Food Security

, Volume 6, Issue 5, pp 649–670

Food (In)security and its drivers: insights from trends and opportunities in rural Mozambique

Original Paper

DOI: 10.1007/s12571-014-0381-1

Cite this article as:
Mabiso, A., Cunguara, B. & Benfica, R. Food Sec. (2014) 6: 649. doi:10.1007/s12571-014-0381-1

Abstract

We used multiple rounds of nationally representative agricultural survey data to analyze the trends and drivers of food insecurity in rural Mozambique. Reduced-form Probit models were estimated to explain the observed trends as a function of underlying drivers and factors related to agricultural policy interventions. Despite rapid macroeconomic growth, food insecurity in the rural areas had increased from 42.9 % in 2002 to 47.8 % in 2008. Significant inequalities were also observed in the distribution of food insecurity with a substantial disadvantage to the bottom quintile households and rural households located in the Northern provinces. Limited progress on several drivers of agricultural production and food access as well as geographic disparities appear to explain a significant part of the food insecurity trends and distribution. Whether the indicator was use of improved farm inputs and technology, receipt of agricultural extension services, farm production, or cash income, progress did not occur. This implies that to achieve broad-based food security in rural Mozambique, interventions may need to focus on addressing these drivers to increase agricultural productivity while enhancing resilience to price and weather shocks. Interventions must also be spatially targeted and tailored to each segment of the population.

Keywords

Rural food security Food policy Calorie consumption Rural Mozambique 

Introduction

Mozambique is one of the world’s fastest growing economies, with an average GDP growth rate of 6.5 percent recorded in the last two decades (World Bank 2013). At the same time, volatility in growth has significantly declined in the last decade suggesting increased resilience to macroeconomic shocks (Fig. 1). However the trends in poverty reduction and food security appear to have been considerably less stellar (Arndt et al. 2012; Alfani et al. 2012; Arndt et al. 2006; also see Table 1). Compared to 2003 when the estimated consumption poverty rate was 54.1 percent, the rate stood slightly higher at 54.7 percent in 2008.1 Also, the proportion of food insecure households in the rural areas increased from 43 % in 2002 to approximately 48 % in 2008. The global food price crisis of 2008 is likely to have contributed to this lack of progress in poverty reduction and food security, particularly in the rural southern provinces that are more dependent on grain imports due to lower crop production potential (Arndt et al. 2008; Tschirley and Jayne 2010). This emphasizes that food security continues to be a complex development challenge in Mozambique despite economic progress. This is particularly the case in rural areas which account for 68 percent of the total population and where over 70 percent of the households are engaged in farming (World Bank 2013; IMF, 2007). While farming is their main economic activity, rural households also earn a sizeable share of income (37 percent) outside of farming and nearly a third of their food consumption is sourced from market purchases (Diogo et al. 2008).
Fig. 1

Growth rate of agriculture value added and GDP in Mozambique (1990–2011)

Table 1

Selected macroeconomic indicators for Mozambique

Macroeconomic indicator

1996

2003

2008/09

GINI index

44.5

47.1

45.7

Income share held by highest 20 %

50.7

53.3

51.5

Poverty headcount ratio at national poverty line (% of population)

69.4

54.1

54.7

Poverty headcount ratio at rural poverty line (% of rural population)

71.3

55.3

56.9

Poverty headcount ratio at urban poverty line (% of urban population)

62.0

51.5

49.6

Agriculture, value added (annual % growth)

8.9

5.4

9.1

Exports of goods and services (annual % growth)

28.4

29.5

0.3

GDP growth (annual %)

7.4

6.0

6.8

GDP per capita growth (annual %)

4.3

3.1

4.1

World Bank development indicators databank (World Bank 2013)

As such, government has made efforts to improve food marketing to bolster food security in rural Mozambique. For example, to enable food purchases and dampen volatility in grain prices, the Mozambican government has constructed silos with 50,000 metric ton capacity for grain storage in Tete province and there are plans to expand the capacity to 143,000 metric tons (Tschirley and Abdula 2007; Tostão and Tschirley 2010). In addition, there have been some improvements in the physical infrastructure such as the bridge across the Zambezi River, linking surplus agricultural production areas in the north to deficit areas of the south (Tschirley and Jayne 2010). Such efforts have contributed to Mozambique’s significant reduction in food aid receipts over the course of the last two decades. This is in the face of a series of floods and droughts, which left vulnerable rural households continuing to subsist on grain imports and food aid. Compared to 1992 when approximately 83 percent of Mozambicans were dependent on food aid, largely because of war and drought (Tschirley et al. 1996; Abdulai et al. 2004), annual food aid receipts have declined substantially to less than 200,000 tons in recent years (Fig. 2).
Fig. 2

Food aid deliveries to Mozambique (1988–2011)

The government has also developed several policies and a 5-year poverty reduction strategy (referred to by its Portuguese acronym PARPA - Plano de Acção para a Redução da Pobreza Absoluta), which have potential to enhance food security through various channels. For example, planned strategic investments in irrigation, market infrastructure, extension services, commodity market information, animal traction, and other agricultural technologies could contribute to increased crop productivity, which in turn could enhance food availability and incomes - thus, food security (Mather 2009; Thurlow 2012). Also, planned public investments in roads could reduce transaction costs of agricultural marketing while new investments in education may increase household participation in more remunerative non-farm employment in the long run.2

While these multiple efforts are important, it appears that the emphasis is now shifting away from food security and agriculture towards the energy sector where massive investments are being made. With newfound energy resources, government and foreign direct investments in the energy sector are projected to accelerate economic growth and dwarf the contributions of agriculture to Mozambique’s GDP (Polgreen 2012). Given these prospects for sustained and accelerated economic growth, largely driven by the energy boom, it is not clear if Mozambique will achieve broad-based poverty reduction and food security in the rural areas. This paper argues that, as Mozambique enters its energy boom, it should sustain and increase investments in agriculture that enhance productivity and promote resilience through inter-sectoral linkages to ensure broad-based poverty reduction and food security, particularly in the rural areas where agriculture continues to be the mainstay that undergirds food security. To substantiate the argument, the paper retrospectively analyzes food security trends in rural Mozambique over the period 1996–2008 when overall economic growth exceeded 6 percent per year, with a maximum 12 percent growth in 2001. The analysis also tracks the progress made on the underlying drivers of food security to give a picture of where Mozambique stands on rural food insecurity vis-à-vis its drivers. The findings illuminate areas that the government could focus on to promote broad-based rural food security as the energy boom era takes off.

Measuring food insecurity

While food security is a widely used concept, there is a longstanding debate on how best to measure it.3 Over the years, many indicators and indexes have been proposed. Some key ones include household calorie acquisition, individual calorie intake, a range of dietary diversity indicators,4 food frequency scores, coping strategies index, subjective or experiential indicators and, more recently, the global food security index (Maxwell and Frankenberger 1992; Riley and Moock 1995; Barrett 2002; EIU (Economist Intelligence Unit) 2012). The plethora of indicators and indexes is largely because food security entails multiple dimensions including availability, access, utilization and stability that are not easily captured by a single measure (FAO (Food and Agriculture Organization of the United Nations) 1996; 2009; Webb et al. 2006; Barrett 2010). Moreover, multiple levels of the food security system and time dimensions make the task of measuring food security more complex (Ecker and Breisinger 2012).

In this paper, we primarily measure food insecurity at the household level using the amount of calories acquired by each household relative to its needs. Two main sources of calorie acquisition are considered: household own food production retained and household food purchases. Based on these two main sources, a food insecurity indicator is constructed and the percentage of households that fail to meet a minimum threshold level of calorie consumption are considered food insecure. This measure is calculated by household quintile and geographic region of rural Mozambique. Thereafter, changes in the percentage of food insecure households are tracked over time using the available household survey data.

This approach of measuring food insecurity utilizes key insights from Smith and Subandoro (2007) and Cafiero (2012) and certainly has its limitations largely due to the nature of available data. Calorie acquisition (or deprivation) indicators have been widely used to measure household food insecurity yet this class of measures has also received criticism for two major reasons. First, it is not possible to uncover intra-household allocations of food when using household-level calorie-based indicators (Hoddinott 1999). Second, by focusing on calories or dietary energy requirements, proteins and micronutrient deprivation is ignored, yet these are critically important (Barrett 2002). Nevertheless, in a country where 44 percent of the population suffers from chronic malnutrition (defined as low height for age or stunting) and about a third faces chronic hunger (Grupo de Estudo, 2009; WFP (United Nations World Food Program) 2010), the lack of calories is still an important component of food security to track and can be a useful assessment indicator. Moreover, data limitations do not allow for tracking past micronutrient and protein intake trends. Therefore the use of calorie acquisition is justified and provides policy-relevant information.5

In creating the food insecurity measure, rural households in Mozambique were defined as food insecure if they were unable to obtain their entire dietary calorie requirement from their own farm production and/or food purchases. While households could have acquired calories from food aid or food gifts consumed to meet their dietary calorie requirements, this paper focuses on the first two primary food sources. This is because food aid is, by design, only made available as an emergency when households are unable to meet their food requirements through normal sources. Also, food gifts tend to be given in similar circumstances as food aid and often constitute an irregular and small source of total food consumption (Silva 2008, MPD (Mozambique Ministry of Planning and Development) 2010). Households may also have served some of their food to guests. This would imply that the food consumed strictly by the household members would be less than the food acquired. We assume that, on average, such incidences of food served to guests may partially offset food gifts.

Equation 1 describes the food insecurity indicator that was constructed:
$$ foo{d}^s=\left\{\begin{array}{c}\hfill 1\ \mathrm{if}\ {\displaystyle \sum_i{\theta}_i}\left({Q}_{i r}+\frac{f{ s}_i{Y}_{cash}}{{\operatorname{P}}_{retail}}\right)< Ca{l}_{req}\hfill \\ {}\hfill 0\ \mathrm{otherwise}\hfill \end{array}\right. $$
(1)

where θi is a calorie content conversion factor for food crop i obtained from the food composition tables (Korkalo et al. 2011), Qir is the quantity of food crop i that is produced and retained for home consumption, and fsi is the share of household expenditure for food crop i while Ycash is the total household cash income, and Pretail is the retail price of maize. To be precise, the retail price that ought to be used is that of each food crop i purchased. However, due to limited coverage of crop prices in the price data available we used the retail price of maize. Calreq is the household specific minimum calorie requirement, which is calculated based on household characteristics (age and sex of each household member as well as household size)6 closely mirroring the methodology of the Food and Agriculture Organization of the United Nations (FAO (Food and Agriculture Organization of the United Nations) 2008). The household is food insecure if foods = 1, i.e., it does not meet the minimum caloric requirements, and food secure, otherwise.

It is important to note that this computation uses expenditure shares that vary by province and from year to year, which captures important dynamics in the food markets. However, as the food expenditure shares are obtained from cross-sectional household-level data, they result in an average annual measure of food insecurity that does not capture within-year seasonality of food insecurity, which would have otherwise been captured by variations in food expenditure shares from harvest to lean season within the same year. This is an important consideration if measuring temporal acute food insecurity but is not the focus of this analysis.

It is also worth noting that the measure of food security adopted here is binary and not continuous. This is purposely the case because physiologically there is a threshold minimum quantity of calories needed for household members to sustain basic functional lives. It is also recognized that the food security indicator does not directly include food purchases. Instead, it estimates them using cash income data from the TIA, Trabalho de Inquérito Agrícola surveys, food budget share estimates obtained from the IOF, Inquérito sobre Orçamento Familiar and IAF Inquérito aos Agregados Familiares sobre Orçamento Familiar databases and maize retail prices from SIMA (Sistema de Informação de Mercados Agricolas). We use this approach due to significant data limitations in that no single survey in Mozambique collects the necessary variables needed to calculate and measure food insecurity in rural areas. Thus food purchases are computed then converted into calorie values using food budget shares from the IOF/IAF data and SIMA price data before being added to the calories computed from the household’s own food production retained for consumption from the TIA data. Because we use maize prices as a means of converting expenditure shares to calories, this is likely to be a source of concern and may lead to underestimating the level of food insecurity of households that purchased more expensive sources of calories (such as meats, fruits and vegetables). Nevertheless, this is not of major concern given the limited extent to which poor rural households in Mozambique purchased more expensive calorie sources, even as real incomes increased between 2002 and 2008.7

We also estimated the cost of eliminating the food insecurity by calculating the total amount of cash that each household would need to eliminate their calorie deficit through purchases of imported maize, then summed this across households. Equation 2 shows the formula for the cost of eliminating food insecurity through imported maize purchases, at the household level.
$$ CalorieDeficitCost=\left\{\begin{array}{c}\hfill {P}_{retail} Exch\left\{ Ca{l}_{req}-{\displaystyle \sum_i{\theta}_i}\left({Q}_{i r}+\frac{f{ s}_i{Y}_{cash}}{P_{retail}}\right)\right\} if\ foo{d}^s=1\hfill \\ {}\hfill 0\ \mathrm{Otherwise}\hfill \end{array}\right. $$
(2)

Where Exch is the prevailing real exchange rate between the US dollar and the Mozambican Metical. If a household is food secure, this measure is equal to zero as that household would not need cash to eliminate food insecurity. It is recognized that food insecurity could be eliminated by a variety of calorie sources other than imported maize and thus equation 2 could be modified accordingly. For instance one could compute the total cost as a weighted sum of costs for different calorie sources, with weights derived from the household food group expenditure shares – that is if assuming that the share of calorie sources would remain the same if food insecure households were to receive cash to eliminate their food insecurity. Nonetheless, we assume that the government is likely to use a single source of food (such as maize) to address any food shortages, hence we focus our analysis of food insecurity elimination costs based on equation 2.

Drivers of food security: conceptual framework

Besides measuring and tracking food security trends, it is important to understand its drivers in order to inform food policy. At the household level, food security is understood to be a direct function of the food acquired and consumed, and the subsequent nutrient intake and metabolism by household members. In the context of rural Mozambique the food consumed is primarily sourced from household farm production or purchased in local markets. Therefore farm production constitutes an important driver of food security as it can directly provide food for rural households to consume as well as generate income through farm output sales, which can be used to access food from the market.

However, farm production itself is influenced by multiple factors, which can be considered to be the underlying or indirect drivers of food security. In rural Mozambique these drivers include agricultural input use, access to commodity and credit market systems (Mather 2009; Benfica et al. 2014); use of improved agricultural technologies, such as mechanization, irrigation, productivity enhancing inputs, and access to agricultural extension services (Mather et al. 2008; Benfica et al. 2014). Negative shocks such as cyclical floods and droughts can also be deemed (negative) drivers of food security (Joubert and Tyson 1996; Usman and Reason 2004). Household assets can help alleviate such shocks and enhance food security through production and consumption pathways. Some assets such as farm implements and animal draft power are factors of farm and off-farm production which affect food availability, while others such as bicycles and radios can be used to access markets and market information. Still other assets influence utilization and stability dimensions of food security e.g. water and sanitation facilities and food storage facilities.8 Assets at the household level can also provide collateral that allows households to access formal and informal credit, relaxing cash constraints which have implications for both production and consumption (e.g. the purchase of farm inputs and consumption smoothing in times of stress). Thus, it may be useful to track trends in asset ownership as it relates to food security in rural Mozambique.

Food purchases are also an important source of food consumed which influences food security. Income, food prices and access to food markets all affect food purchases and thus are considered underlying drivers of food security. The levels of these drivers vary geographically in rural Mozambique. For instance, recent evidence shows that cash income in rural Mozambique is mainly accessed through non-farm employment (Cunguara and Hanlon 2012) and households in the southern provinces have access to better non-farm employment opportunities. In addition southern provinces have better road infrastructure, which facilitates better access to food markets. Households in the southern provinces also have relatively higher inflows of remittances, which can significantly contribute to total household income and be used for food purchases, particularly as a coping strategy during shocks or as investments into income generating activities (Walker et al. 2004). Furthermore, better education levels prevail in the southern provinces (Government of Mozambique 2006) increasing the likelihood of more remunerative rural non-farm employment in this region. Meanwhile, households in the central and Northern provinces predominantly rely on agricultural production and marketing to obtain cash income partly because of greater agricultural potential and lower education levels. Thus disparities and spatial variation in geophysical, social, natural, and human capital are likely to indirectly affect food security levels as well (Heltberg and Tarp 2002; Donovan and Tostão 2010).

Relative food prices, volatility and seasonality are important determinants of food purchases. High food prices can make it difficult for households to purchase food and acquire calories. In the case of households that derive most of their income from non-farm employment, the level of food prices can be particularly important. Food price levels are also important for households that farm food crops. For those that are net sellers of food crops a higher price may translate into higher cash income, potentially implying improved food security while for households that produce some food crops, but not enough to meet their calorie requirements, higher food prices can mean relatively higher food purchasing costs and lower food security. Similarly, high volatility in food prices can present difficulties in stabilizing household food consumption, which implies unstable intake of calories. In addition, seasonality in food prices, which is associated with the farm production cycle, can imply seasonal food insecurity for some households. This is especially the case for rural farm households that are net buyers of food, whose market participation is seasonal and can be disadvantageous.9

The degree of household market participation and market integration with domestic, regional and global markets can also be important underlying determinants of food security. Closely associated with these are government policies such as price stabilization policies, input and trade policies which can make it easier (or harder) for households to buy farm inputs and/or food for consumption leading to improved (or decreased) food security.

Delivery of public services and investments in agriculture, transport and telecommunications infrastructure are also important underlying drivers of food security that operate through multiple channels. Remoteness and poor connectivity may mean less market participation and exacerbated impacts of localized food price shocks on household food security.10 Moreover, poor infrastructure reduces incentives for future farm production and marketing due to high transaction costs and may constrain farmers’ participation in remunerative regional and global export markets. Important to note is that government policies and public investments can affect the underlying drivers as well as food security itself directly. Clearly changes in a variety of these drivers and policies are important to analyze when assessing trends in food security. This paper assesses the trends of some of these drivers depending on data availability.

Data

The data used in this analysis are drawn from multiple rounds of the available national agricultural surveys in Mozambique, commonly known as TIA, which is the Portuguese acronym for Trabalho do Inquérito Agrícola.11 The surveys were conducted by the Department of Statistics within the Directorate of Economics of the Ministry of Agriculture of Mozambique. The TIA samples are rural and stratified by province and agro-ecological zone, making them representative at both levels. Therefore, all analyses and results presented in this paper are population-weighted to account for the stratified sampling. Sample size varied between 3,891 households in 1996 and 6,248 households in 2008. The surveys were generally similar, but differed slightly in terms of some of the questions asked over the years. For example, data on the use of improved seeds and access to credit were only collected starting in 2005 (TIA05). Also, specification of the crops that received fertilizer application was only recorded in the 2008 round (TIA08).

It is important to note that the TIA96 had methodological drawbacks, particularly with respect to the data of cassava production. Data were collected using a single recall questionnaire, which is inadequate for a crop that is harvested several times a year. Therefore, TIA96 data were used sparingly in this study. TIA02, TIA05, and TIA08 were the three most comprehensive surveys utilized. They combined the annual household demographic and agricultural and livestock production components with detailed data on major household income components. The income components included five main sources of cash income, namely livestock sales, remittances and pensions, wage income, non-farm self-employment earnings, and crop production sales. Thus, cash income as defined in this study refers to all cash received by the household.

For consistency and comparability purposes, all income data were inflated to 2008 prices. The inflators used to adjust the 2002 and 2005 income levels were constructed using the method described in Mather et al. (2008). The Instituto Nacional de Estatística consumer price index (CPI) data were not used because they reflect urban food prices, based on the urban consumption basket, whereas the current study focused on rural households. Hence, household food consumption quantities, defined in terms of the region-specific food basket derived from the consumption expenditure surveys (IAF), were used to adjust incomes. These consumption quantities were valued using 2002 Sistema de Informaçao de Mercados Agricolas (SIMA) retail prices, then the basket was revalued with 2005 and 2008 SIMA prices to update the cost of an identical (fixed) consumption basket. The consumption quantities were therefore weights for the commodity prices. Thus the inflators were fixed because the quantity-based weights were not allowed to change over time. It is important to note that the SIMA price data have wide geographic coverage and capture changes in food prices over time and space relatively well.

Trends in food security and the underlying drivers

Food security trends

To assess progress on food security and its drivers, summary statistics and trends on the food insecurity indicator and the various drivers were analyzed. Food insecurity in rural Mozambique, as measured by equation 1 was lowest in 2002, with about 43 percent of rural households being food insecure, but has been increasing ever since (Table 2). In 2008 about 48 percent of households were food insecure. These results are consistent with the official poverty incidence estimates in rural Mozambique, which show an increase in rural households in poverty from 55 percent in 2002/2003 to 57 percent in 2008 (MPD (Mozambique Ministry of Planning and Development) 2010).
Table 2

Percentage of food insecure households by year

Quintiles of total maize production

Percentage of food insecure households (foods) - Equation 1

Cost of eliminating food insecurity per capita/day in 2008 US$ (CalorieDeficitCost) - Equation 2

2002

2005

2008

2002

2005

2008

Bottom quintile

55.81

61.09

60.38

0.10

0.11

0.11

2

52.96

56.55

59.63

0.08

0.09

0.09

Middle quintile

46.43

48.47

51.35

0.07

0.09

0.08

4

37.33

42.24

41.32

0.05

0.07

0.05

Top quintile

12.18

14.43

16.39

0.04

0.05

0.04

Total

42.93

45.44

47.79

0.08

0.09

0.08

Authors’ calculations based on TIA02, TIA05 and TIA08

Households have been ranked in quintiles by the amount of maize produced. Also, it is assumed that all household food insecurity would be eliminated through purchases of imported maize

The results also show that an overwhelming majority of rural households in the bottom three quintiles (over 50 percent) were food insecure. As expected, food insecurity decreases as one moves from the bottom quintile to the upper quintile, but the change is noticeably greatest between the fourth and the top quintile. This is correlated with the median per capita cash income distribution (see Table 6), which shows the top quintile having about twice as high a median per capita income as households in the fourth quintile. Similarly, maize production by the top quintile is about three times higher than that of farm households in the fourth quintile (Table 4). Therefore, food security strategies are likely to differ significantly between households in the top quintile and those in the bottom four quintiles.

Results in Table 2 also reveal that the cost of eliminating food insecurity, as measured by equation 2, slightly increased across the board between 2002 and 2005, probably as a result of the 2005 drought. However, between 2005 and 2008, the cost of eliminating food insecurity for those households in the top three quintiles decreased marginally while it remained the same for the bottom two quintiles. This points to disparities in the incidence of food insecurity and suggests that bottom quintile households are less likely to recover from drought shocks or improve their food security over time. Thus they may need targeted assistance over time. Based on these estimates, we calculated that each Mozambican in rural areas would need, on average, $0.08 per day to eliminate food insecurity. This might not seem to be much at first glance, but is actually greater than the median per capita cash income per day in 2008 (See Table 6: $25.38/365 days = $0.07), and is also roughly a quarter to a fifth of the official poverty line, depending on the region in the country. Thus, while the Food Security and Nutrition National Strategy described in PARPA II aims at increasing food availability from farmers’ own production and improving farmers’ ability to purchase food, these results indicate that this did not happen between 2002 and 2008. On the contrary, food insecurity increased despite rapid macroeconomic growth in the same period. This finding is consistent with results obtained from other welfare measures for Mozambique (MPD (Mozambique Ministry of Planning and Development) 2010; Arndt et al. 2012; Alfani et al. 2012).

To understand the causes of the observed food insecurity trends, we analyzed trends in the drivers of food security. The following section presents these trends in the drivers and discusses them in relation to the conceptual framework as well as government policy and investments.

Drivers of own food consumption

Agricultural production is an important driver of own food consumption and food security in rural Mozambique, which in turn is influenced by a variety of factors; these include land area cultivated, farm inputs use, technology adoption and receipt of agricultural services. Descriptive analysis of the TIA data shows that while maize production has increased by about 20 percent between 2002 and 2008 (Table 3) it is unevenly distributed and land productivity at the farm level remains relatively low. Throughout the 2002–2008 period the bottom quintile accounted for less than three percent of total maize production, while the top quintile accounted for close to two thirds of total production. This skewed distribution barely changed even as total maize production increased during that period. Also, the mean land area cultivated increased from about 1.3 ha in 2002 to 1.8 ha in 2008 per farm household. Although maize yields in 2008 averaged 0.8 tons per ha, the top quintile’s yields were estimated at about 1.9 tons per ha which is about twice as much as the fourth quintile’s yield levels (1.0 tons per ha) and almost 19 times the yields of the bottom quintile (Table 4).
Table 3

Farm production and use of farm inputs, agricultural technology and services

 

TIA02

TIA03

TIA05

TIA06

TIA07

TIA08

Maize production (1000s Metric tons)

1,115

1,181

942

1,396

1,134

1,214

Share of maize production (%) – Upper quintile

56.0

55.4

58.4

57.4

57.8

59.2

Share of maize production (%) – 4th quintile

21.9

22.2

21.2

23.3

22.5

19.8

Share of maize production (%) – middle quintile

13.6

13.2

12.4

11.6

12.1

12.6

Share of maize production (%) – 2nd quintile

6.2

6.9

6.1

5.9

5.8

6.2

Share of maize production (%) – Bottom quintile

2.2

2.3

1.9

1.8

1.8

2.2

Mean Cropped area (ha)

1.3

1.4

1.7

1.7

1.6

1.8

Mean maize yield (tons/ha)*

0.7

0.9

0.6

1.1

0.9

0.8

Kilocalories produced per ha

2,307

2,643

1,935

2,424

2,189

1,961

Used chemical fertilizers (%)

3.7

2.5

3.8

4.6

3.6

3.0

Used chemical pesticides (%)

6.8

5.1

5.4

5.3

6.5

2.6

Used animal traction (%)

11.2

10.9

9.3

12.4

11.5

10.9

Hired seasonal labor (%)

15.5

15.3

17.6

23.8

20.8

19.6

Received extension visits (%)

13.5

13.5

14.8

12.0

10.2

8.3

Received price information (%)

34.5

47.2

40.3

36.3

33.1

34.2

Farmers’ association (%)

3.7

4.5

6.4

6.5

8.3

7.2

Used poultry vaccine (%)

1.9

3.2

3.0

4.1

NA

4.4

Authors’ calculations based on TIAs and MPD (Mozambique Ministry of Planning and Development) (2010)

aNote, the calculation of maize yields is based on dividing maize production by land area cultivated for maize, without estimating land area equivalent ratios for plots that are intercropped

NA, Not available

Table 4

Farm production and inputs use by quintile of total maize production in 2008

 

TIA08: Quintile of total maize productiona

 

Bottom quintile

2

Middle quintile

4

Top quintile

Total

Maize production (1000s Metric tons)

26

75

153

240

718

1,214

Mean cropped area (ha)

1.3

1.5

1.6

2.1

3.2

1.8

Maize yield (tons per ha)

0.1

0.4

0.7

1.0

1.9

0.8

Used improved maize seeds (%)

8.2

8.2

9.3

11.4

12.7

9.7

Used chemical fertilizers (%)

0.8

1.4

2.6

7.4

10.3

4.0

Used animal traction (%)

15.4

16.6

13.0

14.2

19.9

15.6

Hired seasonal labor (%)

11.5

17.8

17.9

25.8

34.5

20.4

Received extension visits (%)

4.9

6.5

9.0

9.5

12.2

8.1

Received price information (%)

28.9

34.3

36.6

39.0

44.2

35.9

Authors’ calculations based on TIA 2008

aAnalysis restricted to maize growers, 78 % percent of the total sample; farms have been ranked in quintiles by the total amount of maize produced

It is also the case that the households in the upper quintile achieved higher production levels mostly because they cultivated larger land areas. Thus while achieving higher yields than the bottom quintile households, their larger cropped area contributed to higher production. As shown in Table 4, total maize production by the top quintile was approximately 28 times greater than that of the bottom quintile in 2008 and mean cropped area of maize was almost three times greater.

Use of improved farm inputs among the top quintile households was also higher, and is likely to have contributed to the higher maize yields obtained by the top quintile households. In 2008, households in the top quintile used 13 times more fertilizers than their bottom quintile counterparts. In addition, over 50 percent more households in the top quintile used improved seed than the bottom two quintiles. Also, more farm households in the top quintile reported having received agricultural extension services (12.2 % compared to only 4.4 % among the bottom quintile households).

In general, receipt of agricultural extension services is low in rural Mozambique; between 2005 and 2008, the proportion of rural households receiving extension services steadily declined from 14.8 to 8.3 %. According to the Ministry of Agriculture, part of this decline can be attributed to disruption in fuel supply for extension workers’ motorcycles. Similarly, the market information system recently experienced a decline in public funding available to pay for radio broadcasts (Mather 2009), which probably resulted in the observed decline in percentage of households receiving price information in recent years. These downward trends suggest that the government did not provide adequate public resources for agricultural extension and market information services support, both likely to have contributed to increased food insecurity.

Statistics on the use of animal traction also showed a gradual decline between 2006 and 2008. Currently only two percent of farmers use tractors in rural Mozambique, mostly in lower agricultural potential areas. Due to the high prevalence of trypanosomosis disease in the Northern parts of the country, animal traction has generally been limited to the southern provinces and a few villages in central provinces. Tick-borne diseases also limit the livestock population and animal traction in the rest of the country. Little effort has been made to address trypanosomosis and tick-borne diseases to enable expanded use of animal traction in areas of high agro-ecological potential (Pingali et al. 1987; Mather 2009). One way of addressing trypanosomosis could be to invest in the production and release of irradiated-sterile male tsetse fly combined with serological surveillance and drug treatment of livestock. To address tick-borne diseases, the use of dipping services can be considered but dipping services are sparse in Mozambique as many dip tanks were destroyed during the war prior to 1992 and have never been rehabilitated. In addition, there is a shortage of dipping tank attendants and a lack of potable water (Alfredo et al. 2005). Additional options to address tick-borne diseases include the application of vaccines in endemically unstable conditions, and the use of tick-resistant breeds of cattle (Norval et al. 1983; Latif and Pegram 1992). Other alternatives are the adoption of conservation farming techniques that entail reduced or no tilling, as well as mechanization for land tilling (e.g. tractors, small-scale power tillers, etc.).

While the adoption of improved agricultural technologies and the use of modern farm inputs are likely to increase food production and hence food security for the bottom quintile, they are unlikely to be sufficient to close the existing food security gap. Our calculations indicate that a family of five, the average household size in rural Mozambique, would have to adopt a package of improved technologies that increases their household incomes by about $241 per year, almost 10 times the median cash income in 2008, which seems quite difficult and highly unlikely. In addition to adopting improved technologies, bottom-quintile farmers would have to expand their cropped area as achieving the productivity levels of the top quintile just on their current land area would not be enough. This suggests that they may have to either expand land area or seek employment outside farming if they are to rapidly achieve food security. In contrast, calculations suggest that food-insecure households in the top and fourth quintiles have the potential to significantly improve their food security by adopting improved agricultural technologies and increasing use of modern farm inputs. This analysis finds that these households in the top and fourth quintiles currently lie just below the ‘food insecurity line,’ and use less farm inputs compared to their food secure counterparts.

Market-based drivers of food purchases

The type of crops as well as the amount sold affect farm cash income, which in turn influences the ability of households to purchase food. The type of crops and amount sold varied significantly by household income level. The poorest farmers, defined as those in the lowest quintile of per capita cash income, tended to sell staple crops such as maize, whereas relatively wealthier households sold cash crops in addition to staple crops (Table 5). Note, however, that fewer households in the top quintile sold cotton (a traditional cash crop in Mozambique), except when compared to the bottom quintile. This is consistent with previous findings (Pitoro et al. 2009; Benfica 2012), which showed that some non-cotton growers were wealthier than cotton growers. In addition to selling a wider variety of cash crops, the top quintile farmers sold them in greater quantities, probably due to the larger scale of their farms and were thus likely to be more resilient to shocks that can cause food insecurity. Note that when bottom-quintile households were able to sell surplus crops, they typically sold extremely low quantities implying that their level of market participation does little to improve cash income and in turn food security. Farmers in the bottom quintile also rarely cultivated tobacco or sesame, which are major cash crops with higher gross margins in Mozambique.
Table 5

Crop output market participation by quintile of per capita cash income and region

Quintile/region

% of households who sold the following cropsa

Mean quantity sold in kilograms per household

Maize

Cotton

Tobacco

Sesame

Tomato

Maize

Cotton

Tobacco

Sesame

Tomato

Bottom

5.0

0.5

0

0.4

1.1

59.9

42.5

0

18.7

197.2

2

17.6

4.9

0.4

5.4

2.6

133.0

213.1

44.7

51.2

309.2

Mid

20.4

6.5

2.2

7.4

4.5

192.6

448.8

198.7

83.3

520.2

4

19.6

4.5

4.2

9.0

5.9

301.4

658.3

369.6

112.7

1,471.0

Top

18.9

2.4

4.9

7.7

5.2

952.2

1,041.2

986.0

407.4

2,088.3

North

20.5

5.0

2.2

6.1

2.9

228.5

416.9

409.4

132.0

585.1

Central

14.1

3.5

3.7

9.5

6.5

802.4

730.2

745.7

197.5

1,806.0

South

4.9

0.2

0.1

0

3.1

182.8

33.9

1,412.3

0

827.2

Total

16.2

3.8

2.2

5.9

3.8

346.8

484.1

555.2

157.6

1,129.2

 

Position in food crop markets in 2008 (%)

 
   

Autarky (neither buy nor sell)

Buy only (Net buyer)

Buy and sell (Net buyer)

Sell only (Net seller)

Sell and Buy (net seller)

   

North

  

34.4

48.7

2.9

9.2

4.7

   

Central

  

22.7

57.2

3.2

12.2

4.7

   

South

  

17.5

77.6

0.7

1.6

2.5

   

Total

  

26.1

57.6

2.7

9.3

4.3

   

Authors’ calculations based on TIA08

Households have been ranked by per capita household cash income to generate quintiles

There are also regional differences in output market participation in terms of the types of crops sold, percentage of households selling and quantities sold. Higher percentages of households sold maize and cotton in the Northern provinces whereas in the central provinces the percentage of households selling tobacco, sesame, and tomatoes12 were predominant. Farmers in the central provinces had the largest volumes of sales on average, including those of maize and cotton. While statistics show the southern provinces as selling the largest quantities of tobacco, the figure reported in Table 5 can be misleading. A closer look at the data shows that this quantity of tobacco sales is accounted for by only three large-scale sellers (i.e. 0.1 percent of all households in the South), and all of them were located in Maputo province. In general there were relatively lower quantities of crops sold by farmers in the Southern provinces, likely due to the lower agro-ecological potential there. Despite high and comparable agro-ecological potentials between the Northern and Central provinces, crop sales were significantly lower in the Northern provinces. This could be associated with poor road infrastructure and low market access in the Northern provinces. Indeed, the development of rural markets is identified by the government as one of the main interventions needed to promote rural development and food security (Government of Mozambique 2006: p70).

Further investigation of output-market participation showed that most rural households in Mozambique were buyers of food and did not sell any food (approximately 58 percent for maize in 2008). This was especially the case among farmers in the southern provinces – over 77 % were buyers of food and did not sell any farm output (Table 6). Also, as revealed in the TIA 08 data, only a small fraction (4 %) sold a net surplus of food to the market while a significant proportion (26 %) was autarkic in 2008 suggesting a significant reliance on own farm production for consumption and therefore food security. For those households that buy and sell food, they usually sell their crops during and/or right after the harvesting season at prices that are significantly lower than the food prices they eventually pay when buying food during the dry season (Stephens and Barrett 2008). Therefore, profitable participation in the output market is important to assess and this largely hinges on a household’s ability to store food or access post-harvest marketing services as well as cost-effective participation in input markets that drives the ability to produce a surplus. For example, Howard et al. (2003) showed that, in Mozambique, farmers using improved maize seeds earned 24 percent more than non-users if they were able to delay their maize sales until November as opposed to selling in September-October, soon after harvest.
Table 6

Mean and median per capita cash income

Quintile of total maize production

Median per capita cash income (in 2008 constant US$)

Mean per capita cash income (in 2008 constant US$)

2002

2005

2008

2002

2005

2008

Bottom quintile

9.61

22.41

21.15

72.85

94.27

100.22

2nd quintile

10.60

26.19

18.94

66.23

113.40

75.05

Middle quintile

13.18

28.30

21.67

81.61

104.18

88.78

4th quintile

22.88

36.09

31.37

99.82

118.28

90.12

Top quintile

41.48

86.87

58.46

152.01

200.16

174.50

Total

16.30

34.44

25.38

90.89

123.47

102.71

Cunguara and Hanlon (2012)

Regarding trends in cash income, both the mean and median per capita cash income increased significantly between 2002 and 2005, which is in line with previous assessments (Mather et al. 2008; Cunguara and Hanlon 2012). However the incomes decreased in the following periods and results show a large variation in cash income across time and quintiles of maize production (Table 6).

The distribution of per capita cash income among rural households was also positively skewed, with more households earning significantly less than the average cash income as depicted by the comparison of the median and mean incomes in Table 6. Even though cash incomes increased between 2002 and 2008, they remain relatively low. The importance of cash income in attaining household food security through food purchases is tied to relative food prices and the ability to access food markets. Hence, we also analyzed food prices by region.

Figure 3 shows the trends in regional real maize prices. In general prices are higher in the Southern region, which is constantly a food deficit area. All regions’ prices increased significantly in 2002 and in 2008, with sharper increases recorded in 2002 (a drought year). This was especially the case in the Northern provinces where the price levels reached the same high levels of the South due to the drought that was mostly located in the North. In 2008, the year of the global food price crisis, the sharpest price increases occurred in the Central and southern Provinces although these prices remained lower than the Southern prices. These trends imply increased food insecurity when taken together with the evidence of declining incomes between 2005 and 2008 shown in Table 6. Also, price increases were geographically distinct with the Southern provinces experiencing price increases in 2005 largely due to localized drought that year while other regions experienced declining prices that year. Prices in the Northern region generally declined, although spikes were recorded for 2002 and 2008 with the greatest percentage increase being in 2002. Given that, on the whole, food prices increased significantly and household incomes decreased between 2005 and 2008, households, particularly those in the Central provinces, were likely to have suffered severe food insecurity in 2008. Those in the Southern provinces that largely depend on food purchases also likely experienced significant food insecurity as a result of the food price increases in 2008.
Fig. 3

Real maize prices by region (CPI-deflated Meticals per kg – base year 2008)

These results stress the importance of regionalized policy strategies and investments that enhance resilience to food price shocks and weather shocks. Investments in road infrastructure and market information, which are necessary (though not sufficient) conditions for increasing crop market participation of Mozambique’s smallholder farmers are an important strategy the government could use to facilitate regional food trade that helps dampen the effects of localized weather shocks and price shocks (Haggblade et al. 2008). Also, road infrastructure investments have multiple benefits and have been shown to be critical in the profitability of farmers’ use of improved agricultural technologies (Cunguara and Darnhofer 2011). However, market infrastructure development and adoption of improved agricultural technologies to enhance resilience to shocks is barely explored in PARPA. Combining such strategies with trade and input policies that enhance resilience to price and weather shocks could yield significant improvements in food security.

Rural farm households could also take advantage of new cost-effective information and communications technologies given the rapid and widespread prevalence of mobile phones in the country. There is an opportunity to leverage the existing private sector investments in telecommunications to upgrade the market information system and provide innovative market information services to farmers and agricultural traders. However, as depicted in the analysis, farmers’ receipt of market information has declined with the decline in public investments in recent years.

Overall the trends analysis shows that food insecurity has increased in rural Mozambique and that the limited progress on several drivers of food insecurity are likely explanations of this. Whether the indicator is per capita cash income, farm production, the receipt of agricultural services (e.g., agricultural extension services and price information), or the use of improved agricultural technologies such as chemical fertilizers and animal traction, limited progress has been made. This suggests that policy efforts, implemented by the government, were not adequate to increase food security. Nevertheless, it is possible that they may have mitigated the extent of food insecurity particularly attributable to the increases in food price in 2008.

Econometric analysis

Econometric analyses were performed using reduced-form Probit models that regressed the probability of being food insecure (i.e. the binary food insecurity indicator, foods) on the underlying driver variables. The class of Probit models that were estimated can be expressed as follows:
$$ \mathrm{Prob}\ \left( foo{d}^s=1\left| X\right.\right)=\phi \left({X}^{\prime}\beta \right) $$
(3)
where ϕ is the standard normal cumulative distribution function, X is a vector of the underlying drivers of food security and β is a vector of parameters to be estimated by maximum likelihood. Details of specific variables comprising the vector X of food security drivers are based on the conceptual framework and can be partitioned into the two components of equation 1, the first which pertains to own food production and the second pertaining to the ability to purchase food from the market. Two additional components of food insecurity, which are mostly outside the scope of the current analysis are food utilization and stability. The former is minimally incorporated into the analysis through the calorie content conversion factor (θi), while the latter is partially analyzed by looking into the dynamics of food insecurity using pooled inter-temporal data as well as incorporating wealth variables such as livestock ownership, which may be useful in buffering food security shocks over time; otherwise the analysis does not attempt to address these components. The specific variables included in the probit analysis are described in Table 7 alongside the corresponding data sources.
Table 7

Description of variables and data sources

Variable Label

Description

Data source

Foods

Food insecurity indicator (Binary: 1 = Food insecure, 0 = Food secure)

Computed using data from multiple surveys (TIA, IAF/IOF, SIMA data)

θi

Calorie content conversion factor

Food composition tables (Korkalo et al. 2011)

Qr

Quantity of farm production produced and retained for own consumption

TIAs

fs

Share of food expenditure

IAF02 and IOF08

Ycash

Total household cash income

Computed using TIAs and FAO minimum calorie requirement

Pretail

Food retail prices

SIMA

Calreq

Household minimum calorie requirement

Computed using TIAs

Gender

Head’s gender (1 = male, 0 = female)

TIAs

Educ

Head’s years of completed education (years completed)

TIAs

Age

Household head’s age (years completed)

TIAs

AE

HH size in adult equivalent scale (Adult Equivalent ratio)

TIAs

Salaried

HH Head is engaged in salaried activity (1 = yes, 0 = No)

TIAs

Self

HH Head is self-employed (1 = Yes, 0 = No)

TIAs

Area

Cropped area (ha)

TIAs

Cattle

Cattle herd size (count)

TIAs

Goat

Number of goats owned by the household (count)

TIAs

Chicken

Number of chickens owned by the household (count)

TIAs

Seeds

HH used improved maize seeds (1 = Yes, 0 = No)

TIAs

Traction

HH used animal traction (1 = Yes, 0 = No)

TIAs

Fert

HH used fertilizers (1 = Yes, 0 = No)

TIAs

LaborP

HH hired permanent labour (1 = Yes, 0 = No)

TIAs

Labors

HH hired seasonal labour (1 = Yes, 0 = No)

TIAs

Extension

HH received extension services (1 = Yes, 0 = No)

TIAs

PriceInfo

HH received price information (1 = Yes, 0 = No)

TIAs

Assoc

Member of a farmers’ association (1 = Yes, 0 = No)

TIAs

Credit

HH received credit (1 = Yes, 0 = No)

TIAs

HH, Household

In general, two types of reduced-form Probit models were estimated, one where the data for TIA02, TIA05 and TIA08 were pooled, thus an estimation of a pooled Probit model that included a time trend variable with the TIA02 set as the baseline, and second where only the TIA08 data were used. The former was estimated to capture the dynamics of food insecurity and its drivers. While this did not use panel data, the TIA02 and TIA05 included a subsample of the same respondents interviewed in both survey rounds, providing credence to our approach of capturing dynamics of food insecurity. The latter Probit model using the TIA08 was estimated to investigate the effects of a broader set of underlying drivers of food security, given that the TIA08 collected data were on a more comprehensive set of variables. However, because this model is cross-sectional, the effects of changes over time were not captured in this model. All types of models were first estimated using quintile Probit regressions but final results that are presented here focus on the estimation using the bottom four quintiles combined as results from the quintile regressions for these bottom-four quintiles did not differ significantly.

Several household demographic characteristics were included as control variables on the right-hand side of the Probit models. Education level of the household head has been shown to influence cash income earned by the household (Reardon 1997; Garrett and Ruel 1999). Thus the coefficient on education was expected to be negative, implying that more educated households would have higher incomes and likely be more food secure. A variable on sex of household head was included because poverty studies in Mozambique and elsewhere in developing countries have shown that female-headed households tend to be worse-off than their male counterparts (Walker et al. 2004; Boughton et al. 2006; Boughton et al. 2007; Mather et al. 2008). Household size, expressed in terms of the adult equivalent scale (Deaton 1997), was also included as a proxy for labor availability, both for agricultural and non-agricultural activities. Cunguara et al. (2011) showed that increasing the household size usually results in lower economic outcomes because the marginal gain in net income per capita is smaller than the average net income per capita. Thus, household size was also included in the analysis and its coefficient was expected to be positive.

A second set of independent variables that were included in the models consists of participation in non-farm activities by the household head. Here two proxies were used: whether the household head was engaged in salaried or self-employment activities. Previous work has revealed that household head’s participation in non-farm activities increases farmers’ ability to purchase food (Garrett and Ruel 1999; Babatunde et al. 2007). Therefore, the coefficient was expected to be negative as households with non-farm sources of income are less likely to be food insecure.

A third set of independent variables consisted of indicators of household asset ownership. Here, two proxies were used: cropped area and livestock (cattle, goat/sheep and chicken) ownership. Households cultivating larger fields were expected to be more food secure (Tschirley and Weber 1994), thus the sign of the coefficient would be expected to be negative while its magnitude would most likely vary by region, reflecting differences in agricultural potential. A squared term was included for the variable on cropped area to capture potential diminishing marginal returns from land. In terms of livestock, cattle are relatively predominant in the southern provinces, goats are found more frequently in the central provinces, particularly in Tete, and chickens are widespread throughout the country. The coefficient on each of these three variables was expected to be negative, implying that households can sell off some of their animals to purchase food, invest in agricultural activities or invest in small-businesses hence increasing their cash incomes and food security (Reardon and Taylor 1996; Dercon 1998; Benfica 1998).

With regard to the use of agricultural technology, animal traction was included as an independent variable in the models for the southern and central regions, but excluded from the Northern provinces model because of very low prevalence of animal traction in the data probably due to widespread incidence of tsetse fly in the region, a vector of trypanosomosis. Mather (2009) estimates that the use of animal traction increases crop income by as much as 33 percent in the central provinces. There the gains from animal traction are related to increases in both agricultural productivity and expansion of cropped area, whereas in the southern provinces its impact is only related to expansion of crop area (Mather 2009); the coefficient was thus expected to be negative. The use of improved seeds and chemical fertilizers was also expected to have a negative sign on the food insecurity indicator, suggesting that households adopting these technologies would be less likely to suffer from food insecurity. A variable on the use of hired labor was included to capture the potential heterogeneous effects of family versus hired labor (Deolalikar and Vijverberg 1987).

The last set of independent variables in the model estimations pertain to access to services, including the receipt of credit, extension services, price information, and household membership of farmers’ associations. Generally, these variables were expected to have a negative sign, implying that access increases either agricultural production or market participation thus improving food security. In addition, district dummy variables were included to control for spatial fixed effects in agricultural potential, access to non-farm opportunities, and other location specific factors.

Econometric results

The following econometric results are primarily based on the analysis of the TIA08 because the data provided more detail on key agricultural variables, which were not collected in previous rounds of the TIA surveys (e.g., access to credit was not collected in TIA02 and TIA05; the use of improved seeds was not collected in TIA02). However, in order to gain insights into the dynamics of food insecurity in rural Mozambique with respect to the drivers whose variables were collected in the previous TIA rounds, results of pooled Probit model estimations using the three TIA surveys (TIA02, TIA05 and TIA08) are also presented. In all of the econometric analyses it is important to highlight that we present correlates of food insecurity and cannot claim direct causality. As such, cautious interpretation of the results is necessary and should only be confined to correlations. Difficulties in finding suitable instruments (and achieving the identification restrictions, given the number of potentially endogenous explanatory variables, meant that we could not address the endogeneity problem by estimating instrumental variables probit models; therefore it is inappropriate to infer causality from the analysis presented here.

Tables 8 and 9 present results from the Probit estimation for TIA08. The model fits the data relatively well, especially among the top quintile households, with a pseudo R2 of 0.80, and about 91 percent of households’ food insecurity status predicted correctly. Most of the signs of the estimated coefficients are consistent with the literature.
Table 8

Probit model results for the bottom four quintile households, by region in 2008

Dependent variable = Prob (foods)

North: bottom four quintiles

Center: bottom four quintiles

South: bottom four quintiles

Coeff.

Sig.

Mean

Coeff.

Sig.

Mean

Coeff.

Sig.

Mean

Head’s gender (1 = male)

−0.32

b

0.80

−0.18

 

0.75

−0.05

 

0.61

Head’s years of completed education

−0.07

c

2.74

−0.05

b

3.05

−0.10

c

2.79

Head’s age (years completed)

0.00

 

40.03

0.02

 

43.10

−0.01

 

48.49

Head’s age (squared term)

0.00

 

1786

0.00

 

2072

0.00

 

2589

HH size in adult equivalent scale (AE)

1.31

c

2.84

0.42

c

3.27

0.23

c

3.57

HH size in AE (squared term)

−0.12

c

9.02

−0.03

 

12.69

−0.01

b

16.72

Head is engaged in salaried act. (1 = yes)

−0.22

b

0.24

−0.39

c

0.36

−0.51

c

0.34

Head is self-employed

−0.37

c

0.42

−0.44

c

0.36

−0.63

c

0.29

Cropped area in hectares

−0.45

c

1.61

−0.25

b

1.94

0.08

 

1.50

Cropped area in hectares (squared term)

0.03

c

3.74

0.04

c

5.80

−0.01

 

3.96

Cattle herd size

−0.03

 

0.03

−0.04

a

0.51

0.00

 

0.77

Number of goats owned by the HH

0.00

 

0.63

−0.01

 

2.37

0.00

 

1.32

Number of chickens owned by the HH

−0.03

c

3.09

−0.01

b

6.89

−0.02

b

5.35

HH used improved maize seeds (1 = yes)

0.13

 

0.05

−0.02

 

0.18

−0.01

 

0.09

HH used animal traction (1 = yes)

  

0.00

−0.02

 

0.11

−0.22

a

0.44

HH used fertilisers (1 = yes)

0.11

 

0.02

−0.08

 

0.04

−0.14

 

0.02

HH hired permanent labour (1 = yes)

  

0.01

−0.57

 

0.04

−0.39

 

0.03

HH hired seasonal labour (1 = yes)

−0.47

c

0.18

−0.59

c

0.17

−0.53

c

0.20

HH received extension services (1 = yes)

−0.05

 

0.08

−0.02

 

0.10

−0.01

 

0.05

HH received price information

−0.12

 

0.34

−0.08

 

0.39

−0.16

 

0.32

Member of a farmers’ association (1 = yes)

0.10

 

0.08

−0.06

 

0.05

0.25

 

0.09

HH received credit (1 = yes)

−0.49

 

0.02

−0.43

a

0.03

0.07

 

0.02

Constant

−1.11

a

 

−0.53

  

1.49

b

 

Number of observations

1581

  

1106

  

1330

  

Wald chi2

306

  

179

  

211

  

Prob > chi2

0

  

0

  

0

  

Pseudo R2

0.21

  

0.18

  

0.17

  

Percent predicted correctly

72.87

  

70.71

  

69.85

  

Authors’ calculations based on TIA08

District dummies are included but not reported to save space. c, b, and a denotes significance at 1, 5, and 10 %, respectively

Table 9

Probit model results for the top quintile households by region in 2008

 

North: top quintile

Center: top quintile

South: top quintile

 

Coeff.

Sig.

Mean

Coeff.

Sig.

Mean

Coeff.

Sig.

Mean

Head’s gender (1 = male)

1.71

b

0.90

−0.88

b

0.91

1.14

b

0.75

Head’s years of completed education

0.16

 

3.32

−0.04

 

4.01

−0.19

b

2.73

Head’s age (years completed)

0.78

c

42.27

0.04

 

42.58

0.10

 

50.52

Head’s age (squared term)

−0.01

c

1943

0.00

 

1985

0.00

 

2721

HH size in adult equivalent scale (AE)

14.84

c

3.39

1.60

c

3.73

0.46

b

5.17

HH size in AE (squared term)

−1.66

c

13.16

−0.09

c

16.16

−0.02

 

33.55

Head is engaged in salaried act. (1 = yes)

−3.55

c

0.19

−0.95

c

0.28

−0.84

a

0.28

Head is self-employed

−1.65

c

0.43

−0.92

c

0.39

−0.01

 

0.31

Cropped area in hectares

−0.73

 

3.10

−0.41

c

3.46

−0.31

b

2.80

Cropped area in hectares (squared term)

−0.05

 

13.98

0.01

c

21.48

0.01

b

12.43

Cattle herd size

0.65

c

0.11

−0.07

a

1.38

−0.02

 

3.50

Number of goats owned by the HH

−0.38

c

1.37

−0.07

c

3.67

0.01

 

3.31

Number of chickens owned by the HH

−0.06

a

6.33

−0.01

 

11.11

−0.10

c

13.53

HH used improved maize seeds (1 = yes)

−2.06

a

0.10

0.75

a

0.18

−2.28

c

0.11

HH used animal traction (1 = yes)

  

0.01

0.05

 

0.24

−1.06

b

0.66

HH used fertilisers (1 = yes)

−3.65

c

0.07

−0.57

 

0.13

  

0.05

HH hired permanent labour (1 = yes)

  

0.09

1.20

b

0.11

−0.51

 

0.09

HH hired seasonal labour (1 = yes)

  

0.32

−0.99

c

0.36

−0.89

b

0.45

HH received extension services (1 = yes)

−2.74

c

0.11

−0.84

a

0.13

1.51

c

0.08

HH received price information

−0.93

 

0.31

0.12

 

0.53

−0.54

 

0.36

Member of a farmers’ association (1 = yes)

3.34

c

0.11

0.75

a

0.09

−0.85

 

0.19

HH received credit (1 = yes)

−2.67

b

0.05

−1.28

a

0.06

  

0.02

Constant

−54.23

c

 

1.00

  

−1.24

  

Number of observations

219

  

433

  

164

  

Wald chi2

126.56

  

156.53

  

83.66

  

Prob > chi2

0

  

0

  

0

  

Pseudo R2

0.80

  

0.50

  

0.53

  

Percent predicted correctly

90.87

  

88.22

  

80.49

  

Authors’ calculations based on TIA08

District dummies are included but not reported to save space

c,b and a denotes significance at 1, 5, and 10 %, respectively

Socio-demographic control variables such as gender and education of household head as well as household size (measured in adult equivalents) were mostly significant and had the expected signs. Male-headed households were less likely to be food insecure ceteris paribus (coefficient ranging from −0.32 in the North to −0.08 in the South) as were households with more educated heads (coefficient ranging from −0.10 in the South to −0.05 in the Central province) while households with higher adult equivalents had significantly higher likelihood of being food insecure (coefficient ranging from 1.31 in the North to 0.23 in the South among the bottom four quintile households).13 The age of the household head was mostly not statistically significant among the bottom four quintile households and the magnitude was often close to zero suggesting that there were no major differences in food insecurity of households with young household heads and those with older household heads (see coefficients in Tables 8 and 11). However among the top quintile households, especially those in the North, increased age of household head seemed to significantly increase the probability of food insecurity (in Table 9 coefficient equals 0.78, and 0.13 in Table 11 both at 1 % significance level). These results suggest that in the rural Northern parts of Mozambique, age may affect the ability to produce or acquire calories and in turn household food security.

The use of improved agricultural technologies (improved maize seed varieties and chemical fertilizers) in the Northern provinces was not significant for the bottom four quintile households, but was significant for the top quintile households (compare Tables 8 and 9). A similar pattern was observed in the southern provinces. One possible explanation for this is that households in the bottom four quintiles cultivate small parcels of land, and most use intercropping systems and less modern farm inputs. Thus the average effect of adopting these technologies is likely diminished, and households in these quintiles may need to substantially expand their use of modern farm inputs as well as cropped area to realize significant gains from agricultural technology adoption. Indeed, in the Northern provinces where cultivated land areas are larger, the coefficient on cropped area is significant for the households in the bottom four quintiles. The results on the effect of animal traction are slightly different, suggesting that all households in the southern provinces would likely benefit from its adoption. Use of animal traction, however, was not statistically significant for the central provinces, probably due to less variation in the data.

Table 9 presents detailed results for the top quintile households by region for 2008 and shows that cropped area is an important factor reducing the likelihood of food insecurity in the Central and Southern provinces but not in the North. Also, hiring of labor (especially seasonal labor) in the Central and Southern provinces significantly helps lower the probability of food insecurity among the top quintile households. This is likely due to a significant contribution of land and labor in farm production and in turn own-food consumption and marketed surplus that generates income for the top quintile households. There are more markets in the Central and Southern provinces, which might explain why the effect of the cropped land area on food insecurity in the Northern provinces is not statistically significant. The results also show the importance of small livestock in reducing the likelihood of food insecurity. Goats were significant factors enhancing food security among the top quintile households (coefficients are −0.07 and −0.38 for the Central and Northern provinces respectively, as shown in Table 9), but not significant among the bottom four quintiles (see Table 8). This is perhaps because most of the bottom four quintiles do not own goats. Instead chickens are the most important livestock for the poor in all three regions especially in the South (Tables 10 and 11) though the magnitude of effect is very low. However, cattle herd size was also significant among the bottom four quintiles in the central provinces, although the magnitude of the coefficient was also small, relative to that of the top quintile in the same region. Currently, chicken sales among the bottom four quintile households appear insufficient to raise the required cash to either purchase sufficient food or invest in agriculture to increase agricultural productivity and production to meet food security needs.
Table 10

Pooled probit model results for the bottom 80th percentile by region

 

North: bottom 80th percentile

Centre: bottom 80th percentile

South: bottom 80th percentile

 

Coeff.

Sig.

Coeff.

Sig.

Coeff.

Sig.

Head’s gender (1 = male)

−0.12

a

−0.08

 

−0.01

 

Head’s years of completed education

−0.06

c

−0.05

c

−0.08

c

Head’s age (years completed)

0.00

 

0.02

b

−0.01

 

Head’s age (squared term)

0.00

 

0.00

b

0.00

 

HH size in adult equivalent scale (AE)

0.83

c

0.30

c

0.22

c

HH size in AE (squared term)

−0.06

c

−0.01

c

−0.01

c

Head is engaged in salaried act. (1 = yes)

−0.38

c

−0.37

c

−0.64

c

Head is self-employed

−0.35

c

−0.41

c

−0.50

c

Cropped area in hectares

−0.32

c

−0.11

c

−0.04

 

Cropped area in hectares (squared term)

0.02

c

0.01

c

0.00

b

Cattle herd size

−0.03

 

−0.02

a

−0.02

c

Number of goats owned by the HH

−0.01

 

−0.01

a

0.00

 

Number of chickens owned by the HH

−0.02

c

−0.01

c

−0.01

b

HH used animal traction (1 = yes)

0.52

 

−0.25

b

0.01

 

HH used fertilisers (1 = yes)

0.07

 

−0.43

c

−0.26

 

HH hired permanent labour (1 = yes)

−0.72

b

−0.56

b

−0.24

a

HH hired seasonal labour (1 = yes)

−0.42

c

−0.57

c

−0.49

c

HH received extension services (1 = yes)

−0.16

b

0.03

 

−0.09

 

HH received price information

−0.11

b

−0.13

a

−0.15

b

Member of a farmers’ association (1 = yes)

−0.09

 

−0.04

 

0.06

 

Dummy for year = 2005

0.22

c

0.10

 

−0.03

 

Dummy for year = 2008

0.31

c

0.14

a

0.26

c

Constant

−1.04

b

−0.75

 

1.02

a

Number of observations

4779

 

3270

 

4405

 

Wald chi2

701.72

 

436.54

 

454.95

 

Prob > chi2

0.00

 

0.00

 

0.00

 

Pseudo R2

0.15

 

0.17

 

0.15

 

Percent predicted correctly

69.09

 

70.67

 

68.63

 

Authors’ calculations based on TIA02, TIA05, and TIA08

District dummies are included but not reported to save space

c, b, and a denotes significance at 1, 5, and 10 %, respectively

Table 11

Pooled probit model results for the top quintile by region

 

North: top quintile

Centre: top quintile

South: top quintile

 

Coeff.

Sig.

Coeff.

Sig.

Coeff.

Sig.

Head’s gender (1 = male)

−0.26

 

−0.20

 

0.22

 

Head’s years of completed education

−0.01

 

−0.04

 

−0.13

b

Head’s age (years completed)

0.13

c

0.06

a

−0.06

 

Head’s age (squared term)

0.00

c

0.00

a

0.00

 

HH size in adult equivalent scale (AE)

1.60

c

0.63

c

0.34

c

HH size in AE (squared term)

−0.13

c

−0.02

c

−0.01

a

Head is engaged in salaried act. (1 = yes)

−0.67

b

−0.60

c

−0.47

a

Head is self-employed

−0.67

c

−0.19

 

−1.08

c

Cropped area in hectares

−0.38

c

−0.12

b

−0.09

 

Cropped area in hectares (squared term)

0.02

 

0.00

 

0.00

 

Cattle herd size

−0.06

 

−0.01

 

0.01

 

Number of goats owned by the HH

−0.02

 

−0.02

b

−0.04

b

Number of chickens owned by the HH

0.00

 

−0.01

 

−0.03

b

HH used animal traction (1 = yes)

  

−0.17

 

−0.54

a

HH used fertilizers (1 = yes)

−0.94

b

−0.58

b

−1.48

c

HH hired permanent labour (1 = yes)

  

0.32

 

−0.62

a

HH hired seasonal labour (1 = yes)

−0.61

c

−0.85

c

−0.24

 

HH received extension services (1 = yes)

−0.12

 

−0.19

 

0.74

b

HH received price information

−0.22

 

−0.10

 

−0.44

 

Member of a farmers’ association (1 = yes)

0.59

b

0.37

a

−0.36

 

Dummy for year = 2005

0.84

c

−0.09

 

0.27

 

Dummy for year = 2008

0.98

c

0.09

 

1.00

c

Constant

−13.70

c

−1.81

a

2.34

a

Number of observations

789

 

1495

 

511

 

Wald chi2

214.11

 

245.66

 

129.35

 

Prob > chi2

0.00

 

0.00

 

0.00

 

Pseudo R2

0.34

 

0.30

 

0.35

 

Percent predicted correctly

88.21

 

86.89

 

82.19

 

Authors’ calculations based on TIA02, TIA05, and TIA08

District dummies are included but not reported to save space

c, b, and a denotes significance at 1, 5, and 10 %, respectively

The statistically significant coefficient on whether the head is engaged in non-farm activities signals the likely importance of non-farm cash income in ensuring household food security. For households in the bottom four quintiles, the magnitude of the coefficient on self-employment (and head’s education) was larger for the southern provinces, which reflects the higher non-farm employment opportunities and lower agricultural potential in that region; thus non-farm employment should be part of the long term strategy for reducing food insecurity and poverty in that region (Cunguara et al. 2011). Finally, with the exception of the receipt of credit in the central provinces, access to agricultural services (extension and price information) was not statistically significant among the bottom four quintile households. With regard to the receipt of extension, the lack of statistical significance might be related to farmers’ inability to adopt the technical recommendations provided by the extension workers (Walker et al. 2004; Mather 2009; Cunguara and Moder 2011). Low levels and variation in the data may also explain the lack of an effect of credit. In the Northern and Southern provinces only 2 % received credit and the impact was not significant, compared to 7 % in the central provinces, where receipt of credit had a somewhat significant effect. The lack of statistical significance in the result on receipt of price information was somewhat surprising, and contradicted results of Mather (2009), which estimated that receipt of price information increased crop income by 23 percent and 31 percent in the central and southern provinces, respectively. For the top quintile households, the receipt of credit and extension was positive and statistically significant both in the northern and central provinces.

Results of the Probit estimation for the pooled model are presented in the Tables 10 and 11. These results are mostly similar to those obtained in the Probit model for TIA08. Year dummies in the pooled regression suggest that food insecurity in 2008 was significantly worse than in 2002 and 2005. Also, ownership of chickens was more strongly correlated with food security among the bottom quintile households, while goats had a stronger relationship with food security for the top quintile households.

Conclusions and policy implications

Using a set of nationally representative household surveys, this paper assessed food security trends in rural Mozambique and whether progress was made in the drivers of food security in rural areas. Results show that rural food security declined between 2002 and 2008 and progress on the drivers of food security did not happen; whether the indicator analyzed was per capita cash income, use of modern farm inputs and technology, receipt of agricultural extension services or agricultural output and productivity.

The analysis of the drivers of food insecurity in rural Mozambique shows that trends differed noticeably by quintile of total maize production and region. The top quintile and central provinces fared relatively better in terms of food security and its drivers. In the short run, adoption of improved technologies appears to be an important driver of food security and should be promoted, especially among households in the top quintile of maize production. This recommendation is grounded on the finding that top quintile households were more likely to successfully adopt improved technologies and consequently become less food insecure. Because the food security gap is relatively large among households in the bottom quintile and the majority of households in the bottom quintile depend heavily on food purchases to meet their calorie requirements, non-farm employment opportunities are likely to increase their cash incomes and would appear more beneficial in addition to adoption of improved farming technologies. Likewise, the main strategy for reducing food insecurity in the southern provinces, where agricultural production potential is low, should probably be creation of non-farm employment for all quintile categories.

The use of targeted improved technologies suitable for the specific agro-ecological conditions in the south also proves to be of great importance. In particular, animal traction and/or mechanization are likely to enhance food security among those in the top quintile of maize production because they can cultivate relatively larger fields. Meanwhile, the use of animal traction was not significant among the poorest households partly because they farmed smaller fields and land is a binding constraint to production for these households. Therefore, if the poor in the south are to realize significant food security gains directly from farm production and adoption of improved technologies they may need to first expand their cropped areas.

Since not all smallholder farmers are likely to be commercially viable in the short to medium term, less competitive households may fail to increase their farm sizes and incomes through farm output market participation and may continue to experience food insecurity. Therefore, enabling them to enter more remunerative non-farm labor markets and engage in alternative entrepreneurial opportunities is critical (Boughton et al. 2007). Given the country’s projected energy boom the government may seek strategies that facilitate the engagement of these poor households in the energy sector or to leverage the energy sector’s linkages with other sectors of the economy, in particular agriculture which accounts for the majority of rural livelihoods, so as to benefit the poorer households. Also, as poor households are often confined to low-return non-farm activities, finding strategies for breaking barriers to entry into higher-return non-farm activities will be critical in this regard. In the long term this may entail increased and sustained investments in education (Reardon 1997), which can lead to increased labor productivity and growth over the long term.

Footnotes
1

That increase was dominated by trends in rural areas where the headcount ratio increased from 55.3 % to nearly 56.9 %. Urban poverty dropped from 51.5 to 49.6 %.

 
2

This commitment is also built into the CAADP process that commits to higher and sustained spending in the agricultural sector through productivity and marketing enhancing investments.

 
3

For recent reviews and perspectives, see Headey and Ecker (2013), FAO and IFAD (2012), Cafiero (2012) and Carletto et al. (2012).

 
4

These include the food variety score, the dietary diversity score and WFP’s food consumption score.

 
5

The recommendation however is that dietary diversity data, which have been shown to be strongly correlated with micronutrient and macronutrient intake, be collected on a regular basis to enable better analysis of trends in food and nutrition security (Ruel 2003; Headey and Ecker 2013).

 
6

Cafiero (2012) and Smith and Subandoro (2007) adeptly argue that data on body sizes (height and mass) as well as physical activity levels of household members would be essential to more accurately estimate the household-specific minimum calorie requirement. In addition, one would have accordingly to adjust the estimate for pregnant and lactating women, if present in the household. Unfortunately, the household surveys used in this study do not contain all the necessary data. Therefore, our estimates of food insecurity are computed assuming a median mass and height (body mass index) for each age and sex and corresponding moderate physical activity levels as defined by the FAO (Food and Agriculture Organization of the United Nations) (2008).

 
7

While Engel’s law would predict that food shares decrease with increases in income, recent studies have found that the poor and wealthy may spend similar shares of their income on food. In the Mozambican context the data suggest that poorer households in fact spend higher shares of their income on food, primarily because of differences in the composition of their food basket (Mather et al. 2008), and poorer households consume more calorie-dense foods that are cheaper. This implies that our conversion of food shares into calories using maize prices is less likely to be problematic among the poorer households, but may underestimate food insecurity among the wealthier households whose food baskets may include more expensive sources of calories.

 
8

Poor sanitation facilities can result in diseases and co-morbidities that lower food utilization e.g. intestinal parasites and diarrhea that reduce calorie intake and nutrient absorption. Similarly, poor food storage can lead to aflatoxin contamination and other food safety challenges that lower food utilization and stability of food supply.

 
9

This means that households that have cash flow constraints and credit constraints and do not have the ability to store food may sell farm output when prices are low, during the harvest season, only returning to the market to purchase food when prices are high.

 
10

For instance when floods occurred in 2000, households in remote areas of Mozambique did not have access to the rest of the country and had limited access to food from regional and international markets.

 
11

A brief description of the TIA surveys can be found at the following website address: http://www.aec.msu.edu/fs2/mozambique/survey/index.htm. Mather (2009; 2012) and MPF/IFPRI/PU (Mozambique Ministry of Planning and Finance/International Food Policy Research Institute/Purdue University) (2004) provide additional descriptions of the TIA data.

 
12

Depending on the region, other crops can dominate as cash crops, such as cashew in Inhambane. For the most part the spatial distribution of crop sales is a result of the distribution in climate and agro-ecological conditions that suit different crops’ production.

 
13

Marginal probability effect for female household head status is about an 11 % increase in food insecurity while the marginal effect of an additional adult equivalent is 46 % for the bottom four quintile households in the North. These marginal effects are evaluated at the means for continuous explanatory variables, with the dummy variables set at 0. The formula for converting the Probit regression coefficients to marginal probability effects is \( m \arg inaleffect=\frac{1}{\sqrt{2\pi}} \exp \left(-\frac{1}{2}{\left({\beta}^{\prime } x\right)}^2\right)\times \beta \), where β is the vector of coefficients shown in the tables and x is the vector of independent variables, evaluated at the means or at 0 in the case of the dummy or categorical variables. For dummy variables the marginal effect is the difference between the predicted probability with and without that dummy variable set to 1.

 

Copyright information

© Springer Science+Business Media Dordrecht and International Society for Plant Pathology 2014

Authors and Affiliations

  1. 1.International Food Policy Research InstituteWashingtonUSA
  2. 2.Department of Agricultural Food and Resource EconomicsMichigan State UniversityEast LansingUSA

Personalised recommendations