5,500
70%
5,000
1993
1994
1995
1996
1997
1998
1999
2000
2001
School attendance 13 to 17 years old
2002
2003
Real GDPpc
2004
2005
2006
0
5,000
1993
1994
1995
1996
1997
1998
1999
Children mortality
2000
2001
2002
Real GDPpc
2003
2004
2005
2006
Upper bounds of crisis effects:
evidence from a “worst-case” scenario
 The evolution of outcomes over the 19992001 recession and the large 2001-2002
crisis reveals a specific effect.
 Not the long(ish) term elasticity: a “crisis”
elasticity.
 Computing a “worst case scenario” impact:
Panel of outcomes at the province (25) –
but calibrate the timing of the “exposure
period” to obtain an upper bound.
Y j t    T  D  T * D   ' X jt   ' F j   jt





Yjt denotes the outcome of interest for province j
in time t.
T is dummy variable that is equal to 1 during the
crisis period (1999-2002, or equivalently 20002003),
D identifies the last year in each period (crisis and
reference).
 jt is an idiosyncratic shock uncorrelated with the
regressors.
 captures the difference in outcomes attributable
to the crisis (interaction term in DiD estimation).
Results...
Not working
School
School
Not working
nor attending
School
attendance 6 to
School
attendance 13
Dependent
nor attending school 13 to 17
attendance 6 to 12 years old in attendance 13 to 17 years old
Variable
school 13 to 17 years old in the
12 years old the first income to 17 years old
in the first
years old
first income
quintil
income quintil
quintil
T = 0 1995-1998
1995-1998
1995-1998
1995-1998
1995-1998
1995-1998
2000-2003
2001-2004
2001-2004
2001-2004
2001-2004
T = 1 2000-2003
0.0041
0.0026
0.1134
0.1573
-0.0825
-0.1309
T
(1.82)*
(0.58)
(4.72)***
(4.04)***
(4.43)***
(3.64)***
0.0035
0.0047
0.0633
0.0857
-0.0519
-0.0900
D
(1.65)
(1.01)
(2.96)***
(2.54)**
(3.23)***
(2.98)***
-0.0092
-0.0146
-0.0882
-0.1132
0.0707
0.1090
T*D
(2.48)**
(1.77)*
(3.24)***
(2.54)**
(3.44)***
(2.63)**
0.0026
0.0037
0.0044
-0.0009
0.0047
0.0111
Male
(1.80)*
(1.12)
(0.55)
(0.06)
(0.88)
(0.71)
0.0427
-0.2773
0.8782
1.6829
-0.2919
-0.0158
Age
(0.25)
(0.67)
(0.93)
(1.00)
(0.40)
(0.01)
0.8858
0.9990
0.2392
-0.0833
0.1403
-0.0894
Constant
(8.04)***
(3.85)***
(0.42)
(0.08)
(0.32)
(0.08)
Observations
R-squared
84
0.549
84
0.351
84
0.828
84
0.712
84
0.806
84
0.636
Maternal
Mortality
Children
Mortality
1995-1998
2000-2003
-0.9466
(2.62)**
-0.4304
(1.07)
1.1508
(1.91)*
0.2052
(0.68)
-17.4014
(0.48)
5.8367
(0.27)
1995-1998
2000-2003
-5.7840
(9.48)***
-2.8863
(6.58)***
2.8380
(3.94)***
0.1703
(0.79)
-47.6226
(1.42)
39.0193
(2.09)**
84
0.811
84
0.925
Effect on poverty according to the elasticity approach
Poverty
Dependent
Children 1.25
Variable
USD
T = 0 1995-1998
T = 1 1999-2002
0.0016
T
(0.20)
0.0038
D
(0.53)
0.1448
T*D
(9.52)***
-0.0028
Male
(0.50)
-0.4138
Age
(0.65)
0.3613
Constant
(0.92)
Observations
R-squared
81
0.917
Poverty
Children 2.5
USD
1995-1998
1999-2002
0.0202
(2.29)**
0.0189
(2.27)**
0.2189
(12.89)***
-0.0140
(2.23)**
-1.4391
(1.66)
1.3216
(2.69)***
81
0.968
Poverty
Children 4 USD
Poverty 1.25
USD
Poverty 2.5
USD
Poverty 4 USD
1995-1998
1999-2002
0.0296
(2.74)***
0.0266
(3.48)***
0.2084
(12.41)***
-0.0283
(4.82)***
-1.4397
(1.96)*
1.9521
(4.48)***
1995-1998
1999-2002
0.0028
(0.62)
0.0014
(0.32)
0.0861
(8.75)***
-0.0004
(0.12)
-0.4562
(0.96)
0.2746
(1.00)
1995-1998
1999-2002
0.0153
(1.94)*
0.0094
(1.25)
0.1623
(9.51)***
-0.0044
(0.79)
-0.9845
(1.14)
0.7254
(1.53)
1995-1998
1999-2002
0.0204
(1.87)*
0.0116
(1.20)
0.1984
(8.86)***
-0.0170
(2.77)***
-1.2957
(1.30)
1.3952
(2.70)***
81
0.981
81
0.917
81
0.952
81
0.967
Conclusion
Conclusion
Regional and time variation in Argentina
provides evidence of permanent effects of
crises through maternal and infant health.
 Asymmetric effects (hysteresis?).
 Middle income country: no major negative
effects on education (some ambiguous),
but negative effects in upper bound
estimates.

Conclusion
Current crisis: difficult to state without
credible GDP (and other) data.
 But probably milder than worst case
estimates.
 Protect social expenditure during crises,
even in middle income countries.

Thanks!
Fixed Effect Model
Projections
Variable
Level at 2006
Official GDP Growth from
2006q4 to 2008q3
Partial elasticity
Extrapolated level at 2008q3
Official GDP Growth from
2008q3 to 2009q1
1
Alternative scenarios
2
3
Extrapolated level to Official
GDP Growth at 2009q1
Extrapolated level to
scenario 1
Extrapoled level to
scenario 2
Extrapolated level to
scenario 3
Poverty
Poverty
children 1.25 children 2.5
USD
USD
8.1%
19.9%
Poverty
children 4
USD
36.7%
Poverty 1.25
USD
Poverty 2.5
USD
Poverty 4
USD
Maternal
mortality
Children
mortality
4.6%
11.5%
22.8%
4.8
22.9
-0.4675
5.0%
-0.5398
15.3%
-3.3174
4.3
-4.0781
22.3
13.9%
-0.4192
2.2%
-0.6610
10.7%
-0.5859
28.5%
-0.2389
1.3%
-0.5%
-1.0%
-5.0%
-10.0%
2.4%
11.0%
28.8%
1.4%
5.2%
15.5%
4.4
22.3
2.6%
11.3%
29.1%
1.5%
5.4%
15.8%
4.4
22.4
4.3%
14.0%
31.4%
2.5%
7.3%
18.0%
4.5
22.5
6.4%
17.3%
34.4%
3.7%
9.6%
20.7%
4.7
22.7
...and new projections
Variable
Level at 2006
Official GDP Growth from
2006q4 to 2008q3
+
Partial elasticity
Extrapolated level at 2008q3
Official GDP Growth from
2008q3 to 2009q1
1
Alternative scenarios
2
3
Extrapolated level to Official
GDP Growth at 2009q1
Extrapoled level to
scenario 1
Extrapolated level to
scenario 2
Extrapolated level to
scenario 3
Poverty
Poverty
children 1.25 children 2.5
USD
USD
19.9%
8.1%
Poverty
children 4
USD
36.7%
Poverty 1.25
USD
Poverty 2.5
USD
Poverty 4
USD
Maternal
mortality
Children
mortality
4.6%
11.5%
22.8%
4.8
22.9
-0.3499
-0.5651
6.6%
-0.3980
-0.6576
17.3%
-7.6
0.2
3.7
3.9
-10.7
23.4
13.9%
-0.3993
-0.4358
2.5%
-0.5389
-0.7624
12.4%
-0.4470
-0.7013
30.4%
-0.2417
-0.2365
1.3%
-0.5%
-1.0%
-5.0%
-10.0%
2.7%
12.7%
30.8%
1.4%
6.9%
17.6%
3.7
23.5
2.9%
13.1%
31.1%
1.5%
7.2%
17.9%
3.7
23.6
4.7%
16.2%
33.9%
2.4%
9.4%
20.6%
3.7
24.0
6.9%
20.0%
37.4%
3.6%
12.2%
23.8%
3.7
24.5
Previous evidence
Rich countries
Middle-income countries
Poor countries
Education outcomes
Positive impact
 United States
Health and nutrition outcomes
Positive impact
 United States
Ambiguous impact
Examples of positive impact
 Mexico
 Brazil
 Argentina
 Peru
 Nicaragua
Examples of negative impact
 Costa Rica
Ambiguous impact
Examples of positive impact
 Colombia
Negative impact
 Indonesia
 Cote d’Ivoire
 Malawi
 South Africa
 (Nicaragua)
Negative impact
 Nicaragua
 India
 Cote d’Ivoire
 Zimbabwe
 Ethiopia
 Tanzania
 Cameroon
 South Africa
Examples of negative impact
 Peru
 Mexico
 Russia
Source: Ferreira and Schady, Aggregate Economic Shocks,
Child Schooling and Child Health, WBRO and PEGNET Presentation, 2009.