Jump to ContentJump to Main Navigation
Agricultural Input SubsidiesThe Recent Malawi Experience$

Ephraim Chirwa and Andrew Dorward

Print publication date: 2013

Print ISBN-13: 9780199683529

Published to Oxford Scholarship Online: January 2014

DOI: 10.1093/acprof:oso/9780199683529.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 20 November 2019

Economy-wide effects of input subsidies

Economy-wide effects of input subsidies

Chapter:
(p.143) 7 Economy-wide effects of input subsidies
Source:
Agricultural Input Subsidies
Author(s):

Ephraim Chirwa

Andrew Dorward

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780199683529.003.0007

Abstract and Keywords

A large national programme of agricultural input subsidies focusing on staple crops should have important wider market effects in addition to direct beneficiary impacts. This chapter reviews the evidence on the economy-wide impacts of the Malawi subsidy programme, using a variety of sources and analytical methods. While there has been macroeconomic stability during the implementation of the subsidy programme, public debts and fiscal deficits have not been favourable. There has however been strong agricultural GDP growth. Very high maize prices in 2008/9 are not what might be expected, and inter-seasonal maize price variability has been high. Nevertheless there is strong evidence of rural wages rising faster than maize prices, leading to increases in real incomes. While qualitative studies suggest overall improvements in peoples’ welfare there are puzzling results with little reported change in national and rural poverty rates and in some measures of child malnutrition from 2004/5 to 2009/10.

Keywords:   input subsidies, maize prices, economy-wide impacts, poverty, Malawi

7.1. Introduction

Examination of economic theory on subsidy impacts in Chapter 2 stressed the importance of considering wider market effects of large-scale input subsidy programmes for staple crops. This was set out in the conceptual frameworks at the end of Chapter 2, in discussion of possible impacts on Malawian farmers’ livelihoods in Section 4.4 of Chapter 4, and in the Introduction to Part II. Nevertheless, it was noted in Chapter 3 that these impacts have been largely ignored in consideration of both the objectives and impacts of recent subsidy programmes in Africa.

This chapter follows on from discussion of direct subsidy impacts in Chapter 6 and considers evidence of the economy-wide impacts of the FISP. It builds on observation of a range of general beneficial changes summarized in Table 6.3 (increases in maize production, net crop income, food consumption, school enrolment, child health, subjective well-being and physical assets, and decreases in ganyu labour supply). The chapter is organized into six sections after this introduction. The next section outlines different sources of evidence and analytical methods considered in subsequent sections, which consider in turn the main potential economy-wide impacts identified in Figure II.1. We then consider in turn the evidence on subsidy programme impacts on macro-economic stability (in Section 7.3), on maize trade (in Section 7.4), on maize prices and wages (in Section 7.5), and on other variables (in Section 7.6). Section 7.7 then provides concluding remarks.

7.2. Sources of evidence and analytical methods

There are two main tasks in determining both direct and economy-wide impacts of the subsidy programme: first the identification of changes associated with the programme, and second attributing or determining the causal (p.144) dependence of these changes to or on the programme. For direct impacts, there are often opportunities to compare households or areas ‘with and without’ and ‘before and after’ subsidy receipt. A variety of experimental and analytical approaches can then be used to address problems of selection bias—for example, using panel data and fixed effects analytical methods as outlined in Section 6.2 in the previous chapter.

Such approaches are not available when considering economy-wide impacts of a large-scale programme such as the Malawi Farm Input Subsidy Programme, as economy-wide effects spread through the economy ‘with and without’ comparisons cannot be made. We therefore draw on four different approaches when examining possible impacts of the programme:

  • examination of changes in variables where an economy-wide subsidy impact is expected, with consideration of likely changes that might have been expected in the absence of the programme;

  • regression analysis (using panel data) relating the extent and nature of subsidy access in an area to area wide variables;

  • computable general equilibrium (CGE) models; and

  • a partial equilibrium model linking different household livelihood models to maize and labour markets.

These approaches all have their own strengths and weaknesses, and there are different challenges in their implementation and application in different contexts. We discuss each briefly.

Examination of changes is a starting point for any analysis of possible economy-wide programme impacts. Apart from difficulties in obtaining relevant and reliable data on some variables, the major issue here is the problem of attribution of change to a particular cause—or indeed the recognition that where there has been no change this could be the result of the subsidy programme counteracting the effect of some other change. We identify three broad changes that roughly coincide with the implementation of the subsidy programme from 2005/6 to 2011/12 and that would be expected to have widespread benefits: improved macro-economic management, good tobacco prices, and generally favourable rainfall—although there is some variation in each of these within the period. Thus, there were major improvements in macro-economic management from 2004 onwards, which are perhaps best demonstrated by the fall in interest rates, debt, and debt servicing from 2004/5 to 2007/8, but resurgent debt in 2008/9 and subsequent raised interest rates (National Statistical Office, 2011).1 Rainfall was also generally (p.145) favourable over the period 2005/6 to 2010/11. Although there were local incidences of flooding and dry spells that caused local production shortfalls, these did not lead to widespread lowering of yields and production. There were more widespread dry spells in 2011/12: these did not lead to a major reduction in national production estimates by the MoAFS, but substantial areas were affected, leading to estimates of 1.63 million people affected by serious food shortages (FEWS NET, 2012). Similarly, burley tobacco prices rose in 2007 with high prices continuing in 2008 and 2009 before falling back in 2011 and recovering somewhat in 2012. Total burley tobacco incomes follow a similar pattern though sometimes with a lag as farmers expand (contract) tobacco areas following high (low) prices (see Figure 7.1).

Improved macro-economic management and good rainfall are probably best considered as necessary but not sufficient conditions for rapid positive change in the rural economy and in rural people’s livelihoods, and it is therefore necessary to look elsewhere for drivers of such change. World Bank (2010a) argues that Malawi’s growth is export-led with weak linkages between maize production and economic growth due to limited sales and thus limited multipliers. This is supported by high correlation between annual per capita export volumes and per capita GDP from 1960 to 2007, and low correlation between annual per capita maize production and per capita GDP from 1990 to 2005. This argument places considerable emphasis on the role of tobacco in driving broad-based growth. It ignores,

Economy-wide effects of input subsidies

Figure 7.1. Burley tobacco prices and sales values, 2000–11

Source: Tobacco Control Commission.

(p.146) however, the importance of low maize prices and high maize productivity to real incomes of large numbers of net food-buying farmers. It may well be the case that in the past there has been very limited correlation between per capita maize production and per capita GDP. However, low growth in maize productivity may be one of the causes of Malawi’s low GDP growth, with exports the major driver of the limited growth that there has been. In teasing out the effects of high tobacco prices from those of the subsidy programme it may be instructive to note that tobacco prices started to rise with the 2007 harvest, whereas the subsidy programme affected the 2006 harvest. This suggests that changes prior to the middle of 2007 are unlikely to have been driven by improved tobacco prices.

Regression analysis relating area wide subsidy access and other variables is the second approach that may be used to investigate some of the subsidy programme’s economy-wide impacts. This approach only works for changes in relatively local markets that are not well integrated. There are a number of studies that show that Malawian maize markets are relatively well integrated (Chirwa and Zakeyo, 2006; Meyer, 2008). Less is known about the integration of labour markets, but one would expect integration to be low in markets for ganyu labour (with intermittent demand, the importance of social relations, and relatively immobile suppliers of ganyu). Ricker-Gilbert (2011) exploits this to investigate the impact of subsidy receipt on wages across different communities. This approach has not, to our knowledge, been applied to investigate other possible economy-wide subsidy impacts, but it would appear to have potential application in studying impacts of subsidy-induced labour market changes, but not subsidy-induced maize market changes. There are challenges to the collection of reliable data on rural wages. However, these are reduced where panel data are employed since analysis of changes in wages among respondents between surveys demands consistency in wage measurement within responses by the same households at different times, rather than consistency in wage measurement across households (where standardized definitions of tasks are very problematic).

Computable general equilibrium (CGE) models should in principle be the best approach for investigating economy-wide programme impacts. Apart from the costs involved in developing such models, there are two main difficulties in the use of this approach: obtaining reliable data on the wide range of variables and relationships that make up such models, and constructing models to properly capture key features of the livelihoods, markets, and economies modelled. These challenges are also discussed in Chapter 9, where possible uses of CGE models in benefit–cost analysis are examined. We note here, however, that more conventional CGE models, including those that have been developed for Malawi, may be considered ‘top down’ models in the sense that in broad terms they start from analysis of national accounts (p.147) to develop a Social Accounting Matrix (SAM) for the economy as a whole, and then work down to model in fairly broad terms the main sectors in the national accounts (Dorward et al., 2004b; Benin et al., 2008; Douilleta et al., 2012). This approach does not, however, adequately describe the seasonal constraints and market failures affecting poorer households and hence the potential role of the subsidy in addressing the low productivity maize production trap, as argued earlier in Chapter 4 (see also Dorward et al., 2004b). An alternative approach, the ‘Local Economy-wide Impact Evaluation’ or LEWIE model, is currently being developed to address this problem (Taylor, 2012) with an illustrative application to analysis of the Malawi subsidy programme (Filipski and Taylor, 2011). Although this method is expected to provide valuable insights into economy-wide impacts of the subsidy in the future, existing information from CGE models is limited. Nevertheless, the use by Buffie and Atolia (2009) of a ‘standard’ CGE model to investigate FISP impacts provides useful insights given its assumptions, and these insights are made more useful by Filipski and Taylor (2011) in their examination of potential changes in analysis when some of these assumptions are changed.

Partial equilibrium models that link different household livelihood models to maize and labour markets are the fourth broad approach used in investigating economy-wide subsidy impacts. These suffer from similar but narrower challenges as CGE models as regards the need for reliable data on a wide range of variables and relationships. We use an Informal Rural Economy’ or IRE model (described briefly in Chapter 1 and more fully in Dorward and Chirwa, 2012b) to explore both the direct impacts of the programme (as described in Chapter 6) and economy-wide impacts which arise from impacts on maize and labour supply and demand and hence on maize prices and wages.

7.3. Macro-economic environment and role of input subsidies

The contribution of the subsidy programme to economic growth and macro-economic aggregates is difficult to disentangle as there are so many factors that affect macro-economic stability. However, we compare macro-economic developments between 2000 and 2005 (the pre-FISP period) with those between 2006 and 2011 (the FISP period).

The macro-economic environment since the introduction of the Farm Input Subsidy Programme has remained relatively stable with high growth rates and low inflation (Chirwa et al., 2011d). Table 7.1 shows the average performance of the economy between 2000 and 2011. With respect to growth in gross domestic product, the official figures record average growth between 2000 and 2005 (pre-FISP period) of 1.7% per annum compared to 7.8% per annum in the subsidy period, with the agricultural sector an important driver (p.148)

Table 7.1. Economic growth performance, 2000–10 (% per annum)

Indicator

2000–5

2006 –11

2005

2006

2007

2008

2009

2010

Real GDP growth

1.7

7.8

3.3

6.7

8.6

9.7

7.7

6.7

Real agricultural growth

2.1

10.3

7.8

12.3

12.3

11.8

10.4

6.6

Manufacturing growth

0.1

5.4

4.3

4.6

3.6

9.9

4.8

4.2

of recent growth: in the pre-FISP period, the agricultural sector GDP only grew by 2.1% per annum compared to 10.3% per annum during the subsidy period. Allowance for bounce-back in 2006 following the bad rains in the previous year lowers this figure by a little over 1%.

However data on GDP growth in the FISP period need to be interpreted cautiously for two reasons. First, annual crop production estimates are used in the calculation of GDP. As was discussed earlier in Chapter 6, there are concerns that crop production estimates in the FISP period may be somewhat over-estimated. If this is the case then GDP will also be over-estimated, given the close relation between the performance of the agricultural sector and the economy. Determining the possible scale of such over-estimates is difficult but it is unlikely to lead to an inflation of GDP by more than 1.5% over the FISP period as a whole. This could then be a small contributor to higher GDP growth in the FISP period. Possible distortions to the GDP deflator as a result of under-estimates of CPI are, however, possibly much larger. National Statistical Office (2012) use an unexplained revised estimate of inflation of just under 130% between the 2003/4 and 2010/11 Integrated Household Surveys. Comparison of the published CPIs for 2003 and 2010, 154.3 and 319.8, (Reserve Bank of Malawi, 2012) suggests inflation of approximately 107%, in which case the real GDP in 2011 might be some 17% lower than indicated by current published estimates. This could reduce annual GDP growth estimates by some 2.5 percentage points.2 If these calculations are broadly correct, they have very serious implications for Malawi’s growth record since 2005. Nevertheless, agricultural sector performance in the FISP period would still be impressive, averaging a little under 8%—or a little under 7% if the 2006 recovery from bad 2005 rains is factored in. (p.149)

The agricultural sector has therefore grown consistently during the subsidy period, and this has helped overall economic growth. Growth rates in both gross domestic product and in agricultural output may be partly attributed to the subsidy programme, although as discussed earlier they may also be partly attributed to high tobacco prices and to improved macro-economic management, and have been aided by the good rains experienced in most seasons since 2005/6. The dry spell that hit some parts of the country in 2009 and lower tobacco prices and higher interest rates since then have contributed to the marked decline in agricultural growth after 2009.

There have also been fiscal implications of the subsidy programme, particularly as much of the financing comes from domestic revenues. As detailed in Chapter 5, the high costs of the programme have increased resource allocations to the agricultural sector and the subsidy accounts for a significant proportion of the national budget. Chirwa et al. (2008) note that due to the subsidy programme, Malawi became one of the first African countries to achieve a 10% budgetary allocation to the agricultural sector, in accordance with the African Union (AU) and New Partnership for African Development (NEPAD) target for stimulation of agricultural growth. However, this increased spending on the agricultural sector has come at the expense of increasing budget deficits and increasing debts. Figure 7.2 shows official figures for trends in fiscal deficit/GDP ratio and public debt/GDP ratios from 1999 to 2010. The deficit as a proportion of gross domestic product has worsened during the period of implementation of the subsidy programme. The deficit after grants increased substantially in 2008 and 2009, in line with the very high expenditures on the

Economy-wide effects of input subsidies

Figure 7.2. Fiscal deficit/GDP ratio, 1999–2010

Source: Computed by authors based on Reserve Bank Database.

(p.150) FISP in 2008/9 reported in Chapter 5 (with other expenditures related to the 2009 elections) before a surplus was registered in 2010.

There has also been some worsening of domestic debt, which increased from 8.2% in 2006 to 15.7% in 2010. The rising costs of the subsidy programme from 2006/7 to 2008/9 and, as noted earlier in Chapter 5, its general failure to keep within budget (except for 2009/10) may have contributed to these trends—together with other pressures on public expenditure associated with the 2009 election. However it is also important to consider the budgetary implications of possible maize import and humanitarian distribution costs for government if the absence of the subsidy would have led to lower and possibly insufficient national maize production.

7.4. Maize exports and imports

There are difficulties in obtaining good data on international maize trade flows in Malawi, partly due to the restrictions imposed on maize exports and imports and due to informal and hence unreported trade across borders with Mozambique, Zambia, and Tanzania. There are conflicting estimates of trade across different reports (for example, Jayne et al., 2010; FAOStat; and FEWS NET reports). In broad terms, however:

  • Malawi is more commonly a maize importer than exporter, with relatively small annual informal imports and no formal imports in most years, unless there is a national food shortage demanding large-scale formal imports and increasing informal imports.

  • Informal imports have been fairly constant during the subsidy years, after major imports in 2006 following the poor 2004/5 harvest.

  • Following the large estimated harvest in 2007, the government allowed formal exports of maize amounting to around 400,000 tonnes through the granting of licenses for export, particularly to Zimbabwe (licenses were granted to exporters that had unsold stock from the 2006 harvest, with permission for ADMARC to export up to 100,000 metric tons of maize, with the rest exported by private exporters).

  • This was followed by a dramatic maize price surge, peaking at 90MK/kg and formal imports from South Africa of around 40,000 MT in 2008/9 (Jayne et al., 2010).

  • According to FEWS NET (2011) the government also issued export licences from July 2010 to August 2011, allowing an increased export of maize mainly to Kenya and Zimbabwe, although it has been difficult to track records of these exports.

(p.151)

This suggests that the overall domestic maize supply has improved except in 2008/9 (following the 2007/8 harvest). However these figures need to be interpreted together with information on domestic maize prices, to which we now turn.

7.5. Impacts on maize prices and rural wages

Maize is the main staple food in Malawi. The price of maize has the largest weight in the food price index of the consumer price index. Increased maize production as a result of the subsidy programme should push maize prices down and promote more general price stability and benefit net maize buyers among both beneficiaries and non-beneficiaries of the subsidy programme. Changes in maize prices are therefore a critical determinant of real wages and consideration of changes in real wages thus requires consideration of changes in both maize prices and nominal wage rates. We examine these in turn.

7.5.1. Maize prices

One of the expected benefits of the farm input subsidy is to reduce both the price of maize relative to rural incomes (Dorward, 2013) and its intra- and inter-seasonal price variability. Figure 7.3 shows nominal and real maize prices between 2001 and 2011 in Malawi Kwacha and US dollars per kilogram. Both inter-seasonal and intra-seasonal variability of prices are evident, and prices have not behaved as might be expected. The FISP period has experienced higher average and peak nominal maize prices than the pre-FISP period

Economy-wide effects of input subsidies

Figure 7.3. Nominal and real maize prices in Malawi, 2001–11

Source: Computed by authors based on MoAFS Price Monitoring Data.

(p.152) with prices in the FISP period reaching a peak of 70MK per kilogram (US$.50 per kilogram) in January/February 2009 compared to a peak of 50MK per kilogram (US$0.39 per kilogram) in February 2006. As regards real prices (deflated by the CPI), averages in the pre-FISP and FISP period are not substantially different: as the moving averages show they are a little higher in the FISP period when measured in Malawi Kwacha and a little lower when measured in US$.

Between 2001 and 2011 there were three surges in the price of maize: in 2001/2, 2005/6, and 2008/9. Chirwa (2009) notes that price surges in 2001/2 and 2005/6 are mainly explained by reductions in maize production owing to poor weather conditions (with heavy rains in March and dry spells and floods in some areas exacerbated by low input uptake in 2000/1; and late distribution of inputs and poor rains in many areas in 2004/5).

The surge in maize prices in the 2008/9 market season should not be attributable to such supply-side issues because of relatively good rains and improved access to subsidized seeds and fertilizer. High prices in 2008/9 (and other market seasons) also raise questions about MoAFS’ high national maize production estimates. Figure 7.4 shows 1993/4 to 2010/11 maize prices (average annual prices from MoAFS market surveys, in US$) against estimated quantity consumed per capita, calculated from Ministry of Agriculture and Food Security crop production estimates, census data, and exports and import estimates compiled from various sources.3 This perspective on prices and supply estimates draws attention to an apparent shift in the relationship between net supply and prices from around the 2006/7 production season.

Chirwa (2009) suggests several reasons for the high maize prices in 2008/9 despite estimated high maize production levels in the 2007/8 production season.

  1. 1. Poor quality of information about domestic supply from the government, both in terms of domestic production and stocks in reserves, created speculative behaviour. Even with record reported maize production levels, maize prices remained high, creating uncertainty about domestic supply, with asymmetric information among different agents.

  2. 2. The unsatisfied contract to export to Zimbabwe from the 2006/7 harvest also fuelled speculation that there was a domestic maize shortage. By February 2008, only 302,000 metric tons of maize had been exported to Zimbabwe by the private sector through the National Food Reserve Agency (NFRA), and Malawi failed to satisfy the contract of maize exports to Zimbabwe (FEWS NET, 2008). In addition, there was speculation that (p.153)

    Economy-wide effects of input subsidies

    Figure 7.4. Maize prices and estimated quantity consumed per capita from 1993/4 to 2010/11 production seasons7

    Source: Adapted and updated from Dorward and Chirwa, (2011c).

    exports to Zimbabwe would continue from the 2007/8 harvest, which was also a surplus year according to the MoAFS crop estimates. The private sector was having difficulty in sourcing maize from the market and this sent signals that there were supply shortages and prices began to increase substantially.

  3. 3. The behaviour of state agencies, ADMARC and the National Food Reserve Agency (NFRA), who were offering higher purchase prices to farmers than the private sector, and the government’s imposition of a ban on private trader purchases, also appeared to signal domestic supply shortages.

  4. 4. High international maize prices, amid continued reported surplus maize production, created speculation that the export market for maize would be lucrative, leading to stockpiling and purchasing maize from farmers at higher prices by the private sector.

  5. 5. A further possible cause of tighter markets could be government purchases of maize for stockpiling of the national grain reserve following the construction of a 60,000 MT new storage capacity.

Table 7.2 shows average nominal maize prices in Malawi and major cities between 2001 and 2011, and suggests that during the FISP period the average national prices and the prices of maize in the major cities have almost (p.154)

Table 7.2. Average nominal maize prices, 2001–11

Malawi Kwacha per kilogram

US Dollars per kilogram

Period

National

Mzuzu

Lilongwe

Blantyre

National

Mzuzu

Lilongwe

Blantyre

Pre-FISP

18.12

(8.84)

17.32

(8.48)

18.43

(8.66)

19.16

(10.03)

0.19

(0.09)

0.18

(0.07)

0.19

(0.09)

0.2

(0.09)

Post-FISP

33.53

(13.83)

35.13

(12.12)

35.6

(14.07)

38.02

(19.56)

0.23

(0.1)

0.24

(0.09)

0.25

(0.1)

0.26

(0.14)

Note: Figures in parenthesis are standard deviations in monthly prices.

Source: Computed by authors based on MoAFS Price Monitoring Data.

doubled in Malawi Kwacha terms. Monthly variability of maize prices (measured in standard deviations) has also increased in the FISP period, although the coefficient of variation has fallen slightly. Prices and price variability have also increased in US$ terms, though not as dramatically. It appears therefore that the subsidy programme has not significantly reduced either prices or food price risks.

It seems clear then that maize prices have not fallen in real or nominal terms over the period of FISP implementation. However, this does not necessarily mean that the FISP has not exerted downward pressure on maize prices but rather that other pressures pushed prices up in 2008/9 (as discussed above). In the absence of such pressures maize prices did fall and remain low throughout the 2006/7 marketing season, following the introduction of FISP in 2005/6 and before government interventions tightened the market after the 2007 harvest. Prices were also low throughout the 2010/11 market season, prior to the macro-economic problems that surfaced in mid 2011.

CGE and partial equilibrium models also suggest that FISP should lead to maize price falls in the absence of interventions or macro-economic or other changes affecting maize prices. Simulations by Buffie and Atolia (2009) estimate long-run falls in domestic food prices of 2% to 5%, with higher short-run falls of up to 9%. Filipski and Taylor (2011) estimate that the 2005/6 subsidy would have led to a 5% increase in maize production in a model that assumes perfectly competitive markets (they also unfortunately assume that maize prices are determined exogenously on the world market and therefore do not estimate any price changes). The introduction of credit constraints on input purchases (one part of the low maize productivity trap discussed earlier in Chapter 4) reduces the overall agricultural production impacts of the FISP to a little over 2% (specific impacts on maize production are not stated) but an alternative specification that introduces unemployment and efficiency wages results in the FISP increasing agricultural production by over 13%. The Informal Rural Economy partial equilibrium models also give (p.155) consistent estimates of increases in maize production of 10% to 20% across all households with the FISP, although direct beneficiary impacts tend to be higher (Dorward and Chirwa, 2012b). Aggregate production increases lead to lower maize prices, with median falls across different years varying from 8% to 40% under different scenarios. Given the clear endogeneity of Malawian maize prices, the large maize production impacts of the FISP in these simulations should have substantial price impacts.

It is, therefore, surprising that falls in real maize prices are not observed, despite the upward pressures on maize prices discussed above. This discussion, however, has examined only nominal prices and real prices deflated by the CPI. As noted by Dorward (2011) and Dorward (2013), real consumer prices are often best examined in relation to incomes, and in this regard there is evidence that real maize prices have fallen relative to wages. We therefore turn now to consider evidence of changes in nominal wage rates.

7.5.2. Wages

One of the expected economy-wide benefits of such a large-scale input subsidy programme is its influence on rural wages relative to maize prices. As outlined in Chapter 4, rural wages may increase due to the low supply of casual labour and high demand for labour among labour-hiring farming households. With low and stable maize prices, the increase in rural wages should increase real incomes of poor households. We discussed maize price movements in Malawi before and after the introduction of the FISP in Section 7.5.1. We now consider wage rates in both nominal terms and relative to maize prices.

Despite difficulties in obtaining reliable data on rural wage rates, there is evidence of rising wage rates in rural areas following the introduction of FISP. School of Oriental and African Studies et al. (2008) report survey data findings that median wage rates rose by 33% from 2005/6 (following a poor 2004/5 crop production season) to 2006/7 (following the 2005/6 subsidy and better rains) per day across the two years. Focus group discussions and key informant interviews also reported increases in wages and increased bargaining power for sellers of labour in a tighter market, with a shift from price setting by buyers of ganyu to sellers of ganyu and wage increases of 50% or more in some areas. School of Oriental and African Studies et al. (2008) also report anecdotal evidence of increased ganyu rates from commercial farmers who faced difficulties in obtaining hired labour without paying higher wage rates. FEWS NET (2007) also reported reduced search for employment by ganyu labour in November 2007 with increased scarcity and cost of hired labour reported by farmers wishing to hire labour.

As noted earlier in Chapter 6, Ricker-Gilbert (2011) reports contractions in labour supply over the period 2003/4 to 2008/9, with a large contraction in (p.156) labour supply before 2006/7 followed by a smaller expansion before 2008/9. He does not report changes in wage rates, although with relatively stable labour demand (and he only reports agricultural labour demand) higher real wage rates would be expected. Dorward et al. (2010a) state that focus group discussions and key informant interviews in 2008/9 reported continued improvements in ganyu wage rates and bargaining positions for sellers of ganyu (as in 2006/7). Household survey estimates of changes in nominal wage rates from 2006/7 to 2008/9 suggest increases of around 70%, so that over the period 2005/6 to 2008/9 they just about kept abreast with maize price increases, although this involved a dramatic increase in real wages from 2005/6 to 2006/7 and then a smaller decline from 2006/7 to 2008/9.

Chirwa et al. (2011d) find that nominal wages have continued to increase in rural areas since 2009, and when coupled with falling maize prices this implies increased real wages among households participating in the labour market. Figure 7.5 shows mean farm-gate purchase prices for maize, selling prices for tobacco prices, and ganyu wage rates between 2009 and 2011, as reported by survey respondents in 2011. With respect to maize prices, the overall prices at which households buy maize show strong seasonal effects in all districts (with January prices considerably higher than June prices) and falling prices from 2009/10 to 2010/11 (Figure 7.5a). This is consistent with maize market prices shown earlier in Figure 7.3. Tobacco prices generally fell slightly between 2009 and 2010 (Figure 7.5b), although in Blantyre and Zomba households reported improved tobacco prices. Again this is consistent with national data on tobacco prices in Figure 7.1. With respect to wages (Figure 7.5c), there is a steady increase in ganyu cropping season wages reported by households over the period January 2009 to January 2011, and these increases occurred in all the districts that were surveyed. Wage increases are broadly similar across all districts in percentage terms.

These wage rates and maize price developments were also widely reported in focus group discussions and in life histories of some of beneficiaries. In most life histories of beneficiaries, engaging in ganyu to earn income to purchase food is a common strategy among poor households and improvements in wages and reduction in maize prices made maize more affordable even for poor households. This is confirmed in Figure 7.5d which shows real increases in ganyu wages in terms of its maize grain purchasing power (variability in prices between areas may be due to bias and reporting inconsistencies, but data collection methods attempted to minimize changes in bias between surveys, and hence estimates of percentage change in wages within in each area should be and appear to be more consistent). Overall, the reported maize purchasing power of daily ganyu wages increased by 47% between January 2009 and January 2010, with the highest reported increase of 80% in Ntcheu district and lowest increase of 34% in Phalombe district. (p.157)

Economy-wide effects of input subsidies

Figure 7.5. Average farm-gate maize prices, tobacco prices, and ganyu wages, 2009–11

Source: Chirwa et al. (2011d), computed from FISS3survey data.

(p.158)

It appears then that real wages increased from 2005/6 to 2006/7, fell back somewhat from 2006/7 to 2008/9, and then increased again from 2009 to 2010. How far can this be attributed to the impact of FISP? The four approaches set out in Section 7.2 may be examined to investigate this, in addition to focus group discussions and key informants attributing increased wage rates to the impact of FISP in household surveys (School of Oriental and African Studies et al., 2008; Dorward et al., 2010a; Chirwa et al., 2011d).

Considering first general patterns of change, the two primary explanations for improved rural wages are improved tobacco prices and earnings, and the FISP. It appears that while nominal wage rates increased over the period 2005/6 to 2009/10, real wage rates improved from 2005/6 to 2006/7, fell back somewhat from 2006/7 to 2008/9, and then rose again from 2009 to 2010/11. Tobacco prices rose in 2006/7 and therefore cannot explain increases in nominal or real wage rates from 2005/6 to 2006/7. These increases might be explained by improved crop production in 2005/6 as compared with 2004/5 as a result of both the FISP and improved weather. Increased nominal wage rates from 2006/7 to 2008/9 would be consistent with the effects of increased maize production as a result of relatively good weather and the FISP and/or the effects of higher tobacco prices received from 2007 onwards. Increased nominal wage rates from 2008/9 to 2010/11 would be consistent with the effects of falling maize prices as a result of the removal of maize market distortions with continued high production (with good weather and the FISP) and with continuing relatively high tobacco prices (these remained above 2005/5 levels: as Figure 7.1 shows 2009 and 2010 prices were similar to 2007 prices, higher than 2006 prices, but lower than 2008 prices).

Further evidence of the impact of the FISP on wage rates is provided by Ricker-Gilbert (2011) in his use of regression analysis to examine the relationship between the amount of subsidy received in communities and median ganyu wage rates in these communities. He finds significant (p = 0.01) positive relationships between wages and subsidy receipt, commercial fertilizer price, hybrid maize price, tobacco price, and coefficient of variation of rainfall. He also finds significant (p = 0.01) negative relationships between wages and the standard deviation of subsidy receipt per household and long-run average rainfall. The estimated effects of subsidy receipt and of maize and tobacco prices are particularly interesting, with an average of 50 kg per household estimated to increase nominal wages by 10%, and an increase in 1 standard deviation decreasing nominal wages by 0.1%. An increase in tobacco prices by 50MK/kg (very approximately the increase in 2007, 2008, and 2009 above 2005/6 and earlier prices) is estimated to also raise nominal wages by 10%. An increase in maize prices of 30MK/kg would lead to an increase in nominal wages of a little over 4%. However, since a 30MK/kg increase in maize prices represents an increase of 100% or more over 2003/4 or 2006/7 maize prices, (p.159) and maize accounts for a significant proportion of expenditure of the poorer people who hire their labour out for ganyu work, the small increase in nominal wages will not be enough to prevent a fall in real wage rates.

Positive impacts of FISP are found in both the CGE models discussed earlier in the chapter. Buffie and Atolia (2009) find real unskilled wages rising by up to 5% immediately after the introduction of FISP, and then declining slightly or significantly depending in later years on the sources of programme finance. Filipski and Taylor (2011) also report FISP leading to an increase in rural wages with their ‘perfect competition’ model, although the scale of this increase is not reported. There is no specific mention of wage effects with the model allowing for liquidity constraints. Where they postulate efficiency wages, the impact of the subsidy is not on wage levels but on employment, and the impact of the subsidy on labour income is very large.

The importance of economy-wide effects on real wages is also supported by informal economy model simulations of the two largest livelihood zones (Shire Highlands and Kasungu–Lilongwe Plain) in Malawi between 2005/6 and 2010/11 (Dorward and Chirwa, 2012b). Small increases in wages and larger post-harvest falls in maize prices combine to give increases in the wage to maize price ratio, with average increases in the pre-harvest period of 19% and 5% for the Shire Highlands and Kasungu–Lilongwe Plain respectively, and 73% and 32% in the post-harvest period. These ‘basic scenario’ simulations are consistent with the earlier discussion of potential impacts in Section 4.4, with differences between the two zones explained by the greater proportion of poor households in the Shire Highlands livelihoods zone.

7.6. Other variables

The previous section considered different sources of information on possible subsidy impacts on maize prices and wages. This reflected both the importance of these impacts (indicated, for example, by their central position in the causal impact chains set out in Figures 2.2, 4.2, and II.1) and the availability of information and analysis on these impacts. There are, however, a range of other potential FISP impacts that are important for welfare and economic benefits, but where there is less evidence on impact. In this section we review changes in these variables over the period in which FISP has been implemented. Arriving at firm conclusions regarding change is difficult. A number of data sources are supported by anecdotal evidence in suggesting improvement on many variables over the period 2005 to early 2011, after which there were increasing political and economic difficulties associated with a number of President Mutharika’s policies. However the third Integrated Household Survey (IHS3, National Statistical Office, 2012), reporting on the situation (p.160)

Table 7.3. Household food consumption over the past 1 month, 2006/7–8/9 (%)

Season

Less than adequate

Just adequate

More than adequate

2006/7

38

51

10

2008/9

10

63

27

Note: Interviews with households were conducted in May and June, shortly after harvest.

in 2010/11, contains a number of statistics that present a much less positive picture. Finding consistency across these estimates and other data sources, including those on estimated FISP impacts, is problematic.

7.6.1. Food availability and consumption

National food production and per capita availability indices show increases in national food security in Malawi since the implementation of the FISP in 2005/6. As discussed earlier in this chapter, there are questions about the reliability of some of the national crop production estimates on which these are based (possible discrepancies between maize production estimates and prices were discussed earlier, and there are longstanding queries about inconsistencies between national crop estimates and survey estimates of root crop areas and production).4 Nevertheless, there is general agreement that the food security situation has improved in the country partly owing to incremental use of fertilizers and improved seeds provided under the subsidy programme and partly owing to the good weather conditions that Malawi has had during this period. These trends of national food security are consistent with household survey findings that show improved self-assessment of food security. Table 7.3 shows the proportion of households with different food consumption between the 2006/7 and 2008/9 seasons from a survey of beneficiaries and non-beneficiaries of FISP. Overall, there is a shift to more households reporting adequate and more than adequate food consumption.

However, National Statistical Office (2012) finds that 33% of households experienced situations on food insecurity, with 42% of the rural population being food insecure in 2010/11. Concerns also arise about the large numbers of people continuing to experience food insecurity and needing emergency humanitarian support. This appears to be particularly serious in 2012/13 with estimates that about 1.6 million people in 15 districts, mainly in southern Malawi, will be unable to meet adequate basic food requirements (FEWS NET, 2012). This is mainly due to prolonged dry spells and poor rains: similar situations in previous years have been associated with local droughts or floods and are arguably unrelated to the FISP. (p.161)

7.6.2. Incomes and poverty

Earlier discussion of FISP focus group discussion and key informant reports of FISP impacts on real wages implicitly or explicitly linked this to improved real incomes. Estimates of subsidy impacts on nominal wages (Ricker-Gilbert, 2011) when combined with estimates on maize production also suggest increases in real incomes for poorer non-beneficiaries as well as beneficiaries. There are explicit results from the different CGE models of Buffie and Atolia (2009) and Filipski and Taylor (2011), with Buffie and Atolia (2009) estimating short-term real income increases of the poor (many of whom may not receive subsidized inputs) rising by 4% to 8%, and Filipski and Taylor (2011) estimating that the introduction of liquidity and labour market imperfections into their model increases the income gains of non-beneficiaries from spillover effects. The IRE partial economy model of Dorward and Chirwa (2012b) also estimates substantial real income gains from wage and maize price change impacts, averaging 10% and 3% across all households in the ‘basic scenario’ in the Shire Highlands and Kasungu–Lilongwe Plains respectively. In this, poor non-beneficiary households gain most, non-poor beneficiaries, and poor non-beneficiaries gain direct and indirect benefits respectively, and non-poor non-beneficiaries generally lose from the indirect impacts through higher wages and low maize prices. In the Shire Highland Livelihood Zone (SHI), simulated ‘indirect gains’ to real incomes for target households are considerably higher than the direct gains from subsidy receipt (13% as compared with 7%). These indirect gains are higher than simulated for the Kasungu–Lilongwe Plain (KAS) because of the former’s high rates of poverty incidence, high land pressure, and larger numbers of poor people relying more on sales of ganyu labour and spending a higher proportion of their income on maize purchases.

If the farm input subsidy raises the income of the poor then it should also play a role in poverty reduction. Indeed evidence of the impacts of the FISP on the wage to maize price ratio and the impact of tobacco prices on wages (as reported earlier) should have led to falls in poverty from 2005/6 to 2010/11. Given the scale of the FISP, these should show up in national poverty statistics. Evidence on changes in poverty at the national level is, however, mixed. Table 7.4 shows the trend in different estimates of the poverty head count ratio between 1998 and 2011. The national head count poverty rate in 2005, prior to the implementation of the subsidy programme, was estimated at 52% while in the rural areas it was estimated that 56% of the population were living below the poverty line. Seasonally adjusted model-based estimates from welfare monitoring surveys suggest that the poverty rate increased in 2005/6 following the poor 2004/5 crop season and subsequent food shortages, then declined sharply between 2006 and 2007 before stabilizing from 2007 to 2009 (Chirwa et al., 2012). The urban poverty rate is estimated to have (p.162) fallen by almost half in this period. However, recent poverty estimates based on the 2010/11 integrated household survey suggest that between 2004/5 and 2010/11 the national poverty rate was much higher than predicted by the WMS estimates, and only fell by 2% between 2004/5 and 2010/11, suggesting only a marginal change in the well-being of the population. This would be consistent with high maize prices putting a brake on growth in real incomes in these years: one would then expect low maize prices in 2010/11 to stimulate further growth in real incomes and falling poverty. However, poverty incidence in rural areas is estimated at over 56% in 2010/11 (National Statistical Office, 2012), much higher than expected, with a fall of only 1.5% from 58.1% in 2004/5 to 56.6% in 2010/11.

The very limited fall in estimated poverty incidence from 2004/5 to 2010/11, as reported by National Statistical Office (2012), is difficult to reconcile with estimates of wider changes (in labour supply, wages, crop income, school enrolment, child health, subjective well-being, asset ownership, and experience of shocks) reported in different surveys discussed in this and the previous chapter. However the continued high reported poverty incidence in 2010/11 (with high incidence of ultra-poverty) is consistent with a number of other estimates of change from 2004/5 to 2010/11 (National Statistical Office, 2005a, 2012): increases in moderate stunting, wasting, and underweight and small reductions in ownership of tables, radios, and sickles. On the other hand, these findings are less easy to reconcile with other estimates from these surveys: falls in severe stunting, wasting, and underweight; increases in ownership of beds and bicycles; increases in permanent and semi-permanent housing with decreases in traditional housing; increases in access to improved water sources and use of improved sanitation; and decreases in the proportion of people reporting inadequate food, clothing, and health care.5 Furthermore, examination of changes in poverty incidence across districts shows a wide range of changes.6 These apparent inconsistencies pose important questions that need to be resolved, suggesting inconsistencies in either the IHS2 or the IHS3.

7.6.3. Nutrition

As with the poverty incidence estimates, one would expect economy-wide FISP impacts on real incomes to lead to national improvements in the nutrition status of children in Malawi. Anthropometric indicators, however, present an (p.163)

Table 7.4. Trends in poverty headcount in Malawi, 1998–2011 (%)

1998

2004/5

2005*

2006*

2007*

2008*

2009*

2010/11

National

54.1

52.4

56

58.4

41.3

41

40

50.7

Urban**

18.5

25.4

24

25

11

13

14

17.3

Rural**

58.1

55.9

53

47

44

44

43

56.6

Note:

(*) Predicted poverty rates based on an econometric model using welfare monitoring survey data (Mathiassen, 2006 and National Statistical Office, 2005b) with seasonal adjustments as in Chirwa et al. (2012).

(**) WMS urban and rural figures are not seasonally adjusted.

apparently confusing picture (Table 7.5). This expectation requires first that the FISP has raised the real incomes of poorer households, and second that such increases in real income lead to improved child nutrition. We consider here first evidence of wider changes in children’s nutritional status, and then, more briefly, evidence that higher incomes do lead to improved child nutrition. There are, however, difficulties in interpreting anthropometric measures across different surveys as a new standard population reference for the calculation of under-nutrition measures was developed by the WHO in 2006, replacing the 1977 NCHS/CDC/WHO reference. Estimates calculated using the different reference populations are not comparable (de Onis et al., 2006; National Statistical Office and ICF Macro, 2011), but the 2010 Demographic and Health Survey (National Statistical Office and ICF Macro, 2011) usefully provides estimates derived from both standard population references, and we include both of these in Table 7.5 to provide some indication of the way that the different standard population references may affect comparisons across surveys.

Stunting is the measure that should provide the best indicator of longer term child nutrition, as it is less affected than underweight and wasting by seasonal variations in food intake and disease. Simple comparison of the estimates in Table 7.5 suggests that there has not been much of a fall in under-nutrition from 2000 to 2011. However, allowance needs to be made for the change in standard reference population for the 2010 DHS and 2010/11 IHS3 surveys. These two surveys give similar estimates using the 2006 WHO reference population (47.1% and 48.1% prevalence of moderate stunting, respectively). The DHS2010 estimate translates into a much lower prevalence when calculated using the 1977 NCHS/CDC/WHO reference. There would presumably be a similar adjustment to the 2010/11 IHS3 (making it something like 42.5%) providing no evidence of any increase in stunting and suggesting that if anything it may have fallen. Similar arguments apply with regard to estimates of severe stunting, with much stronger evidence for a fall in the prevalence of severe stunting. (p.164)

Table 7.5. Nutritional status of children under 5 years, 2000–11 (%)

Indicator

2000

2004/5

2006

2009

2010

2011

(DHS*)

(IHS2*)

(MICS*)

(WMS)

(DHS*) (DHS**)

(IHS3**)

Stunting

49

43.2

46

36

41.5

47.1

48.1

Severe stunting

-

17.8

20.5

-

15.4

19.6

14

Wasting

5.5

4.6

3.5

1

3.7

4

11.4

Severe wasting

-

1.3

0.5

-

0.9

1.5

1

Underweight

25.4

22.2

20.5

17

17.8

12.8

30.6

Severely underweight

-

7.4

3.6

-

3

3

1.2

Note: DHS = Demographic and Health Survey, IHS = Integrated Household Survey, MICS = Multiple Indicator Cluster Survey, WMS = Welfare Monitoring Survey.

(*) : compared against 1977 NCHS/CDC/WHO reference.

(**) : compared against WHO Child Growth Standards adopted in 2006.

Measurement of wasting and being underweight requires accurate weighing of children. On both these indicators the 2010/11 Integrated Household Survey (IHS) stands out for very high prevalence estimates as compared with other surveys (even allowing for adjustments for the different reference populations, which are very small for wasting). This contrasts with the percentage underweight estimates in the 2010 Demographic Household Survey (DHS), estimates which are particularly low compared to earlier surveys when standardized against the same reference population. The 2010/11 IHS estimates of the prevalence of severely wasted and underweight children are however relatively low, particularly when possible adjustments are made to the prevalence of severe wasting to standardize the reference population.

Overall, it is difficult to drawn any firm conclusions on changes in children’s nutritional status from these different estimates. There are, however, possible indications that the prevalence of moderate stunting has fallen a little (although it is still very high) but that there has been a more substantial reduction in the prevalence of severely stunted and of moderately and severely underweight children (if we ignore the 2010/11 Integrated Household Survey estimate of the prevalence of moderately underweight children as an outlier).

The various surveys reported in Table 7.5 also have the potential to provide evidence on links between income and child nutrition. Both the IHS surveys report anthropometric results by consumption quintile, and all surveys report estimates separately by region and for urban and rural areas. Neither the 2004/5 nor 2010/11 IHS show any clear decline in stunting or severe stunting in higher income quintiles. However in the 2010/11 IHS the proportion of children who are moderately or severely wasted or underweight falls (p.165) in higher income quintiles. This is the case for the 2004/5 IHS2 estimates of moderate and severe wasting but not for estimates of the prevalence of underweight children. Both the 2006 MICS and the 2010 DHS show declining stunting and underweight prevalence (moderate and severe) with higher wealth quintiles, but only the 2010 DHS shows this for wasting. International cross-country analysis (for example, Headey, 2011a; Webb and Block, 2012) suggests a stronger relationship between income growth and stunting and a weaker one between income growth and wasting at lower income levels.

Examination of regional differences in anthropometric measures in Malawi may shed a little more light on the relationships between these measures and income and on the consistency between different surveys. Both the IHS2 and the IHS3 estimate highest median incomes and lowest poverty incidence in the central region, and lowest median incomes and highest poverty incidence in the southern region, with the northern region in between. With regard to anthropometric measures the different surveys present somewhat inter-regional comparisons as regards particular measures, but in broad terms the northern region seems to have a lower prevalence of under-nutrition while the central region tends to have the highest prevalence, despite higher median income and lower poverty incidence than the other regions.

7.7. Summary

The chapter set out to review the economy-wide impacts of the Farm Input Subsidy Programme in Malawi. These are important to arguments put forward in Part I regarding the importance of such impacts for large-scale input subsidy programmes supporting staple crop production. Examination of these impacts involved identification of changes potentially associated with the implementation of the FISP and consideration of the attribution of these changes to the FISP and to other potential influences (such as macro-economic management, rainfall, and tobacco prices) using four principle approaches: consideration of patterns of change in different variables; regression analysis; simulations with CGE models; and simulations with a partial equilibrium Informal Rural Economy model.

Although most macro-economic indicators show an environment of macro-economic stability, some of the indicators—such as fiscal deficits and domestic debt—have been unfavourable within the period of implementation of the subsidy programme. While good macro-economic management made the implementation of the FISP possible, the FISP may have then contributed both to good agricultural growth and to fiscal deficits and domestic debt (in and following years when FISP costs were not well controlled). There are possible difficulties with published estimates of GDP growth, but even (p.166) allowing for these, agricultural sector performance in the FISP period would still average a little under 8%—or a little under 7% if the 2006 recovery from bad 2005 rains is factored in.

The analysis of the economy-wide effects of the input subsidy programme must recognize some mixed and puzzling results. While there are multiple sources of evidence for the positive effect of the subsidy programme on production, high maize prices from 2007 to 2009 are not obviously consistent with this, but a number of explanations for this are put forward. Evidence for rising nominal and real wages (as measured against maize prices) is very strong, derived from a variety of different information sources and analytical approaches. Regression analysis linking wage rates to subsidy receipts in different areas is particularly revealing as it also provides insights into the effects of changes in tobacco prices and maize prices on nominal and real wage rates. Evidence of the FISP causing increases in real wage rates and consequent rises in real incomes is also provided by qualitative date from rural people and by CGE and IRE modelling. These sources also suggest that there have been or should have been increases in real incomes, especially among poor buyers of maize and sellers of ganyu labour, and consequent falls in poverty incidence. However while there is a substantial body of evidence suggesting that this has been the case, the very recently released Integrated Household Survey poverty incidence estimates for 2010/11 are only very slightly lower than the 2004/5 estimates. We are not currently able to resolve the inconsistencies between these conflicting sets of information.

Notes:

(1) We do not discuss rates of inflation here as National Statistical Office (2012: p. 207) casts some doubt on the reliability of inflation figures.

(2) Although for ease of explanation the effects of possible under-estimation of CPI are averaged across the 2006 to 2011 period, in fact distortions are more likely to have arisen from 2008 onwards when maize prices started rising.

(3) Production estimates from a production season are linked to prices in the following marketing season. Thus, for example, the very high prices experienced in the 2008/9 market season are shown in Figure 7.4 against the 2007/8 production season.

(7) The regression estimate is discussed in Chapter 9.

(4) This is illustrated by the divergence between estimates in Ministry of Agriculture and Food Security (2007) and National Statistical Office (2010a).

(5) These improvements are found across all income quintiles.

(6) These, for example, range from an astonishing almost halving of poverty incidence in six years in Thyolo and Rumphi Districts, from 65 to 37% and from 62 to 37% respectively, to a rise in poverty incidence from 38 to 57% in Lilongwe, from 66 to 82% in Chikwawa, and from 51 to 61% in Mzimba District (next to Rumphi).