Cherry Law, University of Kent
Expert consultation on trade and nutrition
15-16 November 2016, FAO Headquarters, Rome
http://www.fao.org/economic/est/est-events-new/tradenutrition/en/
Trade liberalization and regional dietary patterns in rural india
1. Unintended Consequences Of Trade
Liberalisation on Regional Dietary Pattern
in Rural India
Cherry Law
University of Kent
2. Background
• Nutrition transition in developing countries: the shift of dietary
pattern from the one dominated by traditional staples to the one
high in animal products and other non-cereal food.
• Trade liberalisation is suggested as one of the possible drivers
behind the shift. (Shetty, 2002; Pingali and Khwaja, 2004; Thow
and Hawkes, 2009; Kearney, 2010).
• However, there is limited empirical evidence on how trade may
have affected food consumption pattern.
3. Stylized facts of India
• In 1991, the Indian government launched a wave of extensive
trade reform. The average effective tariff rate was reduced
from around 86% in 1989-90 to about 40% in 1991 (Hasan et
al., 2007)
• The Indian diet has shifted from cereals to animal products
and other food (Rao, 2000; Shetty, 2002; Deaton and Drèze;
2009).
4. Indian food consumption pattern
1970 1980 1990 2000
Cereals 66.3 65.8 63.4 61.7
Starchy Roots 1.7 1.9 1.7 2.0
Fruits & Vegetables 2.8 3.1 3.1 3.7
Animal Products 4.8 5.8 7.1 7.6
Vegetable Oils 4.8 6.5 6.9 8.2
Sugar & Sweeteners 9.3 9.5 9.2 9.7
Others 10.3 7.4 8.4 7.1
Source: FAOSTAT (Animal products include meat, eggs, milk, animal fats, offals, fish and
aquatic products. Others include treenuts, stimulants, alcoholic beverages, spices,
oilcrops, pulses and miscellaneous.)
(% of total per capita calorie intake per day)
5. Distribution of regional average budget
share in rural India, 1987 and 1997
01234
.2 .4 .6 .8
Budget share
1987 1997
Cereals
02468
10
Density
0 .1 .2 .3
Budget share
1987 1997
Eggs/Fish/Meat
Source: the 43rd and 53rd National Sample Survey data
6. Research question
Can the difference in consumption of cereals and animal
products across Indian rural regions from 1987 to 1997 be
attributed to their differential degree of exposure to the trade
reform?
Yes. The trade liberalisation in 1991 had a negative impact on
cereal consumption but a positive one on the consumption of
animal products.
7. Estimation strategy
Following Topalova (2007), the degree of trade protection faced by
rural regions is measured as follow:
• 𝑇𝑛𝑡 : nominal ad valorem tariff faced by industry 𝑛 at time 𝑡.
• 𝑊𝑜𝑟𝑘𝑒𝑟𝑟𝑛,1991 :the number of workers in industry 𝑛 in region 𝑟 in 1991
• 𝑇𝑜𝑡𝑎𝑙 𝑊𝑜𝑟𝑘𝑒𝑟𝑟: total workers in region 𝑟
To address for the bias arisen from the assignment of zero tariff to
non-traded industries, 𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 is instrumented by:
• 𝑇𝑜𝑡𝑎𝑙 𝑇𝑊𝑜𝑟𝑘𝑒𝑟𝑟: total workers in traded industries in region 𝑟
𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 =
𝑛 𝑊𝑜𝑟𝑘𝑒𝑟𝑟𝑛,1991 ∙ 𝑇𝑛𝑡
𝑇𝑜𝑡𝑎𝑙 𝑊𝑜𝑟𝑘𝑒𝑟𝑟,1991
(1)
𝑛𝑠𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 =
𝑛 𝑊𝑜𝑟𝑘𝑒𝑟𝑟𝑛,1991 ∙ 𝑇𝑛𝑡
𝑇𝑜𝑡𝑎𝑙 𝑇𝑊𝑜𝑟𝑘𝑒𝑟𝑟,1991
(2)
8. Estimation strategy
The baseline regression:
• 𝑊𝑟𝑡: the average percentage of food expenditure spent on the aggregated
food group of region r at time 𝑡.
• 𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 : regional measure of exposure to trade liberalization
• 𝐹𝑃𝐼𝑟𝑡 : regional food price index
• 𝐷𝑟: regional dummies.
• 𝜏 𝑡: time fixed effect
𝑊𝑟𝑡 = 𝛼 + 𝛽1 𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 + 𝛽2 𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 ∙ 𝑙𝑛𝐹𝑃𝐼𝑟𝑡 + 𝛾1 𝑙𝑛𝐹𝑃𝐼𝑟𝑡
+𝑑 𝑟 𝐷𝑟 + 𝜏 𝑡 + 𝜀 𝑟𝑡
(3)
9. Data
• Food consumption data: 43rd, 45th to 53rd National Sample
Survey
• Industrial employment data: 1991 Indian Census
• Agricultural tariffs: Topalova (2010) and World Integrated
Trade System
• Manufacturing tariffs: Aghion et al. (2008)
10. Main results
Table 4. Trade liberalization and food consumption in rural India
Note: All regressions are estimated with constant, region and time dummies. Tariff is instrumented by non-scaled tariff. Robust standard
errors clustered at state-year level are given in parenthesis. ***Denotes significant at the 1% level, **at 5% level, *at 10 % level.
Cereals Eggs/ Fish/ Meat
(1) (2) (3) (4)
Tariff 0.279** 0.260* -0.256*** -0.258***
(0.135) (0.152) (0.075) (0.079)
Tariff*Ln FPI -0.133** -0.137** 0.110*** 0.116***
(0.058) (0.065) (0.035) (0.037)
Ln FPI -0.116*** -0.125*** 0.048*** 0.051***
(0.022) (0.022) (0.013) (0.013)
FDI reform 0.166*** -0.064**
(0.036) (0.026)
Delicensing 0.020 -0.003
(0.021) (0.009)
Observation 821 821 821 821
R-squared 0.942 0.944 0.932 0.934
The average tariff cut (14.6 %) is associated with a 1.01 % point increase in the food budget
share on EFM and a 0.53 % point decrease in that of cereals (relative to national trend).
11. Note: All regressions are estimated with constant, region and time dummies and controls for other reforms. Tariff is instrumented
by scaled tariff in column 2 to 4. Robust standard errors clustered at state-year level are given in parenthesis. ***Denotes
significant at the 1% level, **at 5% level, *at 10 % level
Dependent variable Average budget share
Median budget
share
Ln(Calorie)
Definition of tariff
Non-scaled Tarifft
(Hasan et al, 2007)
IV-Tarifft-1 IV-Tarifft IV-Tarifft
(1) (2) (3) (4)
Panel A: Regional consumption of cereals
Tariff 0.060* 0.224 0.390** 0.356
(0.035) (0.150) (0.167) (0.392)
Tariff*Ln FPI -0.032** -0.148** -0.208*** -0.020
(0.015) (0.061) (0.072) (0.166)
Ln FPI -0.124*** -0.005 -0.121*** 0.010
(0.021) (0.019) (0.022) (0.041)
Observation 821 744 821 821
R-squared 0.946 0.937 0.939 0.788
Panel B: Regional consumption of eggs/ fish/ meat
Tariff -0.059*** -0.132** -0.204*** -1.403
(0.017) (0.062) (0.077) (1.163)
Tariff*Ln FPI 0.027*** 0.067*** 0.111*** 0.964*
(0.008) (0.026) (0.035) (0.545)
Ln FPI 0.051*** 0.007 0.046*** 0.142
(0.012) (0.009) (0.013) (0.118)
Observation 821 744 821 819
R-squared 0.939 0.928 0.935 0.825
Robustness check I
12. Note: All regressions are estimated with constant, region and time dummies and controls for other reforms. Tariff is instrumented
by non-scaled tariff . For column 4, only observations prior to 1992 are used and the difference between regional tariffs, FPI and
their interaction from 𝑡 + 6 and 𝑡 + 2 are the regressors. Robust standard errors clustered at state-year level are given in
parenthesis. ***Denotes significant at the 1% level, **at 5% level, *at 10 % level
Dependent variable 𝑊𝑟1991 − 𝑊𝑟1987 𝑊𝑟𝑡
Definition of tariff IV –(Tariff1997-Tariff1992) IV-(Tarifft+6-Tarifft+2)
(1) (2)
Panel A: Regional consumption of cereals
Tariff 0.156 -0.442
(0.840) (0.338)
Tariff*Ln FPI 0.461 0.351
(0.899) (0.392)
Ln FPI 0.044 -0.019
(0.070) (0.025)
Observation 73 371
R-squared 0.098 0.949
Panel B: Regional consumption of eggs/ fish/ meat
Tariff 0.251 0.137
(0.262) (0.165)
Tariff*Ln FPI -0.272 -0.119
(0.322) (0.187)
Ln FPI -0.020 -0.003
(0.042) (0.011)
Observation 73 371
R-squared 0.061 0.924
Robustness check II – Placebo tests
14. Income effect
• Demand for animal products is generally more income elastic
than that for cereals (Mittal, 2007 and Kumar et al, 2011).
• To identify the trade impact on income at regional level,
• 𝑙𝑛𝑀𝑃𝐶𝐸𝑟𝑡: logarithm of regional average per capita monthly
expenditure
𝑙𝑛𝑀𝑃𝐶𝐸𝑟𝑡 = 𝛼 + 𝛽3 𝑡𝑎𝑟𝑖𝑓𝑓𝑟𝑡 + 𝑑 𝑟 𝐷𝑟 + 𝜏 𝑡 + 𝜀 𝑟𝑡 (4)
15. Income effect
Table 7. Trade liberalization and total expenditure in rural India
Note: All regressions are estimated with constant, region and time dummies. MPCNE is the monthly per capita non-food
expenditure. Tariff is instrumented by non-scaled tariff. Robust standard errors clustered at state-year level are given in
parenthesis. ***Denotes significant at the 1% level, **at 5% level, *at 10 % level.
Log real MPCE Log real MPCNE
(1) (2) (3) (4)
Tariff 0.064 0.033 -0.075 -0.094
(0.192) (0.193) (0.311) (0.315)
FDI reform 0.201 0.159
(0.127) (0.202)
Delicensing -0.013 -0.085
(0.062) (0.106)
Observations 821 821 821 821
R-squared 0.966 0.966 0.926 0.927
16. Price effect
• Marchand (2012) finds evidence for a positive linkage between the
state-level domestic price of a good and the tariff faced by the
corresponding industry in India.
Table 8. Trade liberalization and real food prices in rural India
Note: All regressions are estimated with constant, region and time dummies. Other food includes pulses, vegetables, fruits and milk
products. Tariff is instrumented by non-scaled tariff. Robust standard errors clustered at state-year level are given in parenthesis.
***Denotes significant at the 1% level, **at 5% level, *at 10 % level.
Cereals Eggs/Fish/Meat Edible Oils Other food
(1) (2) (3) (4) (5) (6) (7) (8)
Tariff -0.006 -0.033 -0.344 -0.394 0.730*** 0.615*** -0.644** -0.718***
(0.138) (0.138) (0.264) (0.262) (0.245) (0.231) (0.261) (0.261)
FDI reform 0.144 0.288* 0.698*** 0.390***
(0.100) (0.163) (0.158) (0.138)
Delicensing 0.049 0.063 0.073 0.180***
(0.040) (0.066) (0.054) (0.066)
Observation
s 821 821 821 821 821 821 821 821
R-squared 0.978 0.978 0.946 0.947 0.951 0.955 0.965 0.966
17. Determinants of regional food
consumption
Table 9. Determinants of dietary patterns in rural India
Note: Contemporary taste is measured by 𝜃 𝑟𝑡, the regional component of the budget share equation which cannot be explained by
prices and total food expenditure. Robust standard errors clustered at state-year level are given in parenthesis. All regressions are
estimated with constant, region and time dummies.. ***Denotes significant at the 1% level, **at 5% level, *at 10 % level.
Cereals Eggs/ Fish/ Meat
(1) (2)
Log real MPCE -0.053*** 0.008***
(0.009) (0.003)
Log real price
Cereals 0.071*** -0.027***
(0.024) (0.003)
Eggs/ fish/meat -0.028*** 0.005**
(0.006) (0.002)
Edible oils 0.025* 0.003
(0.013) (0.004)
Other food 0.032*** -0.020***
(0.007) (0.003)
Contemporary taste
Cereals 0.446***
(0.023)
Eggs/ fish/meat 0.766***
(0.028)
Observations 821 821
R-squared 0.976 0.987
18. Taste effect
• The reduction in tariffs enhances consumers’ access to
varieties of food that were not previously available to them
(Pingali and Khwaja, 2004).
• The increased interaction with foreign culture through the
opening of trade may create a demonstration effect which
promotes the adoption of western dietary pattern (James,
1987).
19. Regional food taste
Following Atkin(2013), the regional taste indicator (𝜃𝑟) is constructed
as follow:
• 𝑤𝑖ℎ : share of food expenditure that household 𝑖 spent on food group ℎ
• 𝑃ℎ: logarithm of prices of food groups.
• 𝑚 ∶the monthly per capita food expenditure adjusted by the price index (𝑃𝑟
∗
).
• 𝑍𝑖: A vector of household characteristics
• 𝐷𝑟: regional dummies
𝑤𝑖ℎ = 𝜃𝑟 𝐷𝑟 +
ℎ
𝛾ℎ 𝑙𝑛 𝑃ℎ + 𝛽 𝑙𝑛
𝑚𝑖
𝑃𝑟
∗ + δ𝑍𝑖 + 𝜀𝑖
(5)
20. Taste effect
Cereals Eggs/ Fish/ Meat
(1) (2) (3) (4)
Tariff 0.051 0.042 -0.113*** -0.105***
(0.124) (0.122) (0.029) (0.029)
FDI reform 0.036 -0.049*
(0.059) (0.027)
Delicensing 0.050 -0.008
(0.037) (0.011)
Observations 821 821 821 821
R-squared 0.994 0.994 0.983 0.984
Table 9. Trade liberalization and regional food tastes in rural India
Note: Contemporary taste is measured by 𝜃 𝑟𝑡, the regional component of the budget share equation which cannot be explained by
prices and total food expenditure. All regressions are estimated with constant, region and time dummies. Tariff is instrumented by
non-scaled tariff. Robust standard errors clustered at state-year level are given in parenthesis. ***Denotes significant at the 1% level,
**at 5% level, *at 10 % level.
21. Summary
• Regions experiencing higher exposure to foreign competition
consume relatively less cereals and more animal products.
• Trade liberalisation may have both positive and negative
impact on health development in developing countries.
• Apart from income and food prices, food taste is found to be
an important channel of transmission between trade reforms
and food consumption.