assessing the causal impact of tobacco expenditure on households’ spending patterns grieve chelwa...
TRANSCRIPT
Assessing the causal impact of tobacco expenditure on households’ spending
patternsGrieve Chelwa
Emerging Research ProgrammeC A P E T O W N 22 – 26 June 2015
Tobacco and household spending patterns 22–26 June 2015
Presentation Outline
• Introduction & motivation
• Brief review of literature
• Conceptual framework
• Empirical strategy
• Data
• Results
• Summary & conclusion
Tobacco and household spending patterns 22–26 June 2015
Introduction & motivation
• 5 million people die p/a directly (WHO, 2010) & 600,000 passively (Oberg et al., 2010)
• Additional cost of tobacco which until recently received little attention– Tobacco consumption might displace other
goods & services– Likely to be more severe among income-
constrained households
Tobacco and household spending patterns 22–26 June 2015
Introduction & motivation
• Zambia
• Tobacco increasingly becoming important in households’ budgets– Yet households have not become substantially well-off
• Important to find out which goods & services tobacco displaces, if at all any
0
50
100
150
200
250
0
20
40
60
80
100
120
140
160
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Per
capi
ta c
igar
ette
cons
umpti
on
GD
P pe
r ca
pita
in Z
MK1
0,00
0
Per Capita GDP Per capita cigarette consumption
Tobacco and household spending patterns 22–26 June 2015
Contribution
• Use expenditure data from SSA to do this work– Only other study I am aware of is Koch and
Tshiswaka-Kashalala (2008)
• Use the standard instrumental variable from the literature, adult-sex ratio, but relax the strict exogeneity assumption using Nevo and Rosen (2012)
Tobacco and household spending patterns 22–26 June 2015
Preview of findings
• Even after relaxing the strict exogeneity assumption, confirm many of the findings in the literature– Tobacco expenditure crowds out food, schooling,
clothing, water, transportation & equipment maintenance
• But unlike previous studies, I do not find instances of crowding in of alcohol– More likely to be a correlation than a causal
relationship
Tobacco and household spending patterns 22–26 June 2015
Literature review
• Tobacco costs conceptualized in two ways– Costs on the macroeconomy via death, increased
health care expenditure & lost productivity (Chaloupka & Warner; Kang et al, 2003; Max et al, 2004; Liu et al, 2006)
– Displacement costs (crowding out)
Tobacco and household spending patterns 22–26 June 2015
Literature review
• 1st round of studies– Simple comparison of expenditure profiles
• Efroymson et al. (2001)
• Second round– Control for observable confounders
• Busch et al. (2004); Wang et al (2006)
Tobacco and household spending patterns 22–26 June 2015
Literature review
• 3rd round– Control for observable and unobservable
confounders using the method of instrumental variables
• John (2008); Koch and Tshiswaka-Kashalala (2008); Pu et al (2008)
• Block & Webb (2009)
• My approach aligned with 3rd round, except I use the standard instrumental variable with less stringent assumptions
Tobacco and household spending patterns 22–26 June 2015
Conceptual framework
• Following John (2008), assume each household maximizes a single utility fn s.t. budget constraint
• Problem results in Marshallian demands functions
• Suppose that a hhold first spends on tobacco and then allocates expenditure to other goods:
Tobacco and household spending patterns 22–26 June 2015
Empirical strategy
• Estimate a QUAIDS (Banks et al., 1997)
• Possibility that:
– Instrument for d using adult-sex ratio
– Instrument for & using the log of assets & its square respectively
• Relax the strict exclusion restriction later
Tobacco and household spending patterns 22–26 June 2015
Instrumental variables: A digression
• Suppose your model is:
𝑿
𝒀
𝒁Validity requirement
Exclusion restriction
Corr (Z, X) 0
Corr (Z, u) 0
Tobacco and household spending patterns 22–26 June 2015
Data
• Use 2006 round of the Living Conditions Monitoring Survey (LCMS)
• About 18,000 Households
• Nationally representative
Tobacco and household spending patterns 22–26 June 2015
Relaxing the exclusion restriction
• Concede that
• But assume:
• These assumptions much less stringent that standard IV assumptions
Tobacco and household spending patterns 22–26 June 2015
Relaxing the exclusion restriction
• Propose new instrument:
• Where
• For , is a “perfect” instrument
• We do not observe but, given previous assumptions, know it’s in [0,1] interval
Tobacco and household spending patterns 22–26 June 2015
Summary & conclusions
• Establish a pattern of crowding out related to socioeconomic status– Food for poor and rural– Health for rich– Schooling for rich and poor– Clothing for urban, rural and poor– Water for urban and rich– Electricity for urban– Alternative energy for poor– Transportation for urban and rich
Tobacco and household spending patterns 22–26 June 2015
Summary & conclusions
• Unable to find evidence of crowding in of alcohol– More likely a correlation than causal
relationship
• Overall: A broader accounting of tobacco’s costs in Zambia should include:– Under nutrition– Under investment in education, among others