Food Security

, Volume 4, Issue 4, pp 593–606

Short-term determinants of malnutrition among children in Malawi

Authors

    • Faculty of EconomicsUniversity of Pavia
Original Paper

DOI: 10.1007/s12571-012-0221-0

Cite this article as:
Sassi, M. Food Sec. (2012) 4: 593. doi:10.1007/s12571-012-0221-0

Abstract

Short-term determinants of Severe Acute Malnutrition in children in Malawi during the period 2003 to 2009 were investigated in the three regions that compose Malawi – northern, central and southern – through an OLS approach and a first-order autocorrelation model. Explanatory variables were selected according to the definition of food security provided by the 1996 World Food Summit. Monthly changes in the number of children admitted to Nutrition and Rehabilitation Units was the impact variable adopted. The explanatory variables selected included a proxy of household income spent on food and the monthly variation in domestic price of maize, its trend, cyclical, seasonal and irregular components, informal cross-border imports in maize, urea price, non-food price index, and number of Nutrition and Rehabilitation Units. The study integrates recently developed studies on food insecurity in Malawi with regional and monthly perspectives. Results verify that child malnutrition is a chronic problem fuelled by transitory food insecurity, including seasonal and temporary features, with the common determinant being the market dependence of households on food purchases during the lean season. This impact is exacerbated by regional-specific explanatory variables: the variation in seasonal and irregular maize price components and the non-food price index in the central region, along with the cyclical maize price component and net cross-border maize imports in southern Malawi.

Keywords

Child malnutritionFood policyRegional developmentMalawi

JEL Classification

Q18

Introduction

In Malawi, food consumption is generally characterized by a significant risk of malnutrition (Ecker and Qaim 2010). Food crops are insufficiently diversified, there being a high dependence on maize for income and growth but low productivity of the agricultural sector in general. Other problems include poor soil fertility, erosion, poor water management, stagnant rural labour markets, high rates of HIV/AIDS, price shocks and mistaken policies, all of which affect the already fragile food security structure of Malawi and compromise the ability of the population to cope with such hazards (FAO and WFP 2002; Harrigan 2008). In this context, children are strongly affected by seasonal and transitory food insecurity and those who are in a transitory state of food insecurity have a high probability of becoming chronically hungry if the problem is not properly addressed.

The current level of malnutrition presents a challenge to the achievement of the Millennium Development Goals (Government of Malawi, Ministry of Agriculture and Food Security 2010). As reported by the World Food Programme (2010), chronic malnutrition in Malawi has not been reduced significantly since 1990; it is persistent, widespread and severe. In this context, children’s nutritional status is a serious problem. Nearly half of all children under 5 years of age are stunted. This was 49 % in 1992, 48 % in 2004 and, according to the last Malawi Demographic and Health Survey, 47 % in 2010 (National Statistical Office and ICF Macro 2011; National Statistic Office 2005; 1993). More than 50 % of these children are severely stunted and, in some districts, the prevalence of acute malnutrition reaches 5 % (National Statistical Office and ICF Macro 2011; Sahley et al. 2005; World Food Programme 2010). This precarious food security situation is directly linked to poverty, a consequence of the slow economic development of the country and of its macroeconomic problems. Malawi is one of the poorest countries in the world and, in 2011, its human development index was 0.400, which gives the country a rank of 171 out of 181 (http://hdrstats.undp.org/en/countries/profiles/MWI.html). The country qualifies for full Heavily Indebted Poor Countries' relief and is dramatically donor dependent (Chinsinga 2007).

Malawi is a landlocked country and is connected to the major ports by poor infrastructure; this situation has a direct effect on the availability of commodities and on their price (FAO and WFP 2002). The economy of the country is heavily dependent upon agriculture, which accounts for about 39 % of GDP, 85 % of the labor force and 83 % of foreign exchange (Chirwa et al. 2006). Apart from the large estate farmers involved in exporting cash crops, the bulk of the sector consists of small agricultural producers, and three-quarters of the population depends on this activity for their main livelihood (Chinsinga 2007).

Smallholder farmers cultivate mainly maize, a crop that is the principal staple food and the focus of food security policy. However, only 20 % of farmers produce a surplus, while the majority are classified as Mlini, or subsistence farmers. The size of their land is inadequate, with 1.2 ha per household (i.e. 0.32 ha per capita), the soil fertility is low and rain-fed agriculture dominates (National Statistical Office of Malawi 2010; Bohne 2009; Minot 2010). Smallholders do not produce sufficient food to last them from one season to another (Ecker and Qaim 2010). During the lean season, when these farmers run out of their maize stock, they have to rely on the market. But here, prices are high owing to shortages and do not decline until after harvest when smallholders have their own maize production and can sell part of it. The gap between the production and consumption needs is covered by casual labor (ganyu); however, income from casual labor is not enough to provide an adequate livelihood for the majority of households. Thus, food security in Malawi is highly vulnerable not only to climatic, economic and social shocks, but also to the seasonal calendar, with seasonal effects that vary from year to year.

Since gaining independence in 1964, food insecurity has represented a high-priority problem for the government of Malawi, and reliable information on its determinants is necessary in order to design suitable policy interventions and to target assistance to those most in need (Lewin and Fischer 2010).

In light of these considerations, this paper focuses on food insecurity in the three regions that compose Malawi — northern, central and southern — with the aim of investigating, on a monthly basis, the short-term determinants of severe acute malnutrition (SAM) in children over the time-period 2003 to 2009.

The study considers monthly time series in order to take into account the role of transitory food insecurity. In this respect, the current paper integrates the recent analyses on malnutrition carried out for the country on an annual basis. These analyses are mainly focused on 2004/2005 due to the rich data provided by the second Malawi Integrated Household Survey (see, for example, Lewin and Fischer 2010; Ecker and Qaim 2010). More generally, the recent empirical literature on Malawi estimates the determinants of food insecurity at the national level (see, for example, Cornia et al. 2012); at a specific community level (see Aiga et al. 2009); in the rural areas as aggregate (see Ecker and Qaim 2010); in the rural regions (see Lewin and Fischer 2010); or describes them at the livelihood zone level (see Malawi National Vulnerability Assessment Committee 2005a; 2003).

The paper integrates the study by Cornia et al. (2012) with the regional perspective. This angle is important in Malawi where geographic, economic and social conditions vary across space. Data availability is often a serious constraint to spatial investigations, particularly in the analyses of time series, and this holds true also for Malawi. The country is divided into 28 districts within three regions, northern, central and southern. Due to a lack of information, the paper compares empirical evidence for Malawi only at the regional level in order to understand the extent of possible spatial disparities in the variables explaining food insecurity; this information is relevant in designing more specifically targeted interventions.

The impact variable adopted is SAM in children. In Malawi, the poor nutritional status of a large number of children provides direct evidence that food insecurity may be getting worse (Harrigan 2008). Children are especially vulnerable to changes in food intake, and for this reason the impact variable adopted, the change in the number of children admitted to the Nutrition and Rehabilitation Units (NRUs), helps to capture aspects of the distribution of wellbeing within the households not adequately reflected by other indicators (Pinstrup-Andersen 2009). In this respect, the analysis integrates the studies developed for Malawi that mainly refer to the aggregate household level.

The paper estimates, through an ordinary least squares (OLS) approach and a first-order autoregressive model, a set of equations whose variables are selected according to the technical concept of food security provided by the 1996 World Food Summit and its basic dimensions. The estimation of these parameters follows a pragmatic approach due to the severe limitations of the dataset. However, the results of the empirical investigation are linked to the targets of food security interventions, defined by the international community, and to the aspects referred to when evaluating the respective policies. For this reason, they introduce observations of interest, particularly from a policy-making point of view.

The paper is arranged as follows: Section 2 introduces Materials and Methods starting from the analytical framework on which the empirical analysis is rooted and, subsequently, introducing the variables selected, and the estimated models; Section 3 presents results; Section 4 presents discussion of the results; and conclusions are reached in Section 5.

Material and methods

Analytical framework

The empirical analysis refers to a conceptual framework based on the technical definition of food security provided by the 1996 World Food Summit: “All people at all times have both physical and economic access to sufficient food to meet their dietary needs for a productive and healthy life” (FAO 1996). The definition reflects the food security discussions developed in the second half of the 1980s (Maxwell and Wiebe 1998); it is the result of a balancing act between the Malthusian concerns of the 1970s, focused on food production, and the concern with consumption and access, as developed through the contribution of Sen (Sen 1981; Dréze and Sen 1989).

The concept introduces, as policy objectives, the following three dimensions: i) food availability. This is reached when all individuals within a country are able to acquire sufficient quantities of safe and nutritious food that meets their food preferences and implies that it is present in reasonable proximity to them, or within their reach; ii) access to food. This is satisfied when households and all individuals within them have adequate resources to obtain appropriate food for a nutritious diet; iii) food utilization. This is fulfilled when there is proper biological use of food as required by a diet that provides sufficient energy and essential nutrients, potable water and adequate sanitation (Sassi 2006). At the top of Fig. 1, nutritional status, which is the condition of health of a person determined by the intake and utilization of nutrients, is affected by a set of interrelated factors of the food economy and the household context, and by confounding factors.
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig1_HTML.gif
Fig. 1

Food security framework

The food economy describes the different ways people have adapted their resources to obtain their food. In this context, the natural, capital and human resources endowment owned by households defines the set of productive activities (food, cash and non-agricultural) they can pursue in meeting part of their income; the total income availability of households also includes public and private transfers or loans that are spent on food and non-food items or saved.

Domestic food production, for subsistence or for the market or food stocks, combined with net formal and informal food imports and food aid, have direct impacts on food availability. Food availability affects food prices and consequently market purchases by households which are contingent upon income. Other factors that affect the nutritional status of individuals are related to food utilization in terms of intra-household food allocation, householders’ knowledge of storage and processing techniques, basic principles of nutrition, health status and proper care of the sick. Food security is a dynamic concept due to its feedback effect on human resources, labor productivity and the potential for householders to earn income. Furthermore, confounding factors frame the food economy and the household context, consisting of international, national and sub-national elements, such as the demographic, economic, political, socio-cultural and environmental conditions and risk, hazard and shock factors.

Data set

The monthly data available for Malawi allow quantifying of only a few aspects that affect child malnutrition (Fig. 1). They are:
  • The number of children admitted to the NRUs, which describes child nutritional status;

  • The domestic price of maize in Malawian kwacha (MWK, 1 US Dollar = 270 MWK), its trend, cyclical, seasonal and irregular components, and a proxy of household income spent on food, which are variables with a direct influence on food access;

  • The informal cross-border imports in maize and one of the most important agricultural production factors, the price of urea (a nitrogenous fertilizer) in Malawian kwacha (MWK), which act on the food availability dimension;

  • The non-food price index, which is a confounding factor;

  • The number of NRUs, introduced as a control variable.

Due to lack of data, three of these variables used in the region specific analysis (i.e., the proxy of household income spent on food, the price of urea in MKW, and the non-food price index) are only measured at the national level. As a consequence, findings are only indicative and not statistically rigorous.

Nutritional status and number of NRUs

The number of SAM children admitted to the NRUs was the impact variable adopted. The admission criteria introduced in 2002 was weight-for-height <70 % of the median or mid-upper arm circumference (MUCA) <11 cm or presence of bilateral oedema (Government of Malawi, Ministry of Health 2006). As the feeding protocols are based on weight-for-height criteria, the number of children admitted to NRUs can be assumed as an indicator of the current nutritional status of children resulting from recent nutritional intake and illness episodes. When the acute forms of malnutrition are taken into consideration, the indicator also reflects severe and life-threatening situations of food shortage (World Food Programme 2010).

However, it should be mentioned that child admissions to NRUs capture only part of the real size of the problem of SAM in children in Malawi. In fact, children without medical complications are treated at home within the Outpatient Therapeutic Programs (OTP). After the 2002 food crisis, the Ministry of Health introduced a Community Therapeutic Care (CTC) approach beyond the NRUs to serve children less than 12 years of age and pregnant and lactating women. The CTC program combines the following four components:
  • Community outreach, which is aimed at community sensitisation, mobilisation, active case-finding and referral and case follow-up;

  • Outpatient Therapeutic Programme, which treats the majority of severely malnourished children with appetite and without medical complications at their home with ready-to-use therapeutic food, systematic medication and a weekly check-up at an OTP site;

  • Nutrition Rehabilitation Unit, which provides intensive inpatient care to a small proportion of SAM children with complications and at the highest risk of death until the patients are stabilised and suitable for OTP;

  • Supplementary Feeding Programme, which provides dry take-home rations to moderately acutely malnourished children and pregnant or lactating women, and to patients discharged from OTP or NRU (Government of Malawi, Ministry of Health 2006).

Furthermore, according to the local sources, only those within an approximately 18-km radius of NRUs generally have access to the care they provide, and often, due to socio-cultural reasons, SAM children are not brought to these centers by members of their families. Thus, the impact variable adopted underestimates the phenomenon under investigation, with consequences on the accuracy of results of the econometric analysis. This aspect should be carefully considered during the interpretation of the estimated correlation coefficients, which should be taken as indicative and not statistically rigorous.

However, despite this limitation, the impact variable adopted is the only indicator regularly collected at the NRU level and on a monthly basis. Furthermore, as recently underlined by Teller (2008), numbers may vary due to the quality of the reporting system. For this reason, several aspects of data quality have been verified. For example, admissions to the program do not represent a constant proportion of the total burden, within and between years.

Data are available starting from 2001, but in 2002 the admission criteria to NRUs changed following the reform of the public health approach introduced by the government of Malawi (Government of Malawi and Health 2006). As a consequence, data after 2002 are not comparable with that from previous months. This fact is illustrated by Fig. 2 which shows a significant increase in child admissions after 2003, in large measure explained by the change in feeding protocols (de Menezes 2004). For this reason, the monthly time series adopted in the analysis starts from January 2003.
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig2_HTML.gif
Fig. 2

Number of child admissions to NRUs in Malawi (2001–2010)

Figure 2 also points out that the 2010 monthly data are unusually low when compared with the previous years. Those months have been excluded due to the fact that the nature of the change cannot be understood. Some local sources unofficially attribute these low numbers to a data collection process which has not yet been completed. Thus, the empirical investigation is restricted to the time period from January 2003 to December 2009. During this period, the number of child admissions was greatest in the central and southern regions, with the central region experiencing the highest variability (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig3_HTML.gif
Fig. 3

Number of child admissions to NRUs in Malawi by region (2003–2009)

An additional limitation of the impact variable, the number of children admitted to NRUs, is that it might be related to the number of centers providing child care in specific months. This number is not stable over time (Fig. 4). It depends on, among other things, the severity of hunger in a specific period, policy considerations and food availability. When this relationship is strong, the dependent variable selected might not represent the intensity and the change in hunger properly. For this reason, following Cornia and Deotti (2008), the number of NRUs is taken as a control variable.
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig4_HTML.gif
Fig. 4

Number of NRUs in Malawi by region (2003–2009)

The monthly child admissions to the single NRUs have been provided by the Ministry of Health of Malawi. From this dataset, the number of children admitted and the number of centers have been aggregated at the regional and the national level.

Food and non-food prices

The monthly price of maize was assumed to be a good proxy of food prices in general as maize is the dominant staple food (Minot 2010). This crop accounts for 46 % of total food quantity, more than 60 % of energy, and almost half of total protein consumption. It is also the source for 67 % of total iron, 65 % of total zinc, and almost 70 % of total riboflavin consumed (Ecker and Qaim 2010).

Maize is cultivated by 97 % of farmers, predominantly smallholders, on 54 % of the arable land; it is grown mainly for subsistence, with less than 20 % being marketed (Minot 2010; Bohne 2009).

During the period analyzed, the maize trade openness index was small over an increasing production (Fig. 5). This is a typical feature of a non-traded commodity and is partly determined by the impact of policies to restrict exports or build local stores in the country. For example, the export of food in 2007 may have been reduced as silos were available to store larger quantities in the country. In one year there were also significant donations to Zimbabwe (Jayne et al. 2008; Minot 2010). As a consequence, maize prices are mainly determined by domestic supply and demand, as confirmed by Cornia et al. (2012). For this reason, the international price of maize was excluded as an explanatory variable in the empirical investigation.
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig5_HTML.gif
Fig. 5

Maize production (tons) and openness index* (2003–2009)

Maize price fluctuations in Malawi are significantly affected by the dynamic seasonal structure of production and by cyclical factors. The latter have assumed a specific importance in the recent price spikes, aggravating the food crisis in the country (see, for example, Chirwa 2009). Thus, the analysis also takes into consideration the trend, seasonal, cyclical and irregular components of maize prices, estimated according to the methodology explained in the next section.

Monthly data for maize prices in Malawian kwacha (MWK) by market have been provided by the FEWSNet National Representative in Malawi. In this dataset, the number of reference markets increases over time (there were 30 in 2003, 42 in 2004, 67 in 2005 and 70 from 2006 to 2009). Furthermore, the time series data at the market level was often incomplete; the most critical year was 2003, when many monthly and market values were lacking. In order to overcome this problem, the study is only based on the 25 markets characterized by the more complete time series. For each of these markets, data have been aggregated by district and then by region according to the simple average method. Table 1 illustrates the analyzed markets by district and region. More precisely, data have allowed aggregation of the markets of two districts in the northern region, Karonga and Mzimba; two districts in the central region, Lilongwe and Ntcheo; and three districts in the southern region, Machinga, Mangochi and Mulanje.
Table 1

Markets used by district and region

Markets

District

Region

Chilumba

Karonga

Northern

Karonga

Embangweni

Mzimba

 

Jenda

Mzimba

Mzuzu

Kasiya

Lilongwe

Central

Lilongwe

Mitundu

Nanjiri

Nambuma

Nkhoma

Nsundwe

Lizulu

Ntcheo

 

Ntcheu

Ntonda

Sharpevaley

Tsangano Turn Off

Liwonde

Machinga

Southern

Ntaja

Mangochi

Mangochi

 

Monkey Bay

Namwera

Luchenza

Mulanje

 

Muloza

The non-food price monthly data was approximated by the average of the non-food items of the national Consumption Price Index provided by the National Statistics Office of Malawi. This information is available only at the country level and for the rural and urban areas separately; thus, the variable was the same in all regions and equal to the aggregate value for the country.

Income spent on food

As no monthly data were available for income spent on food, this variable was approximated by an indicator of households’ market dependence for food purchase. This is equal to 1 from November to March, when the share of households dependent on the market for food is highest, and zero in the other months. In the interpretation of results, it should be taken into consideration that this is a very crude indicator of the degree of market dependence, which varies in relation to many factors; among them, differences in local crop calendar, household wealth, farm size and the size of the preceding harvest. Due to the fact that maize is the main source of food, households in Malawi are seasonally dependent on the market for food purchases during the lean season, which according to the seasonal calendar (http://www.fews.net/Pages/timelineview.aspx? gb = mw&tln = en&l = en) lasts from November to March. Figure 6, based on the data provided by the World Food Programme (2010) for 2009, gives an indication of the severity of this trend in rural areas, where almost 90 % of the country’s population lives (World Bank 2007b).
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig6_HTML.gif
Fig. 6

Rural households who rely mainly on purchase by month (2009)

In Malawi, for the majority of households, income is not enough to fund an adequate livelihood; during the lean season household nutritional status deteriorates dramatically. For this reason the period from November to March is classified as the hunger season. Figure 7 provides a measure of this vulnerability. It shows the 2009 monthly data of the Income Gap (IG), an indicator that has been estimated according to the following formula:
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig7_HTML.gif
Fig. 7

Income gap* in Malawi kwacha (MWK) (2009)

$$ IG=MW-BN $$
(1)
where MW is the minimum wage in rural areas, provided by the ILO, and BN represents the basket needs in Lilongwe, as elaborated by the Centre for Social Concern (CSC) – Malawi.

IG reaches the greatest negative values during the lean season when market dependence for food purchasing is highest. Thus, the variable adopted can also be interpreted as an indicator of household welfare, which deteriorates when the indicator is equal to 1.

Informal trade and agricultural production factors

In Malawi there are two additional components that are important in food security analysis. One is the informal cross-border imports, particularly of maize, which has been a major factor in avoiding widespread hunger during food deficit periods (World Bank 2007a). For this reason the World Food Programme and FEWSNet established a system to monitor informal trade flows that began to operate in July 2004 (Famine Early Warning System Network 2005). Informal cross-border trade is a very important source of livelihood for those involved and their families. Most of the maize is informally carried across the borders to the Malawian side by cyclists for whom this activity represents their main income source (Bata et al. 2005; World Food Programme and FEWSNET 2004). They are locally known as Adyanji which, in the local language, means that without this business they would have nothing to eat. These trade flows contribute significantly to food availability in southern Malawi, where agriculture is poor and food deficits are frequent; in this region they also represent a potential factor for food price stabilization (Bata et al 2005).

The empirical analysis makes reference to the net informal cross-border imports in maize, expressed in metric tons, for the period from July 2004 to December 2009. This information has been obtained by subtracting the informal cross-border export flows from the informal cross-border imports in maize. Both data sets were collected from the monthly Informal Cross-border Food Trade in the Southern Africa Bulletin published by FEWSNet. This source provides data by country of origin and destination, not by entry point. The information at the regional level has been estimated assuming that greater interactions are between Mozambique and southern Malawi; Zambia and central Malawi; and Tanzania and northern Malawi. It should also be taken into consideration that the major source country is Mozambique (World Food Programme and FESWNET 2005). Data on informal trade in maize from January 2003 to June 2004 have been estimated according to the evidence provided by the literature (see, for example, Centre for Regional Agricultural Trade Expansion Support 2003; Whiteside 2002).

The second important component of food security in Malawi is represented by agricultural inputs, both seeds and fertilizers. The stabilization of their prices has always been one of the components of the food security policy of the government of Malawi (Zant 2005; Dorward and Chirwa 2011; Dorward et al. 2008). As no monthly data was available for seeds, the analysis only focuses on fertilizers. They are one of the determinants of a household’s production due to the very low soil fertility of Malawi’s agricultural land, and an important production cost that affects their expenditure capacity. In the 2000s, despite the efforts of the government of Malawi, high prices of fertilizers dramatically compromised the progress of the country towards achieving the objective of food security. At the beginning of the decade, the input support programs, combined with favorable weather, contributed to bumper harvests for three consecutive years. However, in that period the international price of fertilizers and particularly of urea, the most common fertilizer in the country, started to rise. As fertilizer in the country has been fully supplied by imports, the price increase was transmitted into Malawi’s domestic market where it was further exacerbated by the local currency devaluation (Fig. 8).
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig8_HTML.gif
Fig. 8

Urea monthly price index, US$ and MKW (100 = monthly average, 2000), 2003–2009

In this analysis, the focus is on the possible impact of a change in fertilizer prices on the severity of SAM in children, due to its effect on the income of households and thus on their economic access to food. The empirical investigation includes the urea monthly price in US dollars per metric ton, provided by the World Bank and expressed in national currency per U.S. average period through the monthly exchange rate. The sources of these data have been the IMF Statistics Department, and the African Development Bank.

Methodology

The empirical analysis investigates two models that refer to the monthly time series data from 2003 to 2009. The first is specified by the following equation:
$$ \begin{array}{*{20}{c}} {\Delta \ln \left( {NA{{D}_{{i.t}}}} \right)={{\beta }_{1}}+{{\beta }_{2}}\left( {HS{{I}_{{i.t-1}}}} \right)+{{\beta }_{3}}\Delta \ln \left( {PD{{U}_{{i.t}}}} \right)+{{\beta }_{4}}\Delta \ln \left( {N{{C}_{{i.t}}}} \right)+} \hfill \\ {+{{\beta }_{5}}\Delta \ln \left( {UR{{E}_{{i.t}}}} \right)+{{\beta }_{6}}\Delta \ln \left( {IC{{B}_{{i.t}}}} \right)+{{\beta }_{7}}\Delta \ln \left( {PN{{F}_{{i.t}}}} \right)+{{\mu }_{{i.t}}}} \hfill \\ \end{array} $$
(2)
where Δ is the difference operator (so \( \Delta {x_t}={x_t}-{x_{t-1 }} \)), ln is the natural logarithm, i is the region (central, northern and southern), t the month, NAD the number of child admissions to feeding centers; HSI is the dummy variable for households’ market dependence for food purchase, equal to 1 in the months of maximum market dependence for maize demand (November, December, January, February and March), when as previously underlined the income gap reaches the highest negative values, and to 0 in the other months; PDU is the domestic price of maize in MWK, NC the number of NRUs, URE the urea prices in MWK, ICB the informal cross-border trade, PNF the non-food price index and μ the error term.
The second model disaggregates the monthly time series of the domestic price of maize in MWK (PDU) into the four components suggested by the literature. They are:
  • A trend price component (TR), which captures the long-term permanent component of maize prices and could technically include demand and supply conditions, and any unobserved factor that evolves in a fairly continuous manner;

  • A cyclical price component (CY), which expresses the long-term oscillations around the trend, which are not regular and can vary depending on random shocks;

  • A seasonal price component (SF), which gives the short-term fluctuations in the time series that occur periodically at the same time every year and that reflect, for example, the fact that the supply conditions of maize are influenced by climatic events;

  • An irregular or random price component (IR), which captures short-term fluctuations in the time series that are erratic in nature and follow no regularity in their occurrence, such as those related to unforeseen events like floods, earthquakes, wars and famines (Harvey 1990).

The examination of the graph for trends, seasonal price components and statistics tests, such as the F-tests for seasonality, have suggested that these effects interact to give the observed time series, according to a multiplicative model specified as follows:
$$ PDU=TR*SF*CY*IR $$
(3)

The price components have been estimated with two methodologies. The X-12 monthly seasonal adjustment method (Findley et al. 1998) has allowed the calculation of the seasonal (SF), the random effects (IR) and the aggregate long term (TR*CY) components. CY is often of irregular length and difficult to estimate due to the economic assumptions required. In order to overcome the issue, TR has been calculated with the Hodrick-Prescot Filter (Hodrick and Prescott 1997), with a smoothing parameter specified using the power rule of 2 and yielding the original Hodrick-Prescot value of 14,400. Finally, the trend-cycle price component (TR*CY) has been divided by TR for the calculation of CY.

The meaning of the cyclical price component needs to be clarified. As illustrated by Fig. 9, CY assumes a specific importance in the 2002/03, 2005/06 and 2008/09 maize price surge, when it aggravates the food crisis in the country. The causes of food shortages are complex and weather conditions have an important role in an agricultural sector dominated by rain-fed farming. However, the peaks in the cyclical component of maize price were strongly affected by various factors during the three agricultural years, including mismanagement of the country’s strategic grain reserve stock, poor crop estimates, a chaotic delay response in terms of maize imports and late implementation of a fertilizer program (Harrigan 2008; Minot 2010). Figure 9 also shows a positive effect on the cyclical price component on maize prices in 2007. This was related to good rains during the 2006/06 agricultural year, but also to the expanded fertilizer subsidy programme, introduced in 2005/06, which led to a substantial surplus in maize production (Jayne and Tschirley 2009).
https://static-content.springer.com/image/art%3A10.1007%2Fs12571-012-0221-0/MediaObjects/12571_2012_221_Fig9_HTML.gif
Fig. 9

Cyclical component of maize price by region – monthly data (2003-2009)

As the irregular component of maize price captures short-time fluctuations erratic in nature, such as those related to weather, the cyclical price component can be interpreted as representative of policy changes able to produce an expansion or a contraction in maize production with a significant reduction or increase in prices.

The price components in Eq. 3 have been included in Eq. 2 instead of PDU to give the following equation:
$$ \begin{array}{*{20}{c}} {\Delta \ln \left( {NA{{D}_{{i,t}}}} \right)={{\beta }_{1}}+{{\beta }_{2}}\left( {HS{{I}_{{i,t-1}}}} \right)+{{\beta }_{3}}\Delta \ln \left( {N{{C}_{{i,t}}}} \right)+{{\beta }_{4}}\Delta \ln \left( {UR{{E}_{{i,t}}}} \right)+{{\beta }_{5}}\Delta \ln \left( {IC{{B}_{{i,t}}}} \right)+} \hfill \\ {+{{\beta }_{6}}\Delta \ln \left( {CP{{N}_{{i,t}}}} \right)+{{\beta }_{7}}\Delta \ln \left( {T{{R}_{{i.t}}}} \right)+{{\beta }_{8}}\Delta \ln \left( {S{{F}_{{it}}}} \right)+{{\beta }_{9}}\Delta \ln \left( {C{{Y}_{{i,t}}}} \right)+{{\beta }_{{10}}}\Delta \ln \left( {I{{R}_{{i.t}}}} \right)+{{\mu }_{{i,t}}}} \hfill \\ \end{array} $$
(4)

Both the models described by Eqs. 2 and 4 have been first estimated through an OLS approach. Subsequently, the problem of first-order auto-correlation, found for the northern and central regions, has been faced with the first-order auto-correlation model, as discussed in the next paragraph.

Results

Statistic tests have suggested a requirement to estimate Eqs. 2 and 4 with an OLS method for the southern region and to adopt a first-order autoregressive model for the northern and central regions. Table 2 presents the OLS estimates of the two previously introduced models, for each of the three Malawian regions. At the bottom of the table, the statistic tests indicate that, apart from the models for the southern region, there is a problem of first-order auto-correlation, as suggested by the Durbin-Watson test. More precisely, for the models of the northern region the positive auto-correlation is a problem, while for those of the central region, the Durbin-Watson test lies between the bounds used to verify the hypothesis of zero auto-correlation. In the latter case, a conservative approach could be adopted, rejecting the null hypothesis (Wooldridge 2002). However, for both the central and northern regions, the first-order autocorrelation has been solved, estimating a first-order autoregressive (AR(1)) model, with the error term of Eqs. 2 and 4 specified as follows:
Table 2

OLS estimate of the determinants of child admissions to NRUs by region (2003–2009)

 

Central Region

Northern Region

Southern Region

Equation 2

Equation 4

Equation 2

Equation 4

Equation 2

Equation 4

C(1)

−0.08684** (0.0131)

−0.16321 (0.2464)

0.00293 (0.9512)

0.04955 (0.7261)

−0.05774 (0.1120)

−0.03456 (0.8074)

D(HSI)

0.16818** (0.0377)

0.18632* (0.0618)

0.27639** (0.0161)

0.25804** (0.0419)

0.19871** (0.0326)

0.25805** (0.0112)

D(PDU)

0.56795*** (0.0062)

 

0.25178 (0.4093)

 

0.37435 (0.1362)

 

D(PNF)

7.95776*** (0.0004)

8.08989*** (0.0004)

−2.14792 (0.4696)

−2.57385 (0.4071)

3.13653 (0.1599)

3.31548 (0.1456)

D(NC)

1.13189 (0.0000)***

1.12408*** (0.0000)

1.23330*** (0.0000)

1.24094*** (0.0000)

1.09750*** (0.0000)

1.10036*** (0.0000)

D(ICB)

−0.05992 (0.2249)

−0.05923 (0.2480)

−0.03817 (0.5729)

−0.04559 (0.5250)

0.12239** (0.0185)

0.10788** (0.0412)

D(URE)

−0.42304* (0.0648)

−0.36927 (0.1086)

−0.20496 (0.5189)

−0.26361 (0.4387)

−0.17814 (0.4544)

−0.26321 (0.2841)

D(TR)

 

4.48630 (0.5579)

 

−2.03312 (0.7996)

 

−0.71053 (0.9272)

D(SF)

 

0.63384* (0.0774)

 

0.43515 (0.4177)

 

0.00845 (0.9822)

D(CY)

 

−0.22571 (0.6715)

 

0.98706 (0.3022)

 

1.07351* (0.0728)

D(IR)

 

1.01646*** (0.0067)

 

−0.16350 (0.7539)

 

0.21691 (0.7130)

R2 ADJ

0.85904

0.86027

0.77936

0.77397

0.77350

0.77251

F-statistic

84.2874 (0.0000)

57.0956 (0.0000)

49.2763 (0.0000)

32.1994 (0.0000)

47.6720 (0.0000)

31.9395 (0.0000)

Durbin-Watson

2.471030

2.571457

2.755249

2.746339

2.050800

2.103357

p-value in brackets. *Significant at p < 0.10. ** Significant at p < 0.05. ***Significant at p < 0.01

$$ {\mu_{i,t }}=\rho {\mu_{i,t-1 }}+{\varepsilon_{i,t }} $$
(5)
where ρ is the first-order serial correlation coefficient and ε the error term at time t. The fit of the AR(1) model has been compared to the lag model; the former approach shows the lowest values for both the Akaike and the Schwartz information criteria, and thus it is preferred.
Table 3 illustrates the AR(1) estimates for the central and northern regions. Results underline that the overall fit of the models is good. Furthermore, the variables that are statistically significant in explaining the phenomena investigated with the OLS approach maintain their explanatory capacity, even if with a different intensity for the elasticity and the p-value.
Table 3

First-order autoregressive model of the determinants of child admissions to NRUs for central and northern regions (2003–2009)

 

Central region

Northern region

Equation 2

Equation 4

Equation 2

Equation 4

C(1)

−0.111685*** (0.0000)

−0.181426** (0.0409)

0.001063 (0.9770)

0.035528 (0.7078)

D(HSI)

0.348595*** (0.0000)

0.343093*** (0.0001)

0.348901*** (0.0016)

0.388090*** (0.0019)

D(PDU)

0.411705** (0.0111)

 

0.246045 (0.3390)

 

D(PNF)

10.55931*** (0.0000)

10.67533*** (0.0000)

−1.724021 (0.5576)

−1.314377 (0.6808)

D(NC)

1.054347*** (0.0000)

1.037843*** (0.0000)

1.258085*** (0.0000)

1.274620*** (0.0000)

D(ICB)

−0.042189 (0.2888)

−0.031353 (0.4422)

−0.022244 (0.6952)

−0.036816 (0.5462)

D(URE)

−0.298273* (0.0742)

−0.236529 (0.1469)

−0.278352 (0.2537)

−0.364797 (0.1608)

D(TR)

 

4.062737 (0.3883)

 

−1.797514 (0.7414)

D(SF)

 

0.592547** (0.0366)

 

−0.018630 (0.9682)

D(CY)

 

−0.275397 (0.4194)

 

0.992628 (0.1281)

D(IR)

 

0.958149*** (0.0095)

 

0.089618 (0.8538)

ρ

−0.491126*** (0.0000)

−0.517699*** (0.0000)

−0.409000*** (0.0003)

−0.430300*** (0.0002)

R2 ADJ

0.879568

0.884376

0.811454

0.807887

F-statistic

85.51138 (0.000000)

62.95485 (0.000000)

50.80046 (0.000000)

35.06261 (0.000000)

Durbin-Watson

2.126284

2.206281

1.970946

1.961716

Inverted AR Roots

−.49

−.52

−.41

−.43

p-value in brackets. *Significant at p < 0.10. ** Significant at p < 0.05. ***Significant at p < 0.01

In all the models, the statistically significant first-order serial correlation coefficient indicates the operation of other factors which have an influence on the dependent variable, other than the regressors. They negatively affect child malnutrition and explain, in part, the decreasing trend in the number of children admitted to NRUs in the two regions, as can be seen from Fig. 3.

Several reasons are at the basis of this reduction. Among them, there might be a shift toward OTPs and the surplus production of maize recently experienced by Malawi due to good rainfall and to the Government policy on an input-support program, introduced in 2005/06. This latter has benefited poor rural farmers – 1.7 million according to the WFP (2010) estimates – who have been able to buy cheaper fertilizer and seeds with a corresponding positive effect on the level of food production (Dorward and Chirwa 2011).

Discussion

Apart from the expected statistically significant, strong and positive relationship between the number of NRUs and the impact variable, the models estimated show important differences at the regional level, which reflect the specific geographic and socio-economic conditions of each area.

In the northern region, the only additional variable with a statistically significant explanatory capacity of the phenomena investigated is the proxy of households’ market dependence for food purchase (Table 3). Child admissions to NRUs increase during the lean season, when household welfare deteriorates. In this region, poor households show a relatively higher dependence on the market for their food purchases than other household categories; a reliance that is particularly accentuated in northern Mzimba where they are able to eat food they themselves produce for only 6 months of the year. Furthermore, informal cross-border trade does not represent an adequate source of additional food and income; the commercial isolation of the region reduces this possibility to very low levels (Malawi National Vulnerability Assessment Committee 2005a).

Apart from the poor, the other household categories are less dependent on the market for their food security. During periods of downturns in agriculture, they are able to maintain their food intake above the minimum levels from livestock sales (Malawi National Vulnerability Assessment Committee 2005b). However, in normal years, in the two northern districts considered, Karonga and Mzimba, the greater proportion of non-poor households are able to produce maize in excess of their food requirements. Because of this, their income from casual labor, the main additional income source for the poor, tends to be especially low. According to the WFP (2010), it is not even enough to allow poor households to purchase half of their food energy needs.

As illustrated in Table 3, in the central region the model explaining changes in the number of children admitted to NRUs becomes more complex. First of all, the theoretical prescription concerning the role of price and income in affecting malnutrition is confirmed. An increase in maize price is a critical factor for SAM in children. In this region there is a major urban center, Lilongwe, and it is the main tobacco and maize production area. However, maize surplus is almost all grown by the better-off households. On the contrary, in the poor households there is a food production deficit; their food security is strongly dependent on income earned as ganyu on the local agricultural labor market. This reliance on the local agricultural market normally lasts for approximately 6 months a year and results in low pay rates due to the excess of labor supply. Thus, poor households are doubly vulnerable to bad agricultural years: on the side of food prices and on the side of income. When maize production falls short, food has to be imported from other markets with an increase in price and, at the same time, income from agricultural casual labor is reduced (Malawi National Vulnerability Assessment Committee 2005a).

The estimation of Eq. 4 allows a better understanding of the role of food prices in affecting the impact variable in the central region. In fact, SAM in children is particularly sensitive to the seasonal and irregular component of maize prices. The latter aspect is not surprising in an area prone to drought and rainfall irregularities and where most of the country’s poor are found. This underlines the vulnerability of a country where food consumption is almost all provided by home production, irrigation is lacking, and the majority of landholdings are too small to cover household needs.

The amount of rain and its variability have a significant impact on crop yield and a weather shock, even of minor intensity, can have dramatic implications on the incidence and severity of food shortages, and in turn on child malnutrition (Government of Malawi and World Bank 2006).

In the central region, the change in the number of children admitted to NRUs is also significantly and positively correlated to non-food prices. Lewin and Fischer (2010) argue that an increase in non-food prices reduces household purchasing power, causing further deterioration in household food security. According to our results, in northern Malawi, this effect seems to be passed on to children within the household, making their nutritional status more precarious. It should also be noted in the central region the expenditure pattern of poor households has a distinctive feature. As reported by the Malawi National Vulnerability Assessment Committee (2005a, 2003) casual labor in the tobacco sector is an important source of their income. This cash crop is more drought resistant than maize and, thus, income from this sector should reduce the vulnerability of poor households in a bad agricultural year if adopted for building up stocks or saved for future food purchases. On the contrary, this income is normally spent on non-food items as it is received, regardless of the nutritional status of the household members.

A further observation regards the constant term: only in this region is it statistically significant.

Setting all the explanatory variables equal to zero, the number of children admitted to NRUs is expected to be reduced. As previously mentioned, this fact might be the result of specific policies, such as the agricultural input support program, and food security interventions. However, informal sources suggested that the negative sign of the intercept might also be related to negative factors, among which is a progressive reduction in the support for NRUs by the government and donors.

A final consideration regards the change in urea prices. The literature suggests the importance of this factor for explaining nutritional status in Malawi. On the contrary, in our analysis it is weakly significant only in the central region where there is the highest percentage of smallholder farmers purchasing fertilizers (Dorward and Chirwa 2011). The second unpredicted aspect has to do with the negative sign estimated for this explanatory variable. Several reasons might explain this result. The most important seems to be related to the variable adopted in the analysis, the full market urea price. This is generally paid by commercial farmers while, as previously mentioned, smallholders benefit from subsidized prices. Thus, this variable might not properly represent a proxy for agricultural input costs, particularly at the level of poor households.

Table 2 illustrates results for the southern region. As in the other areas taken into consideration, the proxy of households’ market dependence for food purchases has a strong explanatory capacity. However, the elasticity estimated for southern Malawi is the lowest. This result might be related to the fact that households’ livelihoods benefit from the high incidence of the commercial sector; in fact, people can obtain better prices from selling their crops and have more employment opportunities, particularly on the casual labor market (Malawi National Vulnerability Assessment Committee 2005a). Furthermore, this population is less dependent upon maize as their main staple food due to the wider range of crops grown. This feature might also be at the basis of the fact that a change in domestic maize prices does not explain the phenomena investigated. However, Eq. 4 allows us to better qualify this aspect. The change in the number of children admitted to NRUs is positively and more than proportionally correlated with the cyclical component of maize price, which, as previously mentioned, is strongly dependent on policy shocks.

Another feature that distinguishes this region is the significant and positive elasticity of the impact variable to a variation in net cross-border imports in maize. As already underlined, Malawi shares its southern border with Mozambique and during the lean season the population intensifies maize imports from that country as a means for improving their food availability and income earnings.

Conclusions

This paper explores, with reference to the Malawian regions and the time period from 2003 to 2009, the short-term relationship between SAM in children and a set of explanatory variables, selected with reference to the concept of food security provided by the 1996 World Food Summit. In this context, the study suggests some areas in which policy research should be intensified. One lesson from this analysis is that, in Malawi, child malnutrition is a chronic problem fuelled by transitory food insecurity, including seasonal and temporary features. The seasonal component of SAM in children is related to factors affecting the market dependence of households on food purchases during the lean season. This is a common element in all the regions investigated, which is exacerbated by regional-specific features suggesting a territorial aspect of food security.

The impact of maize prices is an example. It is statistically significant in the central and southern regions but with important differences. In the central region, the seasonal and irregular maize price components contribute to the explanation of child malnutrition, underlining the key role played by agriculture as a source of food and income for food insecure households. Thus, policies aiming at intensifying agricultural production and raising land yields, especially among smallholders, may help reduce SAM in children in that area. In this respect the agricultural input subsidy program introduced by the government of Malawi represents a possible example (Cornia et al. 2012).

Finding ways to improve agricultural diversification, particularly with drought resistant crops, in combination with lower transaction costs, better infrastructure, access to irrigation and market efficiency should also be central to a credible food security strategy. In Malawi there has always been an over-emphasis on maize production for food security on the part of the government and donors. The support given to that sector makes it rational for farmers to produce maize; it is considered a form of protection for their households. However, the protection offered by the “maize economy” has not been sufficient to last throughout the year, and this then becomes one of the sources of poverty and malnutrition.

Also in southern Malawi, the effect of the above mentioned economic food access problem on SAM in children is made more severe by changes in maize prices and more precisely its cyclical component. As previously underlined, this is first of all an expression of policy interventions that have accentuated maize supply shortages on the domestic market. In this respect, lapses in the government’s early warning systems, distortions in domestic markets, and mismanagement of food reserves should be kept under control (Chirwa 2009; Minot 2010; Rubey 2003; Jayne et al. 2008).

The role of maize prices in affecting SAM in children as suggested by this paper contradicts the conclusions of Ecker and Qaim (2010). Referring to their empirical evidence, they argue that rising income might contribute to higher and more balanced consumption, instead of making the price of staple foods cheaper. On the contrary, according to our study, improving child nutritional status also requires a specific attention to maize price stabilization, particularly in certain areas of the country.

Turning to the proxy of household market dependence on food purchases as adopted in this paper, it is interesting to notice that Lewin and Fischer (2010), on the basis of a similar dummy variable, find that households are less food insecure during the lean season. The relationship is interpreted as a result of the effect of food assistance provided by the government. If this explanation is true, the opposite sign of the correlation pointed out by our analysis in all the equations estimated might be attributed to the intra-household food distribution mechanisms that penalize children. Poor child feeding and care practices, low education levels, lack of knowledge in food processing and utilization, lack of access to quality health care services, diseases (particularly HIV/AIDS), and adverse cultural beliefs represent important factors that explain the exclusion of children from consumption of high nutritive value foods (Government of Malawi, Ministry of Agriculture and Food Security 2010). The food utilization problem has clearly emerged in the central region where the so called “Centre paradox” finds confirmation: average higher caloric availability and lower levels of poverty coexist with higher levels of child malnutrition (Government of Malawi and World Bank 2006).

A final observation refers to the food security policy approach adopted in Malawi that has traditionally been aimed at self-sufficiency (Harrigan 2008). This paper suggests that behind an increase in SAM in children there is a composite problem of food access, availability and utilization, framed by a highly vulnerable natural and socio-economic environment. As a consequence, the concept of food security should be at the heart of a credible policy-making process aimed at improving child nutritional status in Malawi.

Acknowledgments

The author would like to express her gratitude to Giovanni Andrea Cornia (University of Florence) for the valuable suggestions received during the research project, supported by the UNDP, on Food price volatility over the last decade in Niger and Malawi: extent, sources and impact on child malnutrition, on which this paper is based. The study also benefited from the comments and information provided by many other people; among them, further thanks go to Alexandre Castellano (COOPI, Malawi); Paola Fava (COOPI); James Bwirani (FEWSNet, Malawi); Olex Kamowa (FEWSNet, Malawi); Irene Kumwenda (Masters student, Malawi); Fydess Mkomba (Centre For Social Concern, Malawi); Antonio Lijoi (University of Pavia), Carlo Bernini Carri (University of Pavia) and anonymous referees. The contents of the present paper reflect only the opinion of the author.

Copyright information

© Springer Science+Business Media Dordrecht and International Society for Plant Pathology 2012