• No se han encontrado resultados

Diez preguntas

In document En la desembocadura del Yukon (página 32-42)

Propensity score matching requires an over specified regression model for the selection into the FSP programme. As such, for the standard PSM approach the following baseline

variables, taken from both the PHS 1999/2000 and the SS 2000, were included:38

 Head of household information

o Age: Dummies for Under 18, 18-30, 30-45, 45-60, 60+

o Gender: Dummy for male

o Education: Dummies for None, Primary, Secondary, Higher

 Household demographics39

o Household size

o Number of adults, Number of adult males, Number of children

o Number of members in each education band

 Assets

o Land: Total, total owned, and total cultivated

o Household maize stock – a dummy for whether the household still has maize

in stock from the previous harvest (2000).

o Livestock: Dummies for household ownership of cattle, pigs, goats, sheep and

chicken

o Farming implements: Dummies for household ownership of plough, harrow

and oxcart  Household livelihoods

o Remittances: Amount sent and received from within district, other districts,

and abroad

o Number employed

38

Roughly equivalent variables were used for the post-treatment matching approach. They were drawn from the SS 2004 and SS 2008 respectively, as well as the variables which the research group at Michigan State University have derived from them.

39

o Most important source of off-farm income: Dummies for None, Labour on

another farm, Labour mining, Small business, Food processing,

Salary/pension, Other

o Total employment / labour income

o Crop purchases – Total amount spent on buying crops for consumption since

the last harvest

o Credit – Number of loans, value of loans, and value of informal loans during

the 99/00 agricultural season  Household level inputs

o Dummies for soil conservation farming methods: Mixing of two or more crops

in one field, Improved fallow, Pot holing / planting holes, Crop rotation, Leaving crop residues (except for cotton and tobacco), Other minimum tillage techniques

o Dummies for use of basal fertiliser, top / dressing fertiliser and manure

 Crop level inputs

o Dummy for whether cassava/burley tobacco grown

o Dummies for animal or mechanical power used for land preparation and

whether used for planting

o Dummy for purchase seed for this season

o Dummies for apply basal or top fertiliser

 Farming conditions:

o Climate and soil: Dummies for agricultural zone, pH of soil, pH of soil squared,

exchange aluminium, exchange aluminium squared, dummies for soil type, suitability for high intensity maize production, suitability for low intensity maize production

o Access to market: distance to nearest main road, distance to nearest town40

o District dummies to account for regional variation

B.4

Results

This section presents a series of six tables which provide the results of the impact analysis. The tables follow a standard format.

Columns (1) to (6) relate to average levels and changes in levels of the outcome measures:  Column (1) gives the average level of the outcome measure at baseline (i.e. pre-

treatment) for FSP beneficiaries;

40

There is concern that these variables might constitute potential instruments, in the sense that they introduced truly exogenous variation. As discussed above, the inclusion of instruments in the conditioning set is not advisable when applying Propensity Score Matching methods. While the exogeneity of the variables is questionable, the analysis was carried out with and without these variables, as a sensitivity analysis. The inclusion of the variables was found to have very little effect on the ATT coefficient, and so in the interest of improving matching they were left in.

 Column (2) gives the average level of the outcome measure at follow-up (i.e. post- treatment) for FSP beneficiaries;

 Column (3) gives the average of the change per household between baseline and follow- up levels for FSP beneficiaries;

 Column (4) gives the average of the change per household between baseline and follow- up levels for the matched non-FSP beneficiaries;

 Column (5) gives the estimate of the average change in levels that is attributable to FSP participation (the ATT estimate);

 Column (6) gives the standard deviation of the estimates shown in column (3).

Columns (7) to (10) relate to average growth rates of the outcome measures:41

 Column (7) gives the average growth of the outcome measure between baseline and follow-up for FSP beneficiaries;

 Column (8) gives the average growth of the outcome measure between baseline and follow-up for the matched non-FSP beneficiaries;

 Column (9) gives the estimate of the average growth that is attributable to FSP (the ATT estimate);

 Column (10) gives the standard deviation of the estimates shown in column (7). The tables are divided into three panels:

 The top panel (A), presents estimates that relate to all current FSP beneficiaries;

 The middle panel (B), presents estimates that relate to only those current FSP beneficiaries that were already using fertiliser in 2000 (i.e. pre-treatment);

 The bottom panel (C), presents estimates that relate to only those current FSP beneficiaries that report receiving the FSP support on time42.

T-scores of all ATT estimates are reported in brackets, and significance is indicated based upon them according to the following levels: * = 90%, ** = 95%, *** = 99%. The statistical significance of all estimates presented in the tables below are based upon single runs of the PSM procedure. However, the differences between these and those based on bootstrapped standard errors were found to be small, and the significance of results were not found to change at the 5% level for all important indicators.

41

Average growth is computed as the geometric mean of individual growth rates. Specifically, we define average growth as ln(Xt + 1) – ln(X0 + 1), where Xi is the level of the outcome variable at time i.

42

Households are classed as receiving FSP support on time if both basal and top soil were available to the household through the programme at the time that they were needed.

Table B.1

FSP impact analysis results: 2000-04, standard PSM approach (matching on pre-treatment

In document En la desembocadura del Yukon (página 32-42)