The module 4 Corresponds to impacts on beneficiary households where LAPs seek security and legal certainty about land ownership.

Module 4: Household Livelihoods

Statistical design

Statistical design is a key factor in impact evaluation. It helps to define the sampling scheme, the sample design, the selection and how to expand the results obtained in a sample to the whole range of properties studied (expansion factors).

It is recommended that this design be developed by a specialist, however the aspects that the designer should consider when choosing the sampling type required should be highlighted; this will facilitate dialogue with the specialist and the subsequent interpretation of the data.

1 Confidence and precision parameters should be chosen; for this type of study the recommended confidence level should be 95% and never less than 90%. The precision or margin for error should not exceed 5%. As the error rate is limited, the sample tends to be greater and collection more costly.

2 .The sampling technique should be chosen: the basic technique is simple random sampling; although this is easy to calculate, it tends to generate very large samples that increase the cost of the evaluation. Stratified sampling is recommended for LAPs, establishing homogeneous subgroups within the sample based on territory, in order to have greater precision. For LAPs at least three groups should be set up for the strata:

Treatment groups

Control groups

Urban intervention group

Urban control group

Rural intervention group

Rural control group

This stratification is the key to the analysis because the behaviour of the variables may be expected to be very different in urban and rural areas, particularly in relation to land use, property value and perception of security.

3. Take into account how the projects are implemented. LAPs have had two types of intervention which can have consequences for the effects of RCT processes:

  • Systematic regularization of a whole geographic area, consisting of cadastral mapping, legal diagnosis and the systematic regularization of tenure in parcels that meet the eligibility criteria of the programme.
  • Voluntary regularization, which offers facilities for individuals or groups of individuals who are applying to the programme for regularization processes.

The effects of voluntary regularization may be greater than those of systematic regularization because in the former case beneficiary households are those which have shown personal motivation to obtain a title or other forms of strengthening their tenure rights. It is therefore more likely that they will try to obtain certain types of benefits once the tenure of their land has been regularized (see bibliography for further information).

These two types can in turn form two study groups as follows:

Intervention group with systematic regularization

Intervention group with voluntary regularization

Non-intervention group

These three groups would apply equally to urban or rural cases. In cases where an evaluation is performed when the regularization process has not been completed, a further subdivision could be applied to the intervention group with systematic regularization, as follows:

Group with completed titling
This refers to households in territories where titling has been completed.

Semi-intervention group
This refers to households in territories where cadastral mapping has been started or completed but the RCT process has not started or is incomplete.

Example of sampling scheme

Example of sampling scheme

To evaluate the impact of land rights in various eligible areas in a country (departments, provinces, groups of municipalities, etc.), a sampling scheme should be designed which allows different evaluation methods to be applied, ranging from the classic experimental model to control models.

Statistical design: sampling technique to be applied:

» Stratified, with four strata or independent groups (urban and rural control and treatment)

» Multistage, to reduce collection costs by minimizing the dispersal of the ultimate sampling units. The three selection stages would be:
• One – Municipalities.
• Two - Comarca or districts.
• Three - Parcels.

» With allocation in proportion to size (number of properties), within each stratum.

Aspects to be considered when determining sample size:
» Confidence level: 90% to 95%
» Margin for error: 4% to 6%
» Design effect: between 1.5 and 3.2
» Expected no response rate: 5% to 18%, depending on the type of stratum and the precision of the sampling frame

For a total of more than 100,000 properties, and with the sampling technique described and the parameters to be considered, a very approximate sample size is estimated per stratum of 300 to 400 households, which is equivalent to a total collection of 1,200 to 1,600 evaluation forms.