G. Porro and S. M. Iacus, "Random Recursive Partitioning: A Matching 
Method for the Estimation of the Average Treatment Effect", Journal of 
Applied Econometrics, Vol. 24, No. 1, 2009, pp. 163-185.

The files ip-scripts.zip and ip-data.zip contains scripts and data files
needed to reproduce the figures and tables appearing in the paper. The
contain ASCII files in DOS format. Unix users should use "unzip -a".

The file ip-rda.zip contains the R dump file all.rda. This is a binary file.


*** SOFTWARE ***

The software used is R, which is freely available at
http://www.R-Project.org. A package called RRP which implements the Random
Recursive Partitioning is freely available at http://CRAN.R-Project.org.
Nevertheless, we join here all the scripts and original code.


*** DATA ***

The data sets are the following:

"LL.csv"      (Lalonde data, treated and control experimental units, 722 obs)
"DW.csv"      (Deejia-Whaba subset of LL, 445 obs)
"ST.csv"      (Smith-Todd subset of DW, 250 obs)

"LLvsPSID.csv"	(treated units from LL and control units from PSID data, 
                 2787 obs)
"DWvsPSID.csv"	(treated units from DW and control units from PSID data,
                 2675 obs)
"STvsPSID.cav"	(treated units from ST and control units from PSID data,
                 2598 obs)

All the data sets contain the following 10 variables
treated  : numeric, indicator of treatment
age      : integer, age of subject in the experiment
education: integer, years of education
married  : integer, indicator of being married
nodegree : integer, indicator of lack of possession of degree
black    : integer, ethnic indicator function
hispanic : integer, ethnic indicator function
re74     : numeric, real earnings in 1974
re75     : numeric, real earnings in 1975
re78     : numeric, real earnings in 1978 (the otcome variable of interest)

the column 'X' in the csv files contains labels for each observation.

All the data sets used are also contained in the R dump file:  "all.rda"
To load it into R, use load("all.rda"). To load one of the CSV files, use,
e.g.,  read.csv("LL.csv"), etc.

For data sources, please refer to the references below.


*** SCRIPTS ***

There are the following accessor scripts:
rrp.dist.R		(the RRP algorithm)
mahala.dist.R		(calculates mahalanobis distance)
nnk.att.R		(nearest neighbor matching and att estimation)
optm.att.R		(optimal full matching att estimation)
rrp.att.R		(evaluates att via RRP proximity matrix)

The above scripts are called by the following ones which produce all the
table and figures present in the paper. Figure 2 is an ad hoc picture made
by hand drawing, and it is not data related, so the script does not exist.

script_fig1.R		(script used to generate Figure 1)
script_fig3.R		(script used to generate Figure 3)
script_tab1-7-8.R	(script used to generate Table 1, 7 and 8)
script_tab2.R		(script used to generate Table 2)
script_tab3.R		(script used to generate Table 3)
script_tab4-5-6-9.R	(script used to generate Tables 4, 5, 6 and 9)


*** REFERENCES ***

Dehejia, R., Wahba, S. (2002) Propensity score matching methods for 
	 Non-experimental causal studies, Review of Economics and Statistics, 
	 84(1), 151-161.
	
Lalonde, R. (1986) Evaluating the Econometric Evaluations of Training
	 Programs, American Economic Review, 76, 604-620. 

Smith, J., Todd, P. (2005) Does Matching Overcome Lalonde's Critique of 
       	 Nonexperimental Estimators?, Journal of Econometrics, 125(1-2),
	 305-353.