{smcl}
{* 31aug2026}{...}
{cmd:help opl_policy_eval}
{hline}

{title:Title}

{p2colset 5 18 22 2}{...}
{p2col :opl_policy_eval {hline 1}} Evaluate a user-supplied binary treatment policy after {cmd:opl_tb_cba}
{p2colreset}{...}

{title:Syntax}

{p 8 8}
{cmd:opl_policy_eval} {it:policyvar}

{pstd}
where {it:policyvar} is a numeric binary variable coded 1 for treated
observations and 0 for untreated observations.

{dlgtab:Description}

{pstd}
{cmd:opl_policy_eval} evaluates a user-supplied binary treatment policy after
{helpb opl_tb_cba}. The command uses the treatment-effect variable, treatment
cost, cost-weight parameter, and estimation sample stored by
{cmd:opl_tb_cba}.

{pstd}
For a policy indicator {it:D_i}, the command computes treatment coverage,
total treatment benefit, total treatment cost, and weighted social surplus:

{p 12 12}
Q(D;lambda) = sum_i D_i tau_i - lambda sum_i D_i c_i.

{pstd}
Here, {it:tau_i} is the estimated individual treatment effect,
{it:c_i} is the individual treatment cost, and {it:lambda} is inherited from
the preceding {cmd:opl_tb_cba} estimation.

{pstd}
The evaluation sample is the estimation sample from {cmd:opl_tb_cba},
excluding observations with missing values in the policy, treatment-effect,
or cost variables.

{pstd}
{cmd:opl_policy_eval} may be used to evaluate externally defined policies,
benchmark policies, or policy variables generated by {cmd:opl_tb_cba}, such
as the estimated optimal threshold policy and the first-best policy.

{dlgtab:Requirements}

{phang}
{cmd:opl_policy_eval} must be run after {helpb opl_tb_cba}. The preceding command
must store {cmd:e(tauvar)}, {cmd:e(costvar)}, {cmd:e(lambda)}, and
{cmd:e(sample)}.

{phang}
{it:policyvar} must be numeric and must contain only 0 and 1 in the evaluation
sample.

{phang}
Observations with missing values in {it:policyvar}, the treatment-effect
variable, or the cost variable are excluded from evaluation.

{dlgtab:Policy evaluation measures}

{pstd}
The command computes the following quantities:

{synoptset 22 tabbed}
{synopt:{cmd:N}}number of observations in the evaluation sample{p_end}
{synopt:{cmd:N_treated}}number assigned to treatment{p_end}
{synopt:{cmd:N_untreated}}number assigned to no treatment{p_end}
{synopt:{cmd:coverage}}share of the evaluation sample assigned to treatment{p_end}
{synopt:{cmd:total_benefit}}sum of estimated treatment effects among treated observations{p_end}
{synopt:{cmd:total_cost}}sum of treatment costs among treated observations{p_end}
{synopt:{cmd:Q}}weighted social surplus{p_end}
{synopt:{cmd:ATET}}average estimated treatment effect among treated observations{p_end}
{synopt:{cmd:avg_cost}}average treatment cost among treated observations{p_end}
{synopt:{cmd:avg_surplus}}average weighted surplus among treated observations{p_end}

{pstd}
Treatment coverage is

{p 12 12}
coverage = N_treated / N.

{pstd}
Weighted social surplus is

{p 12 12}
Q = total benefit - lambda x total cost.

{pstd}
Average surplus is defined as

{p 12 12}
average surplus = Q / N_treated.

{pstd}
If the policy assigns no observations to treatment, total benefit, total cost,
and total welfare are set to zero, while ATET, average cost, and average
surplus are returned as missing.

{dlgtab:Stored results}

{pstd}
{cmd:opl_policy_eval} is {cmd:rclass} and stores the following results:

{synoptset 32 tabbed}
{synopt:{cmd:r(N)}}number of observations in the evaluation sample{p_end}
{synopt:{cmd:r(N_treated)}}number assigned to treatment{p_end}
{synopt:{cmd:r(N_untreated)}}number assigned to no treatment{p_end}
{synopt:{cmd:r(coverage)}}treatment coverage{p_end}
{synopt:{cmd:r(total_benefit)}}total treatment benefit{p_end}
{synopt:{cmd:r(total_cost)}}total treatment cost{p_end}
{synopt:{cmd:r(Q)}}weighted social surplus{p_end}
{synopt:{cmd:r(ATET)}}average treatment effect among treated observations{p_end}
{synopt:{cmd:r(avg_cost)}}average treatment cost among treated observations{p_end}
{synopt:{cmd:r(avg_surplus)}}average weighted surplus among treated observations{p_end}
{synopt:{cmd:r(policy)}}evaluated policy variable{p_end}
{synopt:{cmd:r(tauvar)}}treatment-effect variable inherited from {cmd:opl_tb_cba}{p_end}
{synopt:{cmd:r(costvar)}}cost variable inherited from {cmd:opl_tb_cba}{p_end}

{dlgtab:Remarks}

{pstd}
{cmd:opl_policy_eval} does not estimate a treatment policy. It evaluates an
existing binary policy using the treatment-effect and cost information stored
by the latest {cmd:opl_tb_cba} estimation.

{pstd}
The command evaluates the policy on the original estimation sample. It does
not permit an independent {cmd:if} or {cmd:in} restriction.

{pstd}
The reported ATET is the average estimated treatment effect among observations
assigned to treatment by the evaluated policy. It should therefore be
interpreted as a policy-specific average treatment effect among the treated.

{pstd}
Because {cmd:lambda} is inherited from {cmd:opl_tb_cba}, all policies
evaluated after the same estimation are compared using the same cost-benefit
criterion.

{dlgtab:Examples}

{pstd}
Estimate a threshold-based policy before evaluating alternative binary
policies:

{phang2}
{stata "sysuse data_opl_tb_cba_2, clear"}

{phang2}
{stata "opl_tb_cba tau, cost(cost) select(x1 x2) lambda(1) npoints(51) custompolicy(D) custom(.30 .30) generate(opl2) replace"}

{pstd}
Evaluate an externally defined policy:

{phang2}
{stata "opl_policy_eval D"}

{pstd}
Evaluate the optimal threshold policy generated by {cmd:opl_tb_cba}:

{phang2}
{stata "opl_policy_eval opl2_opt"}

{pstd}
Evaluate the first-best policy generated by {cmd:opl_tb_cba}:

{phang2}
{stata "opl_policy_eval opl2_fb"}

{pstd}
Inspect the returned results after any evaluation:

{phang2}
{stata "return list"}

{dlgtab:Acknowledgment}

{pstd}
The development of this software was supported by FOSSR (Fostering Open
Science in Social Science Research), a project funded by the European Union -
NextGenerationEU under the NPRR Grant agreement n. MURIR0000008.

{dlgtab:Author}

{phang}
Giovanni Cerulli{p_end}

{phang}
IRCrES-CNR{p_end}

{phang}
Research Institute for Sustainable Economic Growth,
National Research Council of Italy{p_end}

{phang}
E-mail: {browse "giovanni.cerulli@cnr.it"}{p_end}

{dlgtab:Also see}

{psee}
Online: {helpb opl}
{p_end}