«Edited by Helene Bie Lilleør and Ulrik Lund-Sørensen Practical Action Publishing Ltd The Schumacher Centre, Bourton on Dunsmore, Rugby, ...»
Despite this positive impact of RIPAT on food security, we did not find any significant impact on the prevalence of poverty as measured by a number of different poverty indicators relating to households’ expenditure levels. There is a variety of possible reasons for this finding. Food security in the lean period of the year will increase if households find it easier to smooth their food consumption over the year, even if their total food consumption levels remain unchanged. Furthermore, all households have limited resources which they have to allocate between different types of expenditure (food and The impaCT oF ripaT on Food seCuriTy and poverTy change in the PPI only if the household has experienced a substantial improvement in their level of wealth.
Our second approach is therefore to assess whether some partial poverty indicators have changed, i.e. whether housing standards have improved, the ownership of luxury items has increased, or the proportion of children in the household enrolled in school has risen. Although such poverty indicators are only indirect expressions of the level of poverty within a household and therefore cruder than direct income measures, they can still provide useful information about the overall impact on socio-economic well-being among participating RIPAT households. Because they are simpler, we expect these measures to be of a more dynamic nature than the composite PPI.
5.3 RIPAT households and comparison households To analyse the impact of RIPAT 1 on food security and poverty among participating households, we collected data using the large-scale quantitative household survey described in Chapter 3. The survey covered roughly 90 per cent of all RIPAT 1 and RIPAT 3 households, as well as comparison households from 20 selected comparison villages in the two districts of implementation, Arumeru and Karatu (see Table 3.1 in Chapter 3). For the impact assessment in this chapter we used a total of 1,718 interviewed households, of whom 424 had participated in RIPAT 1 groups and 491 had participated in RIPAT 3 groups. There were 361 households from comparison villages near the RIPAT 1 area in Arumeru and 442 households from comparison villages near the RIPAT 3 area in Karatu.2 We explain in Section 5.5 why we also need information on RIPAT 3 households for the impact assessment of RIPAT 1.
In Table 5.1, we list some of the main characteristics of all the households and villages (both RIPAT households and villages and their comparison households and villages) included in the impact analysis.
The first column in the table lists the overall averages for all the households and villages included. In the subsequent columns, this average is broken down by the four types of household and village that form the basis of our impact analysis, namely RIPAT 1 households and villages (R1) and their local comparison households and villages (C1) in Arumeru District, and RIPAT 3 households and villages (R3) and their local comparison households and villages (C3) in Karatu District.
The majority of households in the sample have male heads. The average age of heads of households is 47 years and they have on average 5.5 years of schooling. There are three children living in the average household, which owns roughly 3 acres of land. The most important food crop for households in the sample is maize, and the most important cash crops are beans in Arumeru and pigeon peas in Karatu District. The majority of the houses have access to well or piped water for drinking, but a substantially larger proportion of the households in Karatu have very poor housing standards in terms of floor, walls, and roofs than is the case in Arumeru. All the villages have primary schools, and many villages also have a secondary school. Many villages have an agricultural extension officer associated with the village. On average, the villages are 10 km from the nearest market where farmers can sell their crops. The majority of villages in Karatu have no irrigation channels, whereas the opposite is the case for villages in Arumeru District.
Although there are several similarities between the two districts, there are also important differences. Karatu is a somewhat poorer and more remote district than Arumeru, with 44 per cent of the population in Karatu having a high risk of being poor The impaCT oF ripaT on Food seCuriTy and poverTy also found clear indications of high adoption rates among RIPAT farmers compared with their comparison farmers in comparison villages. Both RIPAT 1 and RIPAT 3 farmers are significantly more likely to employ zero-grazing husbandry for some of their livestock, and are up to 20 percentage points more likely to be members of a savings and loan group than their comparison farmers in comparison villages. The consequences of this for the impact analysis are discussed in Section 5.5.
5.5 The impact of RIPAT on food security and poverty Having established that RIPAT farmers did in fact adopt components of the basket of options to a greater or lesser extent, the next question is whether this adoption led to the expected impact on food security and poverty as stated in the development goals of the project. That is, do we see an increase in food security and a reduction in poverty among RIPAT 1 participants due to the RIPAT 1 intervention? In order to identify such an impact, we used a statistical method called ‘Difference-in-Differences’, which establishes the difference in impact between RIPAT 1 and RIPAT 3 households, subtracting the difference between their comparison households to take into account regional variations stemming from the fact that the two projects were implemented in two different districts.
Using this method, we exploited the fact that there is a two-year time lag between RIPAT 1 implementation (which started in 2006) and RIPAT 3 implementation (which started in 2008). This allowed us to account for two important and unobserved selection processes, namely that: 1) RIPAT villages were not chosen at random from among all the villages in the two districts, but rather they were selected due to their specific characteristics; and 2) similarly, RIPAT households were not chosen at random but rather volunteered to join the project, and hence might have been more motivated to improve their livelihoods than the average comparison households in the comparison villages.
Therefore, we made use of the fact that RIPAT 1 and RIPAT 3 households are likely to be more directly comparable with each other in certain respects than are RIPAT households with randomly chosen comparison households in comparison villages. However, there will be other respects in which the RIPAT 1 households and RIPAT 3 households differ, because they reflect variations in the districts where they live, for example agro-ecological and socio-economic differences. The comparison households allowed us to control for these regional differences in the impact analysis, as explained in Box 5.3.
Impact on food security When should we expect to see an impact on food security from RIPAT? We cannot expect any impact before the household has adopted the new agricultural technologies and the crops have had time to mature and ripen, which would take 9–15 months in the case of bananas, depending on the local climate. By then it should be possible to detect small impacts on food security if the implementation and production of the new crops have gone according to plan. Unfortunately, many of the RIPAT 3 villages were badly hit by drought in the first year after project commencement. We therefore expect only a very limited impact on food security in RIPAT 3 by January 2011, and not nearly the same impact as in RIPAT 1 (see Box 5.3 for details).
We find a statistically significant positive impact of RIPAT on access to food, measured using the Household Hunger Scale. Using the Difference-in-Differences method, RIPAT 1 58 Farmers’ ChoiCe the graduated bars cross the zero impact line, we cannot statistically distinguish the estimated impact on fish consumption from zero. RIPAT 1 households have, however, also become more likely to eat eggs. This is less surprising, since improved chicken breeds were part of the basket of options in the RIPAT project, as mentioned in Section 5.4.
Similarly, the implementation study also reported improved nutritional quality of the daily diet as exemplified by the quote from a female farmer from Marurani village (see Section 6.3, Chapter 6).
The last food security outcome we examined was the degree of malnutrition among children under five years old. We found that some children in a few villages are considerably less likely to be stunted. This results in an overall average of young children being 27 percentage points less likely to be stunted.3 It suggests that for these children the improved food security, in terms of both better access to food and better nutrition, has been sufficiently persistent to have generated a lasting impact on their lives. RECODA has – with its focus on bananas and chickens, and hence on the production of eggs – hit on one of the best and most cost-effective nutritional combinations that one could obtain. In a study from the Philippines (Banerjee and Duflo, 2011: 26), eating eggs and bananas has been found to be the cheapest way to obtain sufficient calories, and to provide the right proportion of calories from fat and from protein; this might also apply to rural Tanzania.
Impact on poverty Despite the fact that we found clear indications of a strong positive impact from RIPAT on food security outcomes, we did not find any statistically significant impact on the risk of being poor as indicated by the PPI measure, as can be seen from the first bar in Figure 5.4. Although the impact of RIPAT on the risk of poverty appears to be positive (at slightly below 10 percentage points), the shaded bar crosses the zero line, indicating that we cannot distinguish the impact from a zero effect with any reasonable degree of statistical certainty.
There are a number of possible reasons why the study did not reveal an impact on poverty in RIPAT households.
First, it is possible that RIPAT did make an impact on poverty, but that this impact was not captured by the measure of poverty used, the PPI. The PPI measures poverty indirectly, by measuring spending, rather than directly, by measuring income. It does this through 10 simple questions on items owned in a household, the condition of the house, literacy of the spouse, and children’s education. To achieve a change in the overall PPI value would require additional spending on household items, housing quality, or children’s schooling. This suggests that the PPI value for a household tends to change only over a long period or with a substantial change in the level of prosperity.
However, from the PPI data documentation it is possible to extract a list of consumption indicators that are highly correlated with having expenditure levels below the national poverty line. We also examined the results for these separate items of consumption. The average values for one item may change more quickly than the full index, and thus an examination of individual items is more likely to reveal short-term changes. Again, we found no impact from RIPAT 1 on any of the individual indicators, be it housing quality measures, ownership of ‘luxury’ items such as mobile phone, sofas, stoves, lanterns, or watches. Some examples are shown in Figure 5.4. Although the estimated impacts are 60 Farmers’ ChoiCe as they have been encouraged to do by RECODA since 2009; RIPAT participants are 20 percentage points more likely to be members of savings groups than their comparison households. These factors suggest that not only have the households opted to reduce their cash income by reducing their supply of casual labour, they have also increasingly opted to use whatever cash income they have on agricultural investments (through hiring labour) or on savings (through their savings groups), rather than on improving their housing quality or purchasing luxury items. Because all these three changes are already evident among households in RIPAT 3, we cannot employ the Difference-inDifferences method to control for the unobserved selection processes of households and villages in RIPAT 1, but instead have to compare RIPAT households and their comparison households in the same district.
The findings regarding the use of labour within the household are shown in the last two bars in Figure 5.4; these depict the situation for RIPAT 3 households only. Here we see that RIPAT 3 households rely 10 percentage points less on supplying casual labour as one of the most important income sources in the household than their comparison households do. In rural Tanzania, supplying casual labour is strongly stigmatized, and many farmers cut back on this as soon as they can afford to do so, despite it being a rather remunerative source of income. We should therefore expect the reduction in casual labour to be a result of a household being able to do without that income source, although the net effect may be a reduction in overall income level. RIPAT 3 farmers have not only cut back on their supply of casual labour, they are also 10 percentage points more likely than their comparison households to hire labour from outside to work for them, and RIPAT 1 households are 30 percentage points more likely to do so (not shown in Figure 5.4). This impact is pronounced only among the RIPAT households that grow bananas; this is not surprising, since banana cultivation is very labour intensive in the early stages.
A third possible reason is that RIPAT may have made consumption smoothing easier by spreading the yields of agricultural production, and thus income, more evenly over the year, and by introducing the availability of loans and savings through the savings and loans associations. Improved consumption smoothing could account for improved food security even if incomes had not increased overall. Our food security measure is a measure of whether or not the household has experienced hunger in what is defined as the worst month of the year in this respect. That is to say, our ‘no hunger’ measure can also be seen as a measure of whether RIPAT households have been successful in securing enough food for the ‘hungry’ period – i.e. whether they have increased their ability to smooth food consumption over the year. In terms of agriculture, RIPAT promotes the cultivation of banana, a perennial crop that, once fully established, produces food at a more constant rate throughout the year than annual crops. Improved breeds of livestock also largely provide a yield throughout the year with relatively little seasonal variation.