CREP_Policy_Interaction_Working_Paper/body/conclusion.tex

\section{Discussion and Conclusion}
\subsection{Water Conservation Outcomes} 
Combining each of these empirical estimates a picture of the overall groundwater reduction in the subdistrict can be sketched. The interaction between the \ac{CREP} program and the existing pumping fee is the primary result, but the combined output is also relevant for policy choice. Overall effects are summarized by estimating each program using an event study and combining the effects.

The estimated changes due to subdistrict and \ac{CREP} policies are summarized in \cref{RESTBL}, confidence intervals are suppressed for clarity but are provided in appendix \cref{A:CI_TBL}. The \emph{Direct} and \emph{Spillover} columns predict the amount of water saved as a result of the \ac{CREP} fallow program from the methods in \cref{SECCREP} using the well dataset\footnote{There is yearly variation because the number of wells in \ac{CREP} changes, and not all cohorts are in the same lag set. Estimates were made by summing the coefficient for each well in a given lag year and neighbor lag year.}. The \emph{Expected Column} is the total volume of water reduced by \ac{CREP} wells due to the Subdistrict One policies from \cref{REG2011}. This is referred to as \emph{expected} because a \ac{CREP} policy planners would expect that \ac{CREP} will induce this additional volume of water when not considering the interaction with the pumping fee. The total volume of water saved because of \ac{CREP} is the sum of the \emph{Direct} and \emph{Spillover} columns while the total volume of water reduced because of the \ac{CREP} wells fallowing is the sum of all three columns. Under the Subdistrict sections the Fallow column is the predicted direct effect of water savings from the four-year fallowing program as estimated in appendix \cref{A:FALPROG}\footnote{Neighborhood effects are assumed to be zero for this program.}. The \emph{Other Policies} column captures the response of wells not in \ac{CREP} relative to the control group\footnote{The model predicts a yearly event study of water savings of all wells in Subdistrict One compared to the control. These yearly net subdistrict savings are then subtracted by the other policy estimates. Leaving the reduction in water that cannot be accounted for by the \ac{CREP} program and short-term fallow programs.} estimating the yearly effect of all other policies including the pumping fee. Finally, the \emph{No Policies} column is the counterfactual volume expected if none of the water conservation efforts were made. 

\input{Tables/Policy_Estimates.tex}

These results help provide a benchmark of the overall savings. The pumping fee and other subdistrict policies consistently provided the most water savings. This is due to the directed cost of pumping and the fact that all wells within the subdistrict are affected. Over time the direct effect of \ac{CREP} has increased as enrollment enlarges, but the relative importance of the neighbor effects has declined. The neighborhood effect coefficients revert to zero over time, becoming insignificant in five years. The wells enrolled in \ac{CREP} from 2014-2016 have already been active for five years, thus making these cohorts' addition to the spillover effect negligible going forward. Also, as more wells are enrolled then there are fewer wells that are not neighbors to a \ac{CREP} well. Since there is some spatial clustering in the program enrollment, latter \ac{CREP} wells add a smaller number of wells to neighboring treatment group. The progression of the added effect of each policy outcome is provided in \cref{FIG:BAR}.
%\FloatBarrier
    \begin{figure}[h]%
        \centering
        \includegraphics[width=\textwidth]{Policy_Bar_Graph}
	\caption{Water use and conservation in \ac{SBD1}}
        \label{FIG:BAR}
    \end{figure}%
    While the overall water conservation of \ac{SBD1} is substantial the \ac{CREP} and resulting spillover effects are minor compared to the pumping fee and other \ac{SBD1} policies. This smaller \ac{CREP} impact is not due entirely to the original \ac{PES} design but is in part reduced precisely because the \ac{SBD1} polices are successful at reducing groundwater extraction, which adds complications for policy makers seeking to reach conservation targets.

How the outcomes of \ac{CREP} may be different from policymaker expectations can also be found. The volume of groundwater saved by the \ac{CREP} program is found to be 32\% lower due to the existence of Subdistrict One policies. However, it does not appear that either the spillover effects or the policy slippage due to the pumping fee were accounted for in \ac{CREP} plans. There are two unknowns pushing in opposite directions. A policy planner making choices without this knowledge would overestimate the policy benefits of \ac{CREP} by 33\%\footnote{They would overestimate the direct reduction of \ac{CREP} wells by 61.6\%}. In other settings the gains from \ac{CREP} are likely to be underestimated, since there is evidence that it encourages neighbors to cooperate and reduce water output. However, it is important to consider local spatial attributes and overall enrollment levels, when estimating these effects in a \ac{PES} program. In this case study it was found that the effect of adding 32 wells to \ac{CREP} only increased neighborhood effects by 3.27\%.



The \ac{CREP} benchmarks can also be compared to the stated goals of the program. Looking at the year with the largest \ac{CREP} reduction the \ac{CREP} policy has achieved 17\% of the 60,000 \ac{AF} per year savings target. With the acreage intensity falling -0.49 \(\frac{\ac{AF}}{acre}\) short of the rate needed to reach the goal after full enrollment of 40,000 acres. However, the overall efforts of the farms in \ac{CREP} are relevant to outcomes. When evaluating the total amount of water saved by wells in \ac{CREP} the intensity of savings is 0.80\(\frac{\ac{AF}}{acre}\) above the rate needed to reach the \ac{CREP} program goals. The owners of \ac{CREP} wells have contributed to a more stable aquifer even if much of the conservation efforts were made prior to the \ac{CREP} program starting. 



%\FloatBarrier
\begin{landscape}%
        \centering
    \begin{figure}[h]%
        \caption{Cumulative effect of conservation policies }
        \label{FIG:POLICYEFF}
        \includegraphics[width=1.3\textwidth]{POLICY_COUNTER_FACT}
    \end{figure}%
\end{landscape}%
%\FloatBarrier

The net expected water savings due to conservation efforts are provided in \cref{FIG:POLICYEFF}. Despite the declining aquifer levels, the subdistrict has been successful in reducing the volume of water pumped. The upticks in water use after 2015 does physically lower the water table, but these cumulative results suggest that without the subdistrict policies an even more severe drawdown would have occurred. Comparable wells outside the subdistrict responded to climate, prices, and other factors during this period by increasing groundwater use. The marginal value of groundwater increased, partially reflected in the subdistrict's need to raise pumping fee rates.

For policymakers this provides an important case study for developing \ac{PES} programs. While more conservation may always seem better, the existence of highly efficient water saving policies are identified as reducing \ac{PES} effectiveness. The Pigouvian tax on water use was found to increase enrollment in the \ac{PES} while also lowering overall program savings. Complications in the enrollment structure led to these results. Tailoring program goals, and enrollment criteria to the \ac{PES} region is required to avoid unintended effects from policy interaction.  Such programs have pro-social benefits of encouraging neighbors to conserve water, but once again complicating factors can arise. The spatial distribution of parcels is a major factor in outcomes of these neighborhood effects. Taken together, these outcomes provide insights into the complexities of conservation policy interactions.