% Compile this document after all results are produced by 0_run.do using Stata. 
\documentclass{article}
\usepackage[paper=a4paper,landscape,margin=1in]{geometry}
\usepackage{graphicx}
\usepackage{subcaption}
\usepackage{placeins}
\usepackage{amsmath}
\newcommand{\sym}[1]{\ifmmode^{#1}\else\(^{#1}\)\fi} % for *** stars
\begin{document}

\begin{figure}[p]
	\centering
	\includegraphics[trim=150px 0px 120px 130px, clip, width=0.75\textwidth]{Results/Figure1}
	\caption*{Figure 1: Plot of horizontally drilled wells contained in the data set. Blue dots mark the locations of horizontally drilled wells. Grey patches correspond to shale formations and black solid lines are crude oil pipelines. Shape files for shale formations and pipeline infrastructure are provided by the U.S. Energy Information Administration. Map is constructed using Esri World Topographic Map.}
\end{figure}

\begin{figure}[p]
	\centering
	\includegraphics[width=\textwidth]{Results/Figure2}
	\caption*{Figure 2: Examples of well production profiles for different production technologies. Horizontal refers to shale wells and vertical to conventional wells.}
\end{figure}
	
\begin{figure}[p]
	\centering
	\includegraphics[width=\textwidth]{Results/Figure3}
	\caption*{Figure 3: Mean production profiles of conventional and shale wells. Each data point is computed by taking the mean output across all wells at the same point in their respective life cycles. Only wells that began production between 2005:M01 and 2019:M12 are included in the computations.}
\end{figure}

\begin{figure}[p]
	\centering
	\begin{subfigure}[b]{0.495\textwidth}
		\centering
		\includegraphics[width=\textwidth]{Results/Figure4a}
		\caption*{Figure 4a: Completion event}
	\end{subfigure}
	\hfill
	\begin{subfigure}[b]{0.495\textwidth}
		\centering
		\includegraphics[width=\textwidth]{Results/Figure4b}
		\caption*{Figure 4b: Refracturing events}
		\label{fig:corr_refracturing}
	\end{subfigure}
	\caption*{Figure 4: Binned scatterplot showing the relationship between the 3-month spot-futures spread and the maximum output of all wells during their first 6 months of production. Each of the 20 bins contain the same amount of events. Figure 4a shows this relationship at the time of completion. Figure 4b shows this relationship at the time of a refracturing event. Some wells are restimulated more than once.}
\end{figure}

\FloatBarrier

\input{Results/table1_shale.tex}

\input{Results/table1_conventional.tex}

\begin{table}[p]
	\centering
	\caption*{Table 2 (shale only). Shale vs. conventional wells}
	\begin{tabular}{lcc}
		\input{Results/table2_shale.tex}
	\end{tabular}
	\caption*{Estimation results for various baseline model specifications. $\eta_{oil}$ is the coefficient on the WTI spot price, $\eta_{F}$ is on the spot-futures spread, $\eta_{gas}$ is on the Henry Hub gas spot price and the coefficient $(\eta_{oil} + \eta_{F})$ is the sum of the spot and spread coefficients estimated by an auxiliary regression. Columns 1 and 2 are for models in log-level and 3 and 4 on growth form. Columns 1 and 3 are for shale wells while columns 2 and 4 are for vertically drilled (conventional) Texas wells. We include a measure of the monthly growth rate in U.S. crude oil inventories. The macro controls consist of the term spread, the trade-weighted foreign exchange rate, the Chicago Fed National Activity Index, the copper price, the MSCI world stock index, a measure of geopolitical risk and the VIX index. The cubic spline has knots at every 12th production month. N refers to the number of unique wells included in the estimation.}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table 2 (Conventional only). Shale vs. conventional wells}
	\begin{tabular}{lcc}
		\input{Results/table2_conventional.tex}
	\end{tabular}
	\caption*{Estimation results for various baseline model specifications. $\eta_{oil}$ is the coefficient on the WTI spot price, $\eta_{F}$ is on the spot-futures spread, $\eta_{gas}$ is on the Henry Hub gas spot price and the coefficient $(\eta_{oil} + \eta_{F})$ is the sum of the spot and spread coefficients estimated by an auxiliary regression. Columns 1 and 2 are for models in log-level and 3 and 4 on growth form. Columns 1 and 3 are for shale wells while columns 2 and 4 are for vertically drilled (conventional) Texas wells. We include a measure of the monthly growth rate in U.S. crude oil inventories. The macro controls consist of the term spread, the trade-weighted foreign exchange rate, the Chicago Fed National Activity Index, the copper price, the MSCI world stock index, a measure of geopolitical risk and the VIX index. The cubic spline has knots at every 12th production month. N refers to the number of unique wells included in the estimation.}
\end{table}

\input{Results/table3.tex}

\begin{table}[p]
	\centering
	\caption*{Table 4. Regression results on log-level state-level data}
	\begin{tabular}{lccccccccccc}
		\input{Results/table4.tex}
	\end{tabular}
	\caption*{Table 4: Estimation results for each individual U.S. state with data on log-level. Parameters $\eta_{oil}$ and $\eta_{gas}$ are the coefficients on the natural log of WTI and Henry Hub spot prices. $\eta_{F}$ is the coefficient on natural log of the spot-futures spread. $(\eta_{oil} + \eta_{F})$ is estimated by an auxiliary model. All wells are shale wells.}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table 5. Estimation results for shale play level data}
	\begin{tabular}{lccccc}
		\input{Results/table5.tex}
	\end{tabular}
	\caption*{Estimation results for each individual U.S. shale plays with data on log-level. A shale play is a geological formation where unconventional oil reserves are prevalent. Parameters $\eta_{oil}$ and $\eta_{gas}$ are the coefficients on the natural log of WTI and Henry Hub spot prices. $\eta_{F}$ is the coefficient on natural log of the spot-futures spread. $(\eta_{oil} + \eta_{F})$ is estimated by an auxiliary model. All wells are shale wells. Bakken and Permian shale plays cross state lines and we exclude Montana and New Mexico from the estimations, respectively. This to eliminate any possible confounding factors the different jurisdictions can cause.}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table 6. Regression results accounting for firm size and publicly traded firms}
	\begin{tabular}{lcc}
		\input{Results/table6.tex}
	\end{tabular}
	%\caption*{Panel A shows estimation results for a well-level model where we have interacted the oil prices with a dummy variable for whether the well is owned by a large oil firm. In particular, we use the distribution of well ownership to identify the firms in the dataset that are among the top 25\% in term of number of wells in operation. These are Chevron, ConocoPhillips, Continental Resources, Devon Energy, EOG Resources, ExxonMobil and Occidental Petroleum (Oxy). If well ${i}$ is operated by on of these firms, the dummy variable $large_{i}$ is equal to 1. Parameters $\eta_{oil}$ and $\eta_{gas}$ are the coefficients on the natural log of WTI and Henry Hub spot prices. $\eta_{F}$ is the coefficient on natural log of the spot-futures spread. $(\eta_{oil} + \eta_{F})$ is the total price response estimated by an auxiliary model. $(\eta_{oil}^{'}+\eta_{F}^{'})$ is the total additional price response from the interaction terms. Panel B shows estimation results from a well-level model where we have interacted the oil prices with dummy variable $public_{i}$ which is equal to 1 if well $i$ is owned by a publicly traded firm.}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table 7. Regression results for completion and refracturing}
	\begin{tabular}{lcccc}
		\input{Results/table7.tex}
	\end{tabular}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table 8. Estimation results with a sample including the COVID-19 pandemic period}
	\begin{tabular}{lccc}
		\input{Results/table8.tex}
	\end{tabular}
\end{table}

\begin{figure}[p]
	\centering
	\includegraphics[width=0.8\linewidth]{Results/Figure5}
	\caption*{Figure 5. Black dotted line shows the evolution of the point estimate of $(\eta_{oil}+\eta_{F})$ as the sample window is expanded. Grey dashed lines are 95\% confidence intervals.}
\end{figure}

\begin{table}[p]
	\centering
	\caption*{Table A.1. Role of spot futures spread in baseline model}
	\begin{tabular}{lcccc}
		\input{Results/tablea1.tex}
	\end{tabular}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table A.2. Estimation with 2012 sample restriction}
	\begin{tabular}{lcc}
		\input{Results/tablea2.tex}
	\end{tabular}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table A.3. Regression results for spot-futures spreads at longer horizons: Shale only}
	\begin{tabular}{lcc}
		\input{Results/tablea3_shale.tex}
	\end{tabular}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table A.3. Regression results for spot-futures spreads at longer horizons: Conventional only}
	\begin{tabular}{lcc}
		\input{Results/tablea3_conventional.tex}
	\end{tabular}
\end{table}

\begin{table}[p]
	\centering
	\caption*{Table A.4. Regression results for spot-futures spreads based on oil price expectations}
	\begin{tabular}{lccc}
		\input{Results/tablea4.tex}
	\end{tabular}
\end{table}

\end{document}