Aureo de Paula, Gil Shapira, and Petra Todd, "How Beliefs about HIV
Status Affect Risky Behaviors: Evidence from Malawi", Journal of
Applied Econometrics, Vol. 29, No. 6, 2014, pp. 944-964.

The data used in the article are from the Malawi Diffusion and
Ideational Change project (presently named the Malawi Longitudinal
Study of Families and Health (MLSFH)). The waves used in the analysis
(2004, 2006 and 2008) are not publicly available as of May 2013.
Information on the project and on how to obtain the data can be found
at malawi.pop.upenn.edu .

We provide here two data files without identifying information that is
restricted to the information/observations that is required for our
analysis. The file dPST_extra.csv includes the dataset used for the
analysis of the extramarital sex outcome. The sample is of men who are
married in all three data rounds included. The file dPST_number.csv
includes the dataset used for the analysis of the
more-than-one-sexual-partner outcome. The sample is not restricted
only to married men.

The files includes the following variables: 

1. Age in 2008: 'age2008'

2. Region of residence: 'Balaka, Rumphi, Mchinji'

3. Religion: 'moslem, christian'

4. Education category: 'no_school, primary, secondary, higher_educ'

5. Household has a metal roof in the different rounds:
   'metal_roof2006, metal_roof2008'

6. Number of children in the differenet rounds: 'numchild2004,
   numchild2006, numchild2008'

7. Number of children is missing in specific rounds: 'no_numchild2004.
   no_numchild2006, no_numchild2008'

8. Man is married: 'marital2006, marital2008'

9. Polygamy status: 'poly2004, poly2006, poly2008'

10. More than one sexual partner in different rounds: 'more12004,
    more12006, more12008'

11. Extramarital sex (married respondents only): 'extra2004,
    extra2006, extra2008'

12. Likelihood assigned to being HIV-infected, measured in beans:
    'bean_status2006, bean_status2008'

13. Likelihood assigned to spouse being HIV-infected, measured in
    beans: 'spouse_bean2006, spouse_bean2008'

14. Verbal categories assigned to being HIV-infected (0 = no likelihood,
    1 = low, 2 = medium, 3 = high): 'hiv_likeli2004, hiv_likeli2006,
    hiv_likeli2008'

15. Perceived prevalence (out of 10 individuals in the village, how
    many are HIV-infected): 'out_of_ten2004, out_of_ten2006,
    out_of_ten2008' 

The two csv files are zipped in the file pst-data.zip.

The file pst-matlab.zip contains the files used to perform the
modified Arellano and Carrasco (2003), or AC, procedures reported in
our paper.

Program Files:

1. 'main.m': main program that calls the functions in the other .m
file. The program performs the AC procedure under diferent
specifications (different outcomes, different definitions of cells,
misreporting of sexual behavior, etc.)

2. 'objective_mis': objective function for the GMM procedure

3. 'integral.m': numerical integral

Data input files:

These are data matrices called by 'mis_main.m'. The variables and
their order are described in the mis_main.m file

1. 'extra.out': Sample of men who are married in all rounds. Outomce:
   extramarital sex.

2. 'extra_perc.out': Sample of men who are married in all rounds.
   Outomce: extramarital sex. Include perceptions about prevalence in the
   village.

3. 'number.out': Sample of married and unmarried men appearing in all
   rounds. Outomce: more than one sexual partner.

4. 'number_perc.out': Sample of married and unmarried men appearing in
   all rounds. Outomce: more than one sexual partner. Include perceptions
   about prevalence in the village.


The file pst-stata.zip includes the Stata .do files used to create our
dataset, produce descriptive statistics, and produce the logit model
analysis (except the AC procedure and the attrition analysis that
requires a larger dataset).

1. 'for_analysis.do': Organizes and outputs the data in format
   suitable for Matlab, for the different specifications.

3. 'cells.do': provides the size of the different cells used in the
   Arellano & Carrasco procedure. Presented in Tables A5 & A6

4. 'Tables.do': provides descriptive statistics presented in Tables 1,
   2, A7 and figure 1

5. 'logits.do': produces analysis presented in table A1

6. 'logits_no_fe.do': produces analysis presented in table 3


All text files are ASCII files in DOS format. Unix/Linux users should
use "unzip -a".

Please address any questions to:

Gil Shapira
Development Research Group 
The World Bank
Mailstop MC3-311
1818 H Street NW
Washington, DC 20433
USA
EMAIL: gshapira [AT] worldbank.org