Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Food Policy
journal homepage: www.elsevier.com/locate/foodpol
Keywords: Seasonality is a salient feature of rural livelihoods and particularly within agriculture the demand for labor
Rural livelihoods varies with the seasons and weather. In low-income countries, agriculture employs almost two-thirds of the labor
Seasonal labor force and incomes from labor are a major determinant of welfare. Therefore, an appropriate model re-
Rural labor markets presentation of rural labor markets is critical when analyzing agricultural and food policies. Economy-wide
Economy-wide modeling
models are commonly used for ex-ante policy analysis, but have so far ignored the influence of seasonality,
Policy analysis
Bhutan
implicitly assuming separability of seasonal labor demand and supply. This study relaxes that assumption using a
computable general equilibrium (CGE) model calibrated to the Bhutanese economy as an illustrative case. Using
JEL classification: model setups with and without seasonal labor markets, a cereal export ban of India is simulated leading to higher
C680 import prices for Bhutan. Results demonstrate that neglecting the influence of seasons on rural labor markets
J220 systematically biases model results. Assuming homogeneity of labor units, i.e., allowing substitution across
J430 seasons, understates the impacts of policy changes on rural wage rates, distorts households' labor-leisure trade-
O130 off decisions and overstates agricultural supply response. Given the widespread use of economy-wide models, the
results are important for understanding the implications of domestic and global policy changes for agriculture
and welfare in developing economies.
⁎
Corresponding author.
E-mail addresses: a.feuerbacher@uni-hohenheim.de (A. Feuerbacher), cdukpa@moaf.gov.bt (C. Dukpa), grethe@hu-berlin.de, grethehx@hu-berlin.de (H. Grethe).
https://doi.org/10.1016/j.foodpol.2020.101875
Received 25 June 2019; Received in revised form 27 February 2020; Accepted 29 February 2020
0306-9192/ © 2020 Elsevier Ltd. All rights reserved.
Please cite this article as: Arndt Feuerbacher, et al., Food Policy, https://doi.org/10.1016/j.foodpol.2020.101875
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
multi-period models incorporating stochastic sequential decision followed, by details of the scenario implemented and the model settings
making (e.g., Antle, 1983; Berger and Troost, 2014). In the case of in section 4. The results are analyzed in section 5, followed by reports
micro-level models, the absence of seasonality of labor demand distorts on sensitivity analyses in section 6. The paper ends with a discussion of
model outcomes, since it allows the use of more resources than those the conclusions and their implications.
available (Hazell and Norton, 1986, pp. 42–46). Micro-level models
neither capture consumption and forward and backward linkages
2. Background
within economies nor do they depict the operations of labor markets.
But, economy-wide models such as computable general equilibrium
2.1. Seasonality of labor markets
(CGE) models, which are extensively used for policy analysis
(Devarajan and Robinson, 2005), have neglected the seasonality of
Demand for seasonal labor can be derived from a production func-
labor demand. They commonly assume strong separability between
tion x a = f (l, k, n, i , , ) , where x a is a vector of outputs of activity a;
production and consumption decisions, which implies perfect inter-
l, k , and n (labor, capital and land) are vectors of primary inputs, i is a
temporal substitution in decisions by households about labor supply
vector of intermediate inputs, σ is a vector of substitution elasticities
across a year. Such annual model setups sidestep the reality of the
and a vector of efficiency factors. The labor inputs can be defined
seasonal fluctuations of labor demand in agriculture addressed else-
using skill types, e.g., by education or occupation categories, and sea-
where.
sonal types, e.g., by months. Nested constant elasticity of substitution
Only two studies that incorporate seasonal demands for labor in an
(CES) functions allow flexibility in substitution possibilities such that,
economy-wide framework are known. Finnoff and Tschirhart (2008)
for instance, the substitution possibilities between labor of different
model the interrelations between the Alaskan fishery and tourism sector
skill types can be greater than the substitution possibilities between
as non-consumptive services of the marine ecosystem in a bioeconomic
labor of different seasonal types within each skill type.1 The associated
general equilibrium model. Within a three-sector recursive-dynamic
first order conditions identify the factor returns and allow wage rates
CGE model, seasonal wage differentials are identified as workers expect
for each type of labor to vary across seasons reflecting the seasonal
higher payment within the fishing season compensating them for the
patterns of labor demand.
likelihood of facing unemployment in the off-season. Filipski et al.
The elasticities of substitution of seasonal labor depend on the
(2017) analyze the impacts of Saffron price volatility on rural liveli-
economic activity. Cropping activities, e.g., paddy rice, may follow
hoods in a Moroccan village characterized by a strong gender based
rigid sequences defined by nature, as noted by Sen (1966), which im-
division of labor. The rural labor market is segmented by gender and
plies no, or limited, substitution possibilities. Non-farm activities may
two seasons, the Saffron harvest (<1 month) and the growing season
provide complementary employment options for rural labor with a
(rest of the year). This allows analyzing how changes in seasonal labor
counter seasonal pattern relative to cropping activities (Reardon et al.,
demand impact male and female wages and the substitution between
1998). For manufacturing activities, the degree of substitution of labor
seasonal labor and leisure within and outside the Saffron harvest
across seasons will be greater.
season. Both studies rely on case study specific model specifications at
Labor supply adjusts to the (rigid) seasonal labor demand, de-
either the village or sub-national level. They only distinguish between
pending on the ease with which households can trade-off time for lei-
two seasons and give no (Finnoff and Tschirhart, 2008) or limited
sure with participation in the labor force in one period, and vice versa
(Filipski et al., 2017) consideration of the wider effects of labor sea-
in other periods. Economic theory argues that individuals maximize
sonality on rural economies and the agricultural sector. Filipski et al.
utility by trading off consumption and leisure - uh = u (x c, h , leic, h) , where
(2017) highlight bottlenecks of female labor supply during the Saffron
uh is a vector of household utilities, x c, h is a vector of consumption of
harvest season, but they do not include intraseasonal substitution of
commodities, c, by household, h; and leic,h is a vector of household
leisure. Finally, in both studies consumption and leisure are gross
specific domestic services. Domestic services comprise leisure and ser-
complements.
vices from social reproduction (e.g., house maintenance, child care and
This study contributes to the literature by reporting the first
cooking). These are produced outside the System of National Account’s
economy-wide model that incorporates seasonality of labor and leisure
(SNA) production boundary (ISWGNA, 2009 para 1.41), which only
at a national scale, with intraseasonal substitution of leisure. This sea-
encompasses goods and services that could be sold on markets at un-
sonal model is compared to an annual model (with annualized instead
iquely defined prices (ISWGNA, 2009 para 6.24). Leisure and social
of seasonal labor markets). The model and experiments scrutinize the
reproduction thus need to be modelled using household specific activ-
standard (implicit) assumption of economy-wide models that seasonal
ities (see Section 3.1.3).
labor demand and supply are separable, i.e., labor types are measured
The labor force (persons employed, self-employed and seeking em-
in ‘homogenous units’. With some 63% of the labor force in low-income
ployment) is defined internationally as those “individuals willing to
countries employed in agriculture (World Bank, 2019), and incomes
supply labour to undertake an activity included in the SNA production
from labor being a primary determinant of welfare, the study addresses
boundary” (ISWGNA, 2009. 405 para 19.5). The size of an economy’s
how the neglect of seasonal labor demand may bias the results from
labor force, and its composition, can change seasonally as households
economy-wide models.
trade-off hours supplied inside the production boundary and social re-
Bhutan serves as an illustrative case, for which data on seasonal
production activities. This is particularly relevant for economies char-
labor and leisure quantities were collected and compiled as a social
acterized by large variations in the seasonal demand and limited sub-
accounting matrix (SAM) (Feuerbacher et al., 2017). A range of food-
stitution possibilities for social reproduction.
price shock simulations provide evidence that ignoring seasonal labor
Modelling labor supply is more complex because it is necessary to
demand systematically biases the results with implications for agri-
allow for the observed periods of apparent underemployment and those
cultural supply response, income and household welfare. The analyses
periods when there are high demands for labor, e.g., for harvesting and
and various robustness checks demonstrate that core results apply to
weeding. If labor demand is less time dependent, the availability of
countries with large shares of labor in seasonal activities and underline
labor for production activities is less time critical. The time dependency
the importance of considering seasonal labor in economy-wide ana-
of labor used to produce leisure is, arguably, more flexible than labor
lyses, particularly in developing agrarian economies.
The rest of the paper is organized as follows. The next section re-
views the definition of the labor force and seasonal labor and gives an 1
Regular nested CES functions “are globally well-behaved and can provide a
overview of the Bhutanese economy. The third section reviews the local approximation to any globally well-behaved cost function” (Perroni and
model and data used and details about model calibration. This is Rutherford, 1995).
2
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
used for social reproduction. Substituting time use across periods, in Agricultural households in the temperate zones (AEZ3) are closer to
response to changes in wage rates, means taking less leisure during urban centers (particularly Thimphu and Paro in Western Bhutan).
periods of peak labor demand with compensating increases in leisure They have more diversified income sources, particularly the landless
during periods of slacker labor demand.2 Such patterns of labor demand agricultural households. Per capita incomes decline with decreasing
across production cycles are a feature of agricultural production in both altitude and increasing distance from the Western urban areas. While
developed and developing economies and are an explanation for the climate at higher altitudes is a constraint, it allows farmers to
“[apparent] underemployment” in developing countries (Ruthenberg, supply out of season produce to India and Bangladesh (MoEA, 2012).
1971, p. 78).3 The main cash crop exports are vegetables, potatoes and apples from
the temperate regions (AEZ3), and citrus, cardamom and ginger from
the subtropical zones. The cultivation of cereals, particularly rice and
2.2. Country context Bhutan maize, occupies two-thirds of cropland. Cereal yields are low compared
to other South Asian countries, reflecting low adoption levels of agro-
Bhutan, known for its development philosophy of Gross National chemicals (Feuerbacher et al., 2018).
Happiness, has a population of about 735,000 of which 62% reside in Bhutan maintains a free trade agreement with India. With sub-
rural areas (NSB, 2018). The country is landlocked in the mountainous sistence consumption absorbing the largest share of farmers’ cereal
Eastern Himalayas, with its main access routes being south into India. production, Bhutan is reliant on cereal imports from India; 35% of total
Bhutan’s poverty headcount ratio declined from 23% to 8% between cereal demand (Feuerbacher et al., 2017). Increasing the degree of food
2007 and 2017 (NSB and World Bank, 2017). Economic development self-sufficiency, for inter alia, reasons of sovereignty, is a political
largely benefitted the urban population and poverty became a rural priority in Bhutan. Recently, India has imposed export bans on cereals,
phenomenon, particularly in the Southern and Eastern districts of but exempted Bhutan. If India imposed a cereals export ban on Bhutan
Bhutan. Compared to other South Asian countries, Bhutan scores well the import costs would increase sharply due to the higher transporta-
on most nutrition indicators with only 2.8% of the population living tion costs.
under the food poverty line of 2,124 kcal in 2012 (NSB and World Bank,
2012). However, stunted growth, in children younger than five years 3. Data and model
old, and anemia, which affects women and children, remain challenges
(Atwood et al., 2014). 3.1. The model
Bhutan is a peasant economy, despite increasing investments in the
formal economy, with about 50% of the labor force working in agri- The model is a single country comparative-static CGE model de-
culture (NSB and ADB, 2012) – similar to other South Asian countries veloped from the STAGE2 (McDonald and Thierfelder, 2015) and
(World Bank, 2019). Agriculture is dominated by semi-subsistence STAGE_DEV models (Aragie et al., 2017). There are three substantive
smallholders cultivating, on average, 1.2 ha (NSB, 2018). Only 3% of developments.
Bhutan’s land area is used for farming, which is driven by labor and
capital scarcity rather than land scarcity. About 21% of arable land is 3.1.1. Production system
left fallow, with an increasing trend over recent years (MoAF, 2016). The nesting of the production functions is extended to incorporate
Labor shortages are the main farming constraint, reflecting the very low seasonal labor and fertilizer – a land saving intermediate input (see
population density (19 people per km2) (NSB, 2018) and limited scope Fig. 1). The top nests aggregate non-land saving intermediate inputs
for mechanization, e.g., power tillers for land preparation, due to steep and value-added components using CES or Leontief technologies. At the
slopes of rainfed land and narrow rice terraces. second level, Leontief technologies aggregate non-land saving inter-
Bhutan is a “monsoon economy” (Reardon et al., 1998). The onset of mediate inputs and CES technologies aggregate the fertilizer-land and
monsoon rains determines the cropping pattern of rice, with seasonal capital-labor composites. The third level combines land and aggregate
labor bottlenecks observed during the rice transplanting and harvesting fertilizers, where the latter is a combination of chemical and organic
periods. Farm labor demand is relatively low in slack periods, i.e., fertilizers (manure) from the fourth level; both use CES technologies.
outside the main cropping seasons, when farmers pursue non-farming The capital-labor composite is a combination of capital and aggregate
activities such as textile weaving and forestry. There are considerable labor, with aggregate labor being a combination of seasonal and per-
regional differences in monsoon rain patterns and climatic conditions. manent labor (level 4) – seasonal labor inputs are shaded. Permanent
Three major agroecological zones (AEZs) have been identified: the labor is a combination of skilled and unskilled labor, while seasonal
humid subtropical zone (AEZ1, below 1200 m of altitude), the dry labor is an aggregate of farm labor from all 12 months. All the capital
subtropical zone (AEZ2, 1200 – 1800 m) and the temperate zone (AEZ3, and labor components are aggregated with CES technologies. The point
above 1800 m). Agricultural households and cultivated land are almost estimates for the CES elasticities, used at each level of the production
equally distributed across these zones. system, together with the sources used are detailed in the supplemen-
Rural settlements are scattered and difficult to access, despite a tary material (Appendix A).
growing rural road network. Except for pastoralists migrating between
summer and winter camps, there is little evidence and no data on rural- 3.1.2. The demand system
rural seasonal migration. Rural livelihoods are dependent on crop The demand system in STAGE_DEV is extended to a three level CES-
agriculture. Cattle husbandry, which dominates livestock agriculture, is LES-CES nest, where LES is the Linear Expenditure System. In the first
land extensive, but labor intensive; reflecting low productivity and (CES) level households trade-off aggregate leisure/social reproduction
widespread discouragement of slaughtering due to Buddhist beliefs. services and aggregate consumption. At the second (LES) level house-
holds determine the optimum consumption levels of aggregate com-
2
modities, e.g., processed food, services, other goods, subject to
Another form of adjustment in developing, and developed, economies is
minimum (subsistence) levels of consumption. At the third (CES) level
seasonal labor migration; slack periods in some areas coincide with peak per-
households choose optimum mixes of ‘natural’ commodities subject to
iods in other areas.
3
In high-income countries farmworkers, historically, worked shorter weeks relative commodity prices and the constraints of preferences, income,
during slack (winter) seasons and longer weeks in peak seasons and worked available labor resources and subsistence requirements. This setup al-
more overtime during peak seasons. Other examples of adaptation to seasonal lows households to substitute similar goods and services, e.g., rice,
workloads include the use of contractors, e.g., the harvest teams that work their maize and other cereals. The income elasticities of demand for com-
way north harvesting grains in the U.S. modity groups at the LES level were estimated using cross-sectional
3
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
household data from the 2012 Bhutan Living Standard Survey crop calendars and cost of production studies (Appendix D). The de-
(Feuerbacher, 2019). The CES parameters to aggregate the commodity mands for ‘skilled’ and ‘unskilled’ labor and all other factor are not
groups are documented in the supplement (Appendix B). seasonal.
Seasonal farm labor dominates the rural labor market, although
3.1.3. Labor-leisure and social reproduction non-farm workers like teachers, shop owners, etc., also reside in rural
Each household has an activity that produces household specific leisure areas. The rural labor market is also segmented by the three AEZs, to
(including social reproduction), which requires inputs of labor services reflect variations in climate and altitude and differences in growing
owned by the associated households. These production functions are de- conditions in Bhutan. Non-farm labor is classified as unskilled and
fined as CES aggregates of labor by permanent and seasonally employed skilled (permanent) labor. Farm labor is demanded by agricultural, food
labor, i.e., there are labor substitution possibilities in the production of processing and off-farm activities. Non-farm labor is demanded by
household leisure. By defining activities that produce commodities outside manufacturing and service activities. Households are classified as
the production boundary as household specific, i.e., they can only be pro- agricultural if at least one household member is reported to work
duced and consumed by the household, there are uniquely defined prices within agriculture. Agricultural households are sub-divided by AEZs
for commodities and services produced inside and outside the boundary and access to land, i.e., farm and landless households. There are five
(Aragie et al., 2017). The approach in this model is a generalization of a land accounts (rainfed, irrigated, orchard, pasture and forestland),
method used by Fontana and Wood (2000). which are sub-divided by AEZs. There are four capital factors: two li-
This specification means that only those labor services owned by vestock capital accounts (cattle and other animals), and two physical
each household not used to produce leisure can be supplied to other capital factors by ownership (agricultural household and incorporated
activities. This requires that labor markets are cleared, in quantity business enterprise owned).
terms, at the household level. It is common in CGE models (e.g., Hertel, Seasonal activities are those that demand seasonal labor. Their
1997; Lofgren et al., 2002; McDonald and Thierfelder, 2015) to clear shares in total output, required person-days and characteristics of sea-
labor markets by implicitly assuming that the proportions of each type sonal labor demand are reported in Table 1. Cropping activities have
of labor supplied by each household are constant. In the model this is ‘rigid’ labor demands, i.e., operations such as transplanting or har-
defined as a FDf , a FSIins, f , where FDf , a is the factor, f, de- vesting of paddy are performed in specific time windows (Appendix E).
ins
manded by activity, a, and FSIins, f is the factor supplied by institution, Other seasonal activities such as livestock husbandry or textile weaving
ins, which includes households (Aragie et al., 2017). Labor market by farm households are performed with flexibility in labor demand.
clearing at the household level ensures that labor services used by ac- Some crops, in Bhutan, are cultivated in different seasons or with dif-
tivities within the production boundary, and outside the production ferent cropping patterns, e.g., double or single cropping of maize; these
boundary for social reproduction and leisure, cannot exceed those activities were disaggregated accordingly. Cropping activities account
available to the household. This formulation endogenizes the distribu- for 52% of seasonal activities’ output, by value, and 43% of total
tion of income, i.e., the distribution of earned income among house- person-days. The high labor intensity of rice production − 250 person-
holds is a variable dependent on the quantities of labor each household days per hectare4 – means that rice cultivation accounts for ~ 38% of
sells to activities. cropping labor. There are 37.9 million person-days provided by about
170,000 farmers, i.e., an average of 225 working days per person; this
excludes labor used for social reproduction and leisure. Overall, em-
3.2. The data
ployment in seasonal activities accounts for 48% of Bhutan’s labor days.
The available data do not allow for a meaningful separation of the
The database is a SAM for Bhutan in 2012 (Feuerbacher et al., 2017)
accounts for social reproduction and leisure; they are therefore treated
with satellite accounts for quantities of factors and elasticities (see
as a single account – ‘leisure’
Appendix C for Bhutan’s economic structure as represented in the
SAM). The seasonal demands for farm labor were estimated for 12
monthly periods from primary data, seasonal calendars, secondary data
from the (national) 2012 agricultural sample survey (MoAF, 2013), 4
Maize has half the labor intensity per hectare.
4
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Table 1
Seasonal activities represented in the 2012 Social Accounting Matrix for Bhutan.
Activity %-Share in total seasonal Person-days (in Share in total person-days employed in Seasonal labor substitution
output value thousand) production elasticity σ
5
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Table 2
Changes in Bhutan’s import prices of cereals following an Indian export ban on cereals.
Commodity Base ExpBan (Export Ban)
Share in total cereal Import intensity Import unit price Quantity imported Share imported from Increase of cereal import prices (CIP)a
demand (2012) (2012) India (in foreign currency)
(%) (%) (USD/ton) Tons (%) (%)
Source: Author’s calculation based on 2012 Bhutan SAM and World Bank, 2017.
a
Computed by dividing the increase in import cost of 117 US-$ ton-1 by the import unit price times the share imported from India.
wage elasticity of labor supply (Boeters and Savard, 2013). The wage world market to Bhutan is estimated to be 117 USD/ton (World Bank,
elasticity of labor supply is set to 0.15 following Goldberg (2016), who 2019). This is the applied mark-up on world cereal prices (Table 2),
estimated the wage elasticity in rural Malawi.6 The resultant elasticities which increases the carriage insurance and freight paid price (in USD)
of labor-leisure substitution, lei, X , are 1.54 for agricultural and 2.00 for of imported cereals by 35%. Additional scenarios with changes in im-
non-agricultural households. Sensitivity analysis demonstrates the ro- port prices of cereals ranging between −50% and +50% were run to
bustness of model results across a range of elasticities. assess the results’ robustness to the magnitudes of price changes.
Labor demand patterns differ markedly across activities, particularly for
cropping activities with rigid labor demand (Fig. 2). Potential seasonal labor 4.1. Model setups and closure
bottlenecks are driven by the paddy transplanting and harvesting seasons,
which vary by AEZ, e.g., in AEZ1 transplanting is in June-July and har- The impacts of changes in cereal import prices on Bhutan’s economy
vesting in November-December. Livestock activities, comprising more than are simulated with two different behavioral specifications; a model with
a third of total seasonal labor (Table 1), and leisure have more flexible labor seasonal labor data (henceforth, the seasonal model) and a model with
demand. Non-farm activities have highly flexible labor demand and follow a annual labor data (henceforth, the annual model). The seasonal model
counter-cyclical pattern, i.e., their labor demand is lowest during peak depicts seasonality as described in the model structure, with twelve
periods and highest during lean seasons. monthly farm labor accounts per AEZ (as in nest L5.1 in Fig. 1). In
contrast, the annual model has one farm labor account per AEZ: this
3.3.2. Production of leisure excludes seasonality and replicates the standard assumption of homo-
There are no known estimates of substitution elasticities for different genous labor. Imports are modelled as imperfect substitutes for do-
labor types (skilled, unskilled or seasonal) in the production of leisure. An mestically produced goods and services (using CES functions)
elasticity of 0.50 is assumed, i.e., inelastic substitution. Total seasonal leisure (Armington, 1969) as are exports (using constant elasticity of trans-
is a CES aggregate that allows households to substitute seasonal leisure formation functions) (de Melo and Robinson, 1989). The elasticities
across time subject to an (implicit) intertemporal elasticity of substitution used are reported in the supplement (Appendix A). The consumer price
(IES). The intra-seasonal substitution of leisure is analogous to the inter- index (CPI) serves as the model’s numéraire. The small country as-
temporal substitution of labor supply due to short-term (or transitory) wage sumption is used, i.e., fixed world market prices, and the external
changes in the life-cycle labor supply theory (Lucas and Rapping, 1969). balance (foreign savings) is cleared by a flexible exchange rate. The
Most estimates for the IES range between 0.00 and 0.50 (Card, 1994; models are investment-driven; the investment quantities are fixed and
Heckman and MaCurdy, 1982; Lee, 2001), but other estimates range be- the investment-savings account is cleared by flexible household saving
tween 0.01 (Connolly, 2008) and 1.2 (Fehr and Goette, 2007). Few esti- rates. Government savings are fixed in real terms and income tax rates
mates for developing countries use shadow wage estimations. Fafchamps vary (additively) to clear the government account. Alternative macro-
(1993) reports an infinite elasticity, while Skoufias (1996) estimates IES for economic closures are evaluated in the sensitivity analysis.
female workers ranging between 0.057 and 0.079. There are no estimates
for Bhutan. Recognizing that leisure here includes social reproduction, an 4.2. Factor market clearing
elasticity of 0.20 is used (see Table 1). The results are robust when varying
the IES between 0.1 and 1.2. If the IES is infinite, there is, by definition, no A medium-term horizon is imposed for clearing the factor accounts. The
seasonality. supplies of each type of capital are fixed and perfectly mobile across ac-
tivities.7 All labor accounts, skilled, unskilled and seasonal labor, are per-
4. Model scenarios fectly mobile across activities: they are segmented by characteristics (skill-
level, AEZ and time-period) which cannot be altered, i.e., no labor mobility
The core scenario is rooted in a possible policy change: an export across characteristics. Land is perfectly mobile within its land-type across
ban on cereals by India requiring Bhutan to purchase cereals from cropping activities, i.e., the export ban is assumed to persist over the
countries other than India. This scenario is referred to as ExpBan. medium-to long-term allowing farmers to adjust their land allocation. A
During periods of world food price spikes, India has repeatedly applied land supply curve (Eickhout et al., 2009) is implemented to account for the
export taxes or even bans to stabilize domestic food prices (e.g., Villoria 21% of arable land left fallow in Bhutan. Land supplies depend on the land
and Mghenyi, 2016). So far Bhutan was exempted from these measures rental rate and approaches an asymptote of maximum land supply as the
(GoI, 2014). Bhutan has no border crossings with China, its Northern factor price for land goes towards infinity. An inelastic supply of land, with a
neighbor, hence cereal imports from the rest of the world must arrive by price elasticity, nAL = 0.8, is used following the estimate of Eickhout et al.
land through India or by air. The average import cost per ton from the (2009) for the South Asian region. Model results are robust to variation of
this parameter.
6
Barrett et al. (2008) report a wage elasticity of 0.12 which is based on
7
shadow wage estimations. The simulation results hardly differ, if this elasticity Capital is only mobile between activities that employ the respective capital
is used instead. category in the base SAM.
6
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
7
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Table 4
Percentage changes in commodity prices.
SAM Sector or group Import %-changes
sectors name intensity (%)
Import price (in Import price in domestic Producer price Consumer price Output
foreign currency
currency)
Annual Seasonal Annual Seasonal Annual Seasonal Annual Seasonal
Model model Model model Model model Model model
1–10 Agriculture 30.2 12.0 10.2 10.2 1.7 2.1 3.1 3.4 1.1 0.2
1–7 Crops 26.8 23.2 20.9 20.9 2.4 3.1 5.5 6.1 2.7 1.5
1–3 Cereals 34.6 34.6 32.5 32.4 6.6 8.4 12.3 13.7 10.5 7.5
1 Rice, milled 42.7 35.9 33.8 33.8 8.8 11.2 15.5 17.2 16.7 12.0
2 Maize 3.1 51.7 49.4 49.3 3.4 4.8 3.5 4.6 0.2 −0.6
3 Other cereals 41.4 24.8 22.9 22.8 3.2 3.4 10.6 10.6 4.9 5.1
4–7 Other crops 18.7 −0.4 −0.4 −0.6 −0.6 −2.3 −2.4
4 Vegetables 25.2 0.9 0.8 0.3 0.2 −2.5 −2.4
a
5 Potatoes 9.5 see note below 1.0 0.8 1.7 1.6 −5.8 −5.7
6 Spices 23.7 −1.2 −1.2 −1.0 −0.9 −2.3 −2.7
7 Fruits 19.5 −1.0 −1.2 −1.0 −1.2 −0.7 −0.5
8–10 Livestock 35.1 0.5 0.3 −0.4 −0.5 −2.1 −2.2
11–12 Forestry 7.8 0.3 −0.9 0.1 −0.9 −0.7 −0.1
13 Mining 59.6 −1.6 −1.6 −1.6 −1.6 0.1 0.2
14–19 Food processing 46.5 0.9 0.9 0.0 0.0 −5.5 −5.8
20–21 Textile weaving 38.4 −0.7 −1.3 −1.1 −1.4 −1.2 0.0
17–22 Other 81.6 −1.5 −1.6 −1.5 −1.6 0.0 0.2
manufacturing
23 Electricity 0.3 −1.6 −1.6 −1.7 −1.7 0.3 0.4
24 Construction 5.9 −1.5 −1.6 −1.5 −1.6 0.0 0.0
26–32 Services 11.8 −1.5 −1.5 −1.5 −1.5 0.0 0.0
a
As a response to the appreciating nominal exchange rate, the import prices of all commodities except cereals, and export prices of all commodities decline by 1.5
and 1.6%, respectively.
The seasonal model generates consistently greater increase in farm production. The land rent and supply of irrigated land can increase
wages, in all AEZs (Table 5). Surprisingly, at first sight, this is accom- substantially more in the annual model, because it is complemented by
panied by a decrease in farm labor supply in the first two zones and much cheaper labor. The actual difference in wage rates for labor used
only a low increase in labor supply in AEZ3. Both, the differences in in rice cultivation between both models is much higher than reported in
seasonal wage changes and the composition of aggregate leisure are the Table 5, as in the seasonal model rice cultivation requires labor in
main reasons (see next section). With less labor for cereal production, periods of wage spikes (see also the results on activity-specific changes
total arable land increases only by 2.5% in the seasonal model, com- in price of labor and Fig. 5). The land expansion, via the asymptotic
pared to 3.1% in the annual model. Overall, land rents increase land supply curve, is dependent on the change in real land prices (de-
strongly, because of increased demand for irrigated land for rice flated by the PPI), explaining the increases in rainfed land supply de-
spite low or slightly negative changes in nominal land prices.
Table 5
5.3.1. Farm labor wages and supply
Changes in factor prices and factor supply.
The changes in seasonal wages reflect the seasonal pattern of labor
Factor accounts Change in factor price (%) Changes in factor supply (%) requirement for cereal cultivation, predominantly rice (Fig. 3a). For
example, the strong increases in seasonal wages in the low altitude zone
Annual Seasonal Annual Seasonal
(AEZ1) during June, July and November, coincide with the periods of
Skilled labor −1.6 −1.6 −0.2 −0.2 land preparation, transplanting and harvesting of rice. Due to the colder
Unskilled labor −1.5 −1.6 −0.3 −0.3 climate, rice field operations occur about a month earlier in AEZ2 vis-à-
Farm labor – nationala 2.1 3.8 0.5 −0.1
vis AEZ1, and another month earlier in AEZ3. Hence, the pattern of
Farm labor - AEZ1 2.4 4.9 0.6 −0.7
Farm labor - AEZ2 2.8 4.7 0.6 −0.1 seasonal increases in farm wages occur in phase shifts across the AEZs.
Farm labor - AEZ3 1.4 2.1 0.4 0.3 Additional labor demand for maize production further exacerbates
Arable land national 5.4 3.4 3.1 2.5 seasonal labor shortages. Both single and double cropping of maize are
Rainfed land - national −0.9 −1.1 0.2 0.0 activities in the model. Of the two systems, double-cropping of maize
Rainfed land - AEZ1 −1.0 −1.8 0.1 −0.5
Rainfed land - AEZ2 −1.2 −0.6 −0.1 0.4
overlaps most with the rice transplanting period. Consequently, output
Rainfed land - AEZ3 −0.5 −0.9 0.5 0.2 of double-cropped maize decreases at greater rates in AEZ1 and AEZ2 in
Irrigated land - national 25.8 17.8 12.0 9.7 the seasonal model while in AEZ2 the output of single cropped maize
Irrigated land - AEZ1 23.1 14.8 13.0 9.7 increases.
Irrigated land - AEZ2 26.5 17.0 10.8 8.8
The results for seasonal wages cannot be compared with empirical
Irrigated land - AEZ3 28.2 22.9 12.3 11.2
Orchard land – national −2.2 −2.2 −0.9 −0.9 seasonal wage data. Farm labor transactions (hiring labor in and out)
Pasture land - national −5.7 −4.9 0.0 0.0 predominantly take place in months of peak labor demand and are not
Forestland - national −1.9 −1.4 −0.7 −0.2 recorded over time. Particularly in the lean season, there is no spot
Livestock - national 6.2 2.8 0.0 0.0 market for farm labor, i.e., the reported seasonal wages are shadow
Unincorporated capital 12.1 7.8 0.0 0.0
Incorporated capital −1.6 −1.6 0.0 0.0
wages. Nevertheless, the results are consistent with field observations
and records of farmers reporting labor shortages in the rice trans-
a
Changes in aggregate farm labor computed based on farm labor used within planting and harvesting season. Seasonal wages decline in months with
productive boundary, excluding leisure. low or no labor demand by cereal activities (Fig. 3a). Non-cereal
8
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
activities reduce their demand for seasonal labor due to reduced con- how the assumption of perfect intertemporal substitution affects the
sumer demand and a decline in competitiveness. Furthermore, due to mechanisms of farm labor mobility and supply in the annual model.
the rigid pattern of labor demand for cropping, non-cereal crop activ- First, it allows for perfect mobility of farm labor across activities, in-
ities also have to reduce labor demand in those months with little or no dependent of the periods in which labor is supplied or demanded (see
competition with cereal activities. also the results on reallocation of labor). Second, the changes in the
Farm wages increase less in the annual model, but the increased aggregate price of leisure for agricultural households is mostly de-
supply of farm labor is greater (Fig. 3b). The difference is explained by termined by the annual change in farm wage. Consequently, the
Avg. of
20
seasonal model, 4.9
11.9 12.3
10 Annual model, 2.4
0.8 1.5
0
-2.1 -0.8
-2.9 -3.0 -2.9 -3.5 -2.3
-10
30 27.5
20 Avg. of
seasonal model, 4.7 14.2 14.3
0
-2.2 -1.0 -1.0 -1.2
-3.4 -3.1 -2.8 -2.4
-10
20 Avg. of 17.4
seasonal model, 2.1 12.8
10
Annual model, 1.4
3.9
1.6
0
-2.0 -0.7 -1.1 -1.3 -2.1 -1.8
-3.3 -2.2
-10
Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
Months
Seasonal Annual Avg. of
model model seasonal model
Fig. 3a. Changes in farm labor wages. Source: Own model results.
9
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
Months
Seasonal Annual Avg. of
model model seasonal model
Fig. 3b. Changes in farm labor supply. Source: Own model results.
increase in the composite wage results in an increase in the aggregate direction (see Appendix F).
price of leisure (Table 6). In this model, despite substantial increases in The effects of the labor-leisure trade-off are less straightforward in
household income, labor supply by agricultural households increases, the seasonal model. The change in the aggregate price of leisure
reflecting the domination of the substitution effect of labor moving out (Table 6) differs from the change in labor income, mainly because the
of leisure over the expansion effect, which works in the opposite share of leisure in time use varies over seasons (see Fig. 2). In periods of
10
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Note: A negative (positive) expansion effect (E) reported in the last column means that the change in utility is dominant leading to positive (negative) labor supply. A negative (positive) substitution effect denotes that the
labor shortages, seasonal leisure is scarce, because most household time
is absorbed by activities within the production boundary. Conversely,
Seasonal model
in lean months disproportionately more leisure time is available to
Domination of expansion (E) or
E
E
S
S
S
S
S
S
S
wages (see Fig. 3b). Accordingly, the composite price of seasonal leisure
either drops or increases at substantially lower rates than the average
Annual model
S
S
S
% changes in household total labor
Seasonal model
model, the perfect mobility across activities allows for unrealistic re-
allocations of labor from counter-cyclical activities, like forestry and
textile production, to cereal production (Fig. 4). Counter-cyclical ac-
Annual model
0.15
0.23
0.13
0.47
0.10
0.07
37.1% in the annual model. Most striking are the differences of labor
supply from leisure. Consistent with Sen’s observation, the annual
model leads to an effective increase of labor endowment. Assuming
−0.71
−0.28
−0.30
0.78
0.27
1.24
0.61
0.71
0.37
accounts for 18.1% of the reallocated labor in the annual model. In the
seasonal model, with heterogeneity of labor over time, labor supplied
−0.70
−0.30
−0.32
utility
demand for leisure accounting for 6.5% of the reallocated labor. These
mechanisms and channels explain the greater reallocation of labor in
Seasonal model
% change in composite price of
the annual model: 3.0% of total available farm labor flows to cereal
activities, compared to 1.9% in the seasonal model.
−0.46
−0.66
−0.51
−1.64
−1.58
−1.63
0.92
0.63
0.48
of labor demand. Because the changes in seasonal wages can differ, the
price changes of composite farm labor are activity specific. Composite
−0.34
−1.59
−1.54
−1.58
leisure
1.31
0.78
1.90
1.50
0.76
price changes of agricultural labor are highest for activities that depend
on labor supplied during bottleneck periods such as for rice and maize
% change in composite wage
−0.50
−1.55
−1.58
−1.56
2.00
1.11
2.53
1.86
0.98
Determinants of households’ labor supply via the labor-leisure trade-off.
−0.58
−1.52
−1.55
−1.53
model
1.48
0.73
1.98
1.40
0.74
−0.18
−1.01
−0.80
−1.28
−1.12
model
food price rises disproportionately affect their cost of living. The impacts
0.43
0.70
0.66
0.58
−0.04
−1.06
−0.81
−1.25
−1.17
model
Farm household
Farm household
welfare measures. First, the relatively higher farm labor wages and lower
Unskilled households
Landless
Landless
whose consumption and welfare increase more (or decrease less) in the
annual model. And second, the lower aggregate prices of leisure result in
Households
AEZ1
AEZ2
AEZ3
11
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
1200
800
1000 person-days
400
-400
Fig. 4. Net reallocation of farm labor. Note: Measured in thousand person-days across seasonal activities and leisure aggregated at the national level.
15
Seasonal model Annual model
11.9 Avg. of seasonal model
10 9.7
% changes
-5
Rice Double Single Other Veg. early Veg. late Potato Spices Fruits Cattle Other Dairy Community Textile
cropping cropping cereals season season husbandry animals production forestry weaving
maize maize
Rural economic activities in AEZ1
Fig. 5. Changes in the activity specific price of composite seasonal labor within AEZ1.
12
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Fig. 6. Aggregate changes in farm wages, farm labor supply, agricultural supply and agricultural household welfare depending on changes in cereal import prices.
Note: Welfare changes are measured by the Slutsky equivalent variation incl. leisure expressed as a share of households’ base expenditure.
supply hardly increases. In comparison with the annual model, culti- households owning land.
vation becomes less labor intensive (measured in person-days per
hectare) and increases in cereal production are accompanied by 6. Robustness of model results
stronger declines in non-cereal output. The last panel (Fig. 6d) shows
that the systematic bias in the changes of factor prices and supply also Sensitivity analyses were conducted to test the robustness of model
results in different changes in income and welfare for landless and farm
Table 8
Description of sensitivity analyses.
Sensitivity analysis Description
Sigmasea The seasonal labor substitution elasticity, a,t , is varied by ± 50%. The elasticity of formerly rigid activities remains zero when decreasing the elasticity and
is set equal to 0.1 when increasing the elasticity.
Intertemp The IES set at 0.2 is varied to encompass the empirically plausible range from 0.1 to 1.2.
MinMaxLei The income elasticity of labor supply, H , Y of agricultural households is varied between the minimum and maximum possible quantities of leisure. This
results in lower bound income elasticities of labor supply of 0.25 (AEZ1), 0.18 (AEZ2) and 0.14 (AEZ3) and an upper bound elasticity of 1.38 for all AEZs
(such that the time endowment equals 365 days).
LandEla The land supply elasticity, n , is varied by ± 50%
SeasMigr The model is changed to allow for rural-rural seasonal migration among AEZs. The database is modified, assuming that each seasonal activity receives about
10% of its farm labor demand from the other two regions. Two simulations are run assuming seasonal labor to be substitutable across AEZs with a CES
elasticity of 0.8 and 2, respectively.
13
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
Crop Crop Cereal Cereal Arable land Land Farm labor Farm labor Reallocated
output prices output prices supply prices wages supply farm labor
4.0
+50%
3.5
Min Reference (Difference in results
+50%
0.1 between annual-seasonal)
Deviation from reference by
3.0 sensitivity analysis variant
Percentage point difference
0.8
+50%
2.5 Max -50%
2 +50% Min
-50% 0.1 +50%
0.0
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Reference
Sigmasea
IES
MinMaxLei
LandEla
SeasMigr
Fig. 7. Sensitivity analysis of variations in model parametrization. Red bars report the absolute percentage point difference in main indicators between the seasonal
and annual model in the reference setting. Yellow bars show the positive and negative deviation from the reference model in percentage points for changes in model
parameters described in Table 8. Labels at the lower and upper end of the yellow bars refer to the respective change in the model parametrization.
results to changes in model parametrization and setup (Table 8).9 The differences. Accounting for rural-rural seasonal migration substantially
base elasticities used were estimated, derived from the literature (Ap- reduces seasonal wage differentials in the seasonal model and leads to
pendix A), or, where no estimates were available, imposed. The elas- smaller model differences. But the patterns of seasonal labor demand
ticities varied are those that have most influence on the difference in across AEZs are not complementary and peak labor months partially
results between the annual and seasonal model: details of the sensitivity overlap, which limits the potential of labor exchanges (Appendix E).
analyses are reported in Table 8. In addition, a variant of the model
setup is included allowing for seasonal rural-rural migration across 7. Discussion
AEZs with different degrees of spatial labor mobility. Seasonal migra-
tion was not included in the core analysis, due to its minor role for Seasonality affects rural livelihoods in multiple ways, e.g. through
Bhutan and a lack of data. seasonal food shortages and hunger (Vaitla et al., 2009), seasonal in-
The sensitivity analyses demonstrate that changes in main model cidences of diseases and births (Chambers et al., 1981; Devereux et al.,
parameters do not change the signs on the key results although they do 2012) and seasonal labor shortages and underemployment. The season-
influence the magnitudes; hence the main findings from the core analysis ality of labor demand is particularly relevant as labor units cannot be
remain valid (Fig. 7). Variations in the intertemporal substitution of leisure substituted across time. While seasonality of labor is a recognized phe-
(sigmsea) produced the largest changes in the results (measured in per- nomenon within agricultural economics, most economy-wide models
centage points). Given that seasonal leisure includes reproduction activ- have, implicitly, assumed perfect intertemporal labor substitution. This
ities, which have limited substitution possibilities in low income countries, study relaxes this assumption and the results demonstrate the importance
an IES of the upper end (1.2) is high. The differences in model results of accounting for seasonality of labor in low-income economies.
remain substantial and would only become very small when applying a Ignoring seasonality produces a systematic upward bias in model
very large IES (differences in model results disappeared with an IES of >8). results, particularly agricultural supply and welfare changes, due to
The second largest changes come from variations in the seasonal overstating the labor supply response through the assumption of perfect
labor substitution elasticity. When the assumption of non-substitut- intertemporal substitution. This conclusion is logically consistent in a
ability of seasonal labor, i.e. Leontief technology, is relaxed (substitu- constrained optimization problem: the presumption of perfect inter-
tion elasticity of 0.1) cropping activities can substitute labor originally temporal substitution relaxes a constraint. This study identifies two
required in peak periods with labor from other periods in which wages main sources for the biases. First, ignoring the seasonality of labor
increase at lower rates or even decline. In practice, some substitution is (annual model) overestimates the labor mobility between economic
possible among field operations, e.g., reductions in the labor demand activities with different seasonal labor demand patterns. This produces
for transplanting paddy (“sloppier transplanting”) may be compensated unrealistic labor reallocations, e.g., labor used to produce rice, a
by increased diligence when weeding; although such substitutions may summer activity, is drawn from forestry, a winter activity. And second,
result in reduced productivity. This is over and above the flexibility neglecting the seasonality of leisure distorts households’ labor-leisure
offered by reducing leisure time within a month. trade-off decisions, where leisure time is available in slack periods, i.e.,
The remaining variations all result in relatively small changes. The counter-cyclical to the cropping periods. Consequently, models with an
model differences are amplified/reduced with lower/higher calibrated annual labor market specification will report a stronger labor supply
quantities of leisure; greater (assumed) quantities of leisure dampen the response than models with seasonal labor markets.
seasonal constraints. A lower elasticity of land supply mainly reduces These systematic upward biases in agricultural supply responses and
the agriculture supply response in both models. A short-term analysis, structural changes are inter alia relevant when assessing the economy-wide
with no land expansion and reallocation, results in very low model effects of food security (e.g., Golub et al., 2013) or agricultural trade lib-
eralization scenarios (e.g., Anderson et al., 2006) and can be important to
the distribution of income and welfare results. For example, the seasonal
9
In addition, the macroeconomic closure settings were varied by either fixing labor specification results in higher farm wages and substantially lower
the exchange rate or using a savings driven closure. Differences in model results land rents in case of a cereal import price increase in Bhutan, and will thus
were very low and are therefore not reported. produce greater benefits for landless households compared to farm
14
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
7.1. Potential extensions and avenues for further research Declaration of Competing Interest
The analyses can be further developed. Data limitations preclude the None.
differentiation of labor types by gender and age profile and the differ-
entiation of ‘leisure’ activities between ‘pure’ leisure and social reproduc- Acknowledgements
tion. The implications of relaxing these limitations will depend upon the
extent to which roles are rigidly segmented by gender, school holidays This study was funded by a field research grant of the fiat panis
coincide with peak labor demands and the opportunity costs of trading-off foundation (Ulm, Germany) and by a PhD scholarship of the University
leisure and social reproduction for agricultural labor services change. of Hohenheim. Institutional support by the Humboldt-Universität zu
Further research is also needed to improve the understanding of seasonal Berlin and Bhutan’s Ministry of Agriculture and Forests is gratefully
activities in and outside of the production boundary, particularly con- acknowledged. The authors thank three anonymous referees for their
cerning the role of gender and social reproduction (Fontana and Wood, valuable and constructive comments that helped to improve the article.
2000). Such developments may increase or decrease the degree of inter- Furthermore, the authors would like to thank Kinley Tshering, Tulsi
temporal substitution, but they cannot fully relax the constraint imposed Gurung, Phub Dorji, Mahesh Ghimiray, Thomas Daum, Jonas
by imperfect intertemporal labor substitution. Luckmann, Anton Orlov, and Khalid Siddig as well as participants of the
Similarly, the model could be formulated as a dynamic model to 30th International Conference of Agricultural Economists in Vancouver
evaluate, inter alia, the implications of demographic change (see Aragie for their helpful comments on previous versions of this study.
et al., 2017) and the introduction of labor-saving technology changes.
Particularly in case of the latter, a model with seasonal labor markets Appendix A. Supplementary material
would allow for an improved assessment of technologies, which miti-
gate seasonal labor bottlenecks, e.g., the mechanization of paddy Supplementary data to this article can be found online at https://
transplanting. Further model developments in this direction should doi.org/10.1016/j.foodpol.2020.101875.
account for the nuances of seasonality which have not been addressed
in this article, such as the limited substitutability of machinery with References
specific seasonal usability, e.g., a plough cannot be transformed into a
harvester in the short run and vice versa. Machinery also has a limited Anderson, K., Martin, W., van der Mensbrugghe, D., 2006. Doha merchandise trade re-
seasonal capacity, i.e., it can only be used for a certain amount of time, form: what is at stake for developing countries? World Bank Econ. Rev. 20 (2),
while farmers often demand labor-saving machinery at the same time.10 169–195. https://doi.org/10.1093/wber/lhj009.
Antle, J.M., 1983. Sequential decision making in production models. Am. J. Agric. Econ.
In other contexts it would be important to account for seasonality of 65 (2), 282–290. https://doi.org/10.2307/1240874.
irrigation water flows or that increasing heat stress due to climate Aragie, E., Dudu, H., Ferrari, E., Mainar Causapé, A., McDonald, S., Thierfelder, K., 2017.
change reduces labor productivity seasonally, but not necessarily all STAGE_DEV: a variant of the STAGE model to analyse developing countries.
Publications Office of the European Union, Brussels.
year round. These aspects may be addressed using an multi-period Armington, P.S., 1969. A theory of demand for products distinguished by place of pro-
model with sequential decision making as already explored in the duction. IMF Staff Papers 16 (1), 159–178.
micro-level model literature (Antle, 1983; Fafchamps, 1993; Taylor and Arndt, C., Benfica, R., Maximiano, N., Nucifora, A.M.D., Thurlow, J.T., 2008. Higher fuel
and food prices: impacts and responses for Mozambique. Agric. Econ. 39 (August),
Charlton, 2019).
497–511. https://doi.org/10.1111/j.1574-0862.2008.00355.x.
Atwood, S., Nagpal, S., Mbuya, N., Laviolette, L., 2014. Nutrition in Bhutan: Situational
Analysis and Policy Recommendations, Washington DC, USA. <http://documents.
10 worldbank.org/curated/en/397081468227331146/Nutrition-in-Bhutan-situational-
We are grateful to an anonymous referee for highlighting this.
15
A. Feuerbacher, et al. Food Policy xxx (xxxx) xxxx
16