Main Findings Overview

LSMS-ISA data sets and Sampling

Get Raw Data

Sampling weight

The weights reported on the LSMS-ISA surveys are household weights. The surveys followed a two stage sampling process. In the first stage the clusters were selected, given the desired level of stratification of the survey (i.e. rural/urban, district level, agroecological zones, etc), the first stage weights are calculated as the probability of any given EA to be selected out the whole population of EA’s. In the second stage households are selected within a given cluster (EA), the second stage weights are calculated as the probability of any given household in the cluster to be selected. The household weights are the multiplication of the first and second stage weights.

In Stata, in order to take into account the sampling design, we use the following syntaxis:

svyset comm [pweight=wgt10], strata(stratum) singleunit(centered)

Information Available for studying Trees On Farm

Crop Classification

# Total Crops Listed | 54 % Fruit Trees | 9.43% % Tree Cash Crops | 7.55% % Trees for timber and fuel-wood| 5.66%

Information on planting

Information on harvesting

Replication Repository

All the codes and datasets could be directly retrieved from the Github repository: https://github.com/treesonfarm/UGA. If you prefer to get the zip-file of the repository you can follow this link

If you are Git users, you can directly clone our repository:

git clone https://github.com/treesonfarm/UGA.git

If you have any question or comment about our codes and datasets, don’t hesitate to contact us.

Data Sets

We provide a Stata dataset compatible with stata 12 or higher, with the main variables to study trees on farm (UGA_TreesOnFarm_8Nov2016.dta). In this database, you can get the following variables:

Replication Codes

All our codes use relative paths. So, if you would like to include directly our codes, we only need to set the following two global variables:

global path_work  "[where do you want to save your outputs]"
global path_data  "[where do you have the LSMS-ISA Raw data]"

We made our codes to make self-explained, however if you any question about replicating our results we will be happy to clarify your questions.

Crop Classification

The crop classification was made using all the crops listed throughout the different waves. So, although we only used one wave, you should be able to replicate our classification for all waves. Based on the crop-level data set, you can directly include our codes to classify the crops.

include "UGA-CropClassification.do"

As a result, you will have a crop-level dataset with a new string variable tree_type, which indicates the classification for each crop. In particular:

1. Fruit Tree
2. Tree Cash Crops
3. Plant/Herb/Grass/Roots
4. Trees for timber and fuel-wood
5. NA

For more information about our classification see the different publications from our project.

Stock of Trees On Farm

• Part 1: Building plot-level information

In order to build the total stock of trees on farm at household level, we first need to get the aggregation by plot level. In this code, we get all information from planting and then we collapsed our information to the different measures on presence, extension and area planted by the different crop classification.

include "UGA-0_Stock_TreesOnFarm.do"

As a result, you will have a data set at plot level with different the variables describes above in the dataset at plot level.

• Part 2: Building household-level information

Once we have the information aggregated at household-plot level, we proceed to build the household level data set. In this do-file we merge the crop level information with the complementary data set on plot and household characteristics to build the household-level information.

include "UGA-1_SamplingPlotSize.do"

As a result, we will have the same variables as those contain at UGA_TreesOnFarm_8Nov2016.dta.

Total Sells

Another alternative analysis is the harvesting and sells for the crops by our classification. So, once we have the crop-level classification, we build the aggregation for total production harvested and sell by type of crop. You can include our do-file in this manner:

include "UGA-1_Harvesting_Sells.do"

Shapefile

Using the coordinates available from LSMS-ISA, we built the spatial distribution of the stock of trees on farm. Data was generated using the coordinate reference system: WGS 84 (EPSG:4326).

LSMS-ISA surveys provide modified coordinates to protect household confidentiality, by introducing a random distortion of 0-5km from the original location of the rural household. For more details on this type of mUGAod and its implications for statistical inference see Perez-Heydrich et al. (2013).

The shapefile has seven variable at the attribute table:

In order to merge the shapefile with the entire LSMS-ISA data set, you can use the dataset from the stock trees on farm. It will guarantee a 1:1 merge.