Agriculture, Dams, and Weather

Item Type text; Electronic Dissertation

Authors Mirghasemi, Seyedeh Soudeh

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 29/09/2021 10:35:42

Link to Item http://hdl.handle.net/10150/579110 AGRICULTURE, DAMS, AND WEATHER

Seyedeh Soudeh Mirghasemi

BY: =

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF ECONOMICS

In Partial Fulﬁllment of the Requirements For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2015 2

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Seyedeh Soudeh Mirghasemi, entitled Agriculture, Dams, and Weather and recommend that it be accepted as fulﬁlling the dissertation requirement for the Degree of Doctor of Philosophy.

Date: 29 July 2015 Price Fishback

Date: 29 July 2015 Ashley Langer

Date: 29 July 2015 Derek Lemoine

Date: 29 July 2015 Jessamyn Schaller

Date: 29 July 2015

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the ﬁnal copies of the dissertation to the Graduate College. I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulﬁlling the dissertation requirement.

Date: 29 July 2015 Dissertation Director: Price Fishback 3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulﬁllment of requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that accurate acknowledgment of source is made. This work is licensed under the Creative Commons Attribution-No Derivative Works 3.0 United States Li- cense. To view a copy of this license, visit http://creativecommons.org/licenses/by- nd/3.0/us/ or send a letter to Creative Commons, 171 Second Street, Suite 300, San Francisco, California, 94105, USA.

SIGNED: Seyedeh Soudeh Mirghasemi 4

ACKNOWLEDGEMENTS

This journey would not have been possible without such helpful, supportive, and patient committee members, such as Ashley Langer, Derek Lemoine, Jessamyn Schaller, and my amazing adviser, Price Fishback. Thank you all for your guidance and encouragement throughout my graduate studies. I am also grateful for the invaluable support and advice from Sandy Dall’erba, Gautam Gowrisankaran, Gary Richardson, Andrew Gahan, Francina Dominguez, the members of Arizona History Workshop, and the Arizona Environmental Energy Group. Further- more, I would like to thank the Economic History Association, the University of Arizona, the Institute of the Environment, and the University of Arizona’s Department of Economics for their grants and support. Finally, I would like to thank all of my classmates and friends. Graduate life is far easier when you are surrounded by great people and the support they provide. All errors are my own.

Special thanks to my love, Bardiya, my caring parents, Marzieh and Mahmoud, and amazing family members. Thank you for being there and for believing in me and my success. 5

DEDICATION

To my beloved Bardiya, and my lovely parents, Marzieh and Mahmoud 6

TABLE OF CONTENTS

LIST OF FIGURES...... 8

LIST OF TABLES...... 9

ABSTRACT...... 11

CHAPTER 1 Philosopher’s Concrete: Dam Construction, Farmland Values and Agricultural Production in the Western U.S., 1890 - 1920...... 13 1.1 Introduction...... 13 1.2 Historical Background...... 17 1.3 Data...... 19 1.3.1 Census of Agriculture Data...... 19 1.3.2 Major Dams...... 20 1.3.3 Presidential Elections...... 21 1.3.4 Climate and Geographical Data...... 22 1.3.5 Soil Data...... 22 1.4 Empirical Strategy...... 23 1.4.1 Empirical Model...... 23 1.4.2 IV strategy...... 26 1.5 Results...... 28 1.6 Conclusion...... 31

CHAPTER 2 Politics and Dam Construction: Historical Evidence from the Western U.S...... 45 2.1 Introduction...... 45 2.2 Historical Background...... 46 2.2.1 National Irrigation Congress Role...... 46 2.2.2 Passage of Irrigation Bill...... 49 2.2.3 Bureau of Reclamation...... 52 2.3 Data...... 53 2.3.1 Major Dams...... 53 2.3.2 Presidential Election...... 55 2.3.3 Geographic Characteristics...... 55 2.4 Empirical Strategies and Results...... 56 2.4.1 Bureau in 1910...... 56 2.4.2 Results...... 57 7

TABLE OF CONTENTS – Continued

2.4.3 Army Corps of Engineers...... 58 2.4.4 Corps versus Bureau...... 61 2.5 Conclusion...... 62

CHAPTER 3 The Impact of Climate Change on Agriculture: Accounting for Climate Zones in the Ricardian Approach...... 76 3.1 Introduction...... 76 3.2 The Ricardian Setting...... 82 3.3 Data...... 84 3.4 Results...... 87 3.4.1 Climate Regions...... 87 3.4.2 The Results for the Past (1997-2007)...... 87 3.5 Conclusion...... 89

REFERENCES...... 109 8

LIST OF FIGURES

1.1 Number of Dams Constructed by Different Types of Owner...... 42 1.2 Percentage of Dams Constructed by Different Types of Owner.... 42 1.3 Mean of the Height of Dams Constructed by Different Types of Owner in Each Decade...... 43 1.4 Mean of the Maximum Storage of Dams Constructed by Different Types of Owner in Each Decade...... 43 1.5 Percent Change in Farm Value per Acre (1900-1910)...... 44 1.6 Percent Change in Farm Value per Acre (1910-1920)...... 44

2.1 Dams - Bureau and Corps...... 70 2.2 Dams Constructed by the Bureau: (a) One Purpose: Irrigation, (b) Multi Purposes: Irrigation - Hydroelectric, (c) Multi Purposes: Irri- gation - Recreation...... 71 2.3 Dams Constructed by the Corps: (a) One Purpose: Flood Control, (b) Multi Purposes: Flood Control - Hydroelectric, (c) Multi Pur- poses: Flood Control - Recreation...... 72 2.4 Dams Constructed by the Bureau and Corps...... 73 2.5 Dams Constructed by the Corps...... 74 2.6 Dams Constructed by the Bureau...... 75

3.1 Climate Regions...... 107 3.2 Climate Data...... 108 9

LIST OF TABLES

1.1 Summary Statistics, 1900 - 1920...... 33 1.2 Summary Statistics - 1900...... 34 1.3 Pre Trend Test, 1890-1900...... 35 1.4 Federal Major Dams in the West...... 36 1.5 Primary Purpose...... 36 1.6 Association of the % of Votes for Republican in Presidential Elections and Dam Construction...... 37 1.7 Fixed Effect Results: Impact of a Newly Constructed Dam on the Natural Log of the Value of the Land per Acre 1900-1920...... 38 1.8 Fixed Effect and Instrumental Variable Results, Impact of a Newly Constructed Dam on the Natural Log of the Value of the Land per Acre 1900-1920...... 38 1.9 Fixed Effect and Instrumental Variable Results, Impact of a Newly Constructed Dam on the Natural Log of the Value of the Land per Acre 1900-1920...... 39 1.10 Area Irrigated, Capital Invested...... 39 1.11 Fixed Effect Results: Impact of a Newly Constructed Dam on Bushel per Acre and Acres Planted of Major Crops 1900-1920...... 40 1.12 Instrumental Variable Results, Impact of a Newly Constructed Dam on Bushel per Acre and Acres Planted of Major Crops 1900-1920... 40 1.13 Fixed Effect and Instrumental Variable Results, Impact of a Newly Constructed Dam on the Share of Acre Improved, Log of the Livestock per Acre, and Log of the Dairy Value per Acre 1900-1920...... 41 1.14 Instrumental Variable Results, Impact of a Newly Constructed Dam on the Share of Acre Improved, Log of the Livestock per Acre, and Log of the Dairy Value per Acre 1900-1920...... 41

2.1 Federal Dams in the West...... 63 2.2 Primary Purposes...... 64 2.3 Primary Purposes - Pre and Post 1936...... 65 2.4 Association of the % of Votes for Republicans in Presidential Election and Dam Construction...... 66 2.5 Montana and Idaho Dams...... 67 2.6 Logit Estimation...... 68 2.7 Purposes - Pre and Post 1936...... 69 10

LIST OF TABLES – Continued

3.1 Summary Statistics - Northwest...... 91 3.2 Summary Statistics - West...... 91 3.3 Summary Statistics - Southwest...... 92 3.4 Summary Statistics - West North Center...... 92 3.5 Summary Statistics - South...... 93 3.6 Summary Statistics - Southeast...... 93 3.7 Summary Statistics - Center...... 94 3.8 Summary Statistics - East North Center...... 94 3.9 Summary Statistics - Northeast...... 95 3.10 Chow Test...... 96 3.11 Control Only for Fixed Effects in the Model...... 97 3.12 Effect of Climate on Land Values...... 98 3.13 Effect of Climate on Land Values - Including State-Year Fixed Effects 99 3.14 Effect of Climate on Land Values - Including County and State-Year Fixed Effects...... 100 15A Effect of Climate on Land Values - Including State-Year Fixed Effects - Reference Group: West4...... 101 15B Effect of Climate on Land Values - Including County and State-Year Fixed Effects - Reference Group: West4...... 101 16A Effect of Climate on Land Values - Including State-Year Fixed Effects - Reference Group: South...... 102 16B Effect of Climate on Land Values - Including County and State-Year Fixed Effects - Reference Group: South...... 102 17A Effect of Climate on Land Values - Including State-Year Fixed Effects - Reference Group: Southeast...... 103 17B Effect of Climate on Land Values - Including County and State-Year Fixed Effects - Reference Group: Southeast...... 103 18A Effect of Climate on Land Values - Including State-Year Fixed Effects - Reference Group: Center...... 104 18B Effect of Climate on Land Values - Including County and State-Year Fixed Effects - Reference Group: Center...... 104 19A Effect of Climate on Land Values - Including State-Year Fixed Effects - Reference Group: East North Center...... 105 19B Effect of Climate on Land Values - Including County and State-Year Fixed Effects - Reference Group: East North Center...... 105 20A Effect of Climate on Land Values - Including State-Year Fixed Effects - Reference Group: Northeast...... 106 20B Effect of Climate on Land Values - Including County and State-Year Fixed Effects - Reference Group: Northeast...... 106 11

ABSTRACT

The first chapter investigates whether construction of the Bureau of Reclamation dams in the early twentieth century raised farm values and increased agricultural output. I construct a new county-level panel data set from 1890 to 1920 with information on geography, climate, politics, agriculture, and major dams and then evaluate the effect of the Bureau of Reclamation dams on the value of farms and on crop productivity. Using fixed effect panel estimation, I find that new federal dam construction increased the average value of farmland by approximately 6.4 percent. When I apply an instrument to control for potential endogeneity, the effect of Bureau dams on the farmland value increases in size, although the estimate is no longer statistically significant. When examining the crop output, the only crop for which the dams had effects was alfalfa. In the second chapter I investigate the effect of the geographic, economic and political factors on dam construction at the beginning of the Bureau of Reclamation’s operation in the American West. Applying county-level data which have been linked from various data sources for the time period of 1900 to 1910, I show that the percentage of votes for Republicans in presidential elections has a significant and positive effect on major dam construction. The last chapter investigates the effect of climate change on U.S. agriculture using county-level data from 1997 to 2007. Compared to previous contributions, we pay particular attention to the spatial heterogeneity across the climate zones and include the presence of extreme weather events. The lack of consideration for both effects may have led previous works to generate biased estimates and incorrect impact forecasts. Further, while the current approaches use projected climate variables derived from coarse resolution Global Climate Models (GCMs), we use data at a much finer resolution by relying on dynamically downscaled simulation data. 12

Chow-Wald tests indicate the presence of signiﬁcant heterogeneity across zones in the eﬀects of climate on land values. 13

CHAPTER 1

Philosopher’s Concrete: Dam Construction, Farmland Values and Agricultural Production in the Western U.S., 1890 - 1920

“The philosopher’s stone is really the philosophical stone, for philosophy is truly likened to a magic jewel whose touch transmutes base substances into priceless gems like itself.” Manly P. Hall - The Secret Teachings of all Ages

“We had pushed aside foreign countries and native peoples. Now we would push aside the desert.” Bruce Reichert - The Bureau that changed the West

1.1 Introduction

Dam construction played a major role in the development of water resources during the early 20th century in the American West. Over the ﬁrst half of the 20th century, the number of major dams in the west and their maximum capacity increased nearly sixteen-fold and two hundred-fold, respectively. At the same time, the rate of population growth in the West was at least double the rate in the rest of the U.S.1 This was associated with a greater than four-fold increase in national agricultural output from 1900 to 1950. The Newland Reclamation Act of 1902 created the Bureau of Reclamation, which built the lion’s share of dam capacity for irrigation in the West. The initial aim of the act was to improve agricultural production by providing irrigation to arid areas. However, the eﬀect of dam construction on agricultural growth in the west remains

1Except for the 1930s 14

controversial. I construct an extensive historical county-level dataset from 1890 to 1920 and examine the effect of the Bureau’s dam construction on the value of farmland and crop production in the West. Using the the data, I estimate the average gain in agricultural outcomes from treatment for those places that were treated (i.e., treatment on treated) and provide direct evidence that dams had a positive effect on some agricultural outcomes. Whether large public dams tend to have net benefits has been a controversial topic. Studies in the U.S. have found positive effects of hydroelectric dams on the local population and employment growth (Severnini(2014)), on county income and earnings growth (Aleseyed et al.(1998)), and on agricultural productivity (Hansen et al.(2011)). Duflo and Pande(2007) find that dams have reduced poverty and raised productivity in India in the modern era. On the other hand, Eckstein(1958) shows that the benefit of water resource development varies by the location of the site and as characteristics of the region change. Kitchens(2014) investigates the effect of electrification of the Tennessee Valley Authority’s (TVA) large scale hydroelectric dams on economics activities. Comparing the counties with or without hydroelectric dams2, he finds the TVA had an insignificant effect on economic growth in the Southeastern United States. Further, Howe(1968) finds that public investments in waterway improvements do not lead to rapid local economic growth. Reisner (1993), in the highly influential Cadillac Desert, states that due to political pressures and only a shallow understanding of land productivity, climate conditions, and the economic environment, it was mostly political connections that influenced the water projects authorized by Congress (Reisner(1993)). Therefore, the locations of the irrigation projects were determined in great haste and without comprehensive examination.3 One finding consistent with these views is the fact that the Bureau’s share of capital invested in the irrigation projects was around 19 percent, while its share of the total acres irrigated was only about seven percent in 1920.4

2With potential to have hydropower dam 3In fact, during the ﬁrst four years of federal investment in irrigation, 27 projects were authorized, but four of them were abandoned later Widtsoe(1928). 4U.S. Census of Agriculture (1920) 15

Factors influencing farm value have been studied extensively. The traditional approach assumes that farm value measures the discounted anticipated returns to agricultural production (Featherstone and Baker(1987); Burt(1986); Castle and Hoch(1982)). However, some studies show that the market value of farmland might exceed its agricultural production value as a result of urban proximity and potential for recreational use (Barnard(2000)). Other studies estimate farm value based on the potential development and conversion to urban use (Plantinga et al.(2002); Livanis et al.(2006)). Furthermore, studies have investigated the effect of infrastructure investments other than dam construction such as the effect of an expanded railroad network on agricultural land values (Donaldson and Hornbeck(2013), Fogel (1994); Atack and Margo(2011)). Donaldson and Hornbeck(2013) estimate that, in the absence of railroad investments, farmland value in the U.S would have been 64 percent lower. To advance the debate about the net benefits of major federal dams, I investigate the impact of the Bureau of Reclamation’s dam projects on the local economy and agricultural activity from 1890 to 1920. To the best of my knowledge, this is the first quantitative study that attempts to assess the effect of the Bureau of Reclamation dams on agricultural activities. I develop a new historical county-level panel dataset for the census years 1890, 1900, 1910, and 1920 with information on geography, climate, politics, agriculture, as well as with information on Bureau of Reclamation dams and other major dams. I use the data to measure the effect of Bureau dams on farmland value and crop production. Using fixed effects panel estimation, I find that new federal dam construction increased the average value of farmland in the county by approximately 6.4 percent. The estimation results indicate that new dams constructed by agencies other than the Bureau did not have statistically significant effects on the outcomes. This is reasonable, as the Bureau projects entailed vast federal investments compared to the dam construction by other entities. This can be explained by the much larger maximum capacity of the Federal dams compared to other agencies’ dams. Further- more, Bureau dams constructed in the previous decade (1890-1900) did not have 16 any effect on the value of farms. One potential source of endogeneity is that counties that lobbied for the Bureau dams might have anticipated that their agricultural sector would grow faster. I test whether there were differential pre-trends in economic activities in the Bureau counties before federal construction, and I do not find this to be the case. I find suggestive evidence that in the first 20 years of the governmental irrigation movement, federal dams were located in less densely populated areas and in areas where farm value per acre was decreasing. Nonetheless, as a way to further reduce endogeneity, I develop an instrumental variable approach. The instrument is defined as whether the county had the potential to have a Bureau dam, interacted with the political strength of Republicans in the two presidential elections before the Reclamation Act was passed. When I apply the instrument to control for endogeneity, the effect of Bureau dams on farmland value increases in size, although larger standard errors mean that it is no longer statistically significant at the ten percent level. To examine whether the Bureau’s dam construction affected agricultural activities, I estimate models with the production per acre and the average number of acres planted with important crops, the value of livestock and dairy, and the share of improved acres as outcome variables. Looking at the impact of dam construction on the average crop production per bushel, alfalfa is the only crop with a positive and statistically significant coefficient; however, this crop had been actively produced before the dam construction. This verifies that the original dam’s site had low soil quality in addition to being arid because in lands with low nitrogen levels, which result in low quality of most agricultural crops, it is necessary to first plant crops such as alfalfa. These findings are consistent with the narratives of the projects for this period. Christopher McCune, in the Belle Fourche Project Bureau of Reclamation Report, states:

Beginning in 1915, farmers increasingly turned to stock operations, mostly sheep, to try to turn a proﬁt, as alfalfa became the primary crop of the project...One of the ﬁrst reports given on the project lands stated that grain, hay, alfalfa, and perhaps small fruits will constitute 17

the main crops, which was not much diﬀerent than what had already been grown in the region for several years (McCune(2001)).

Furthermore, I do not find a statistically significant effect of dam construction on the value of livestock and the value of dairy products. In this paper, I focus specifically on the Bureau investments, as they were a turning point in the roles played by the private sector and the federal government in the West. The Bureau was created after the government passed the Reclamation Act to allow the government to build larger projects due to lack of finances and engineering skills in the private sector. My results support the fact that dams had positive effects on the local economy, but only in limited ways. These effects might not have been sufficient but they were potentially important for the West.

1.2 Historical Background

Most of the development of the U.S. occurred in the East, and the Western U.S. was comparatively underdeveloped until the end of the 19th century. Although the Western U.S. provided abundant land for raising crops and livestock, farmers found the climate arid and sought new, large-scale irrigation methods to develop the land. The Federal Desert Land Act, also known as the Carey Act, gave permission to private companies in the U.S. to assemble irrigation systems in the Western states and to proﬁt from the sales of water to the irrigators. Congress passed this Act on August 18, 1894, as the federal government decided that the task of irrigation was too large for individual settlers. The new Act delineated a new approach for the disposal of public desert land. The private sector attempted to evaluate these lands in the Western U.S. to ﬁnd an opportunity to establish an agricultural society. Except for in Idaho and Wyoming, the Carey Act was not as successful as intended.5

5In 1908, Idaho received an additional two million acres (8,000 km2) and Wyoming received an additional one million acres (4,000 km2) of land to develop under the Carey Act. Today, approximately six percent of the Carey Act lands irrigated in the United States are in Idaho. 18

Westerners argued for further action by the government to build larger projects due to lack of finances and engineering skill. In 1902, Congress passed the Reclamation Act, which made the federal government, in the form of the Bureau of Reclamation, responsible for irrigation in the U.S. Western states6 (Miller and Miller(1992)). The bill’s goal was to convert arid federal land into a suitable place for living, by constructing dams, power plants, canals, lateral systems, pumping plants, and other water facilities. Building of a dam required prerequisite construction, such as roads and railroad construction. The water projects were to be financed through a Reclamation Fund, which was funded by selling federal land and, later, by selling the water to the irrigators (Reis- ner 1986). To discover the feasibility of the water projects, the geological surveys were prepared by the Bureau of Reclamation, which included all related factors, such as the amount of water flow, elevation of the surface and the streams, and their catchment areas for the dam construction (Newell(1905)). Initially, the Bureau’s ambition was to boost agricultural activities and help the local economy by constructing water projects and delivering water to the arid areas. However, because of political pressure from Congressmen, Senators, and state legislators to acquire water projects, dams might have been constructed in the districts with little potential for agriculture. The water projects mostly had to be authorized by Congress, but the President could veto the bill. According to the history of the Bureau of Reclamation, Michael Robinson7 states:

Initially, little consideration was given to the hard realities of the irrigated agriculture. Neither aid nor direction was given to settlers in carry- ing out the diﬃcult and costly work of clearing and leveling the land, dig- ging irrigation ditches, building roads and houses, and transporting crops to remote markets....The government was immediately ﬂooded with re- quests for project investment as the Local chambers of commerce, real

6Western states served by Reclamation are Arizona, California, Colorado, Idaho, Kansas, Mon- tana, Nebraska, Nevada, New Mexico, North Dakota, Oklahoma, Oregon, South Dakota, Utah, Washington and Wyoming. Texas was not included during the ﬁrst few years. 7The son-in-law of a Commissioner of Reclamation 19

estate interests and congressman were convinced their areas were ideal for reclamation development (Reisner(1993)).

Twenty-four federal irrigation projects were authorized within four years of the passage of the Act, four of which were abandoned later. The ﬁnancing of the Reclamation projects compelled the farmers to meet their repayment obligation in ten years. This proved to be an unrealistic estimate, as 60 percent of the farmers delayed their payments. In some cases, the delays stretched beyond 20 years from the passage of the ﬁrst Reclamation law, and the repayment period was extended to 40 or 50 years.

1.3 Data

The new county-level dataset is assembled from several sources for the 1890 through 1920 census years 8 for the Western U.S.9

1.3.1 Census of Agriculture Data

The US Census of Agriculture data are reported for the following: value of farms, number of acres in farming, improved acres, value of dairy products, value of livestock, production of important crops, 10 and population.11 Table 1.1 shows the summary statistics of the data. For counties with or without Bureau dams, Bureau counties, on average, were less densely populated and had fewer farms compared to non-Bureau counties. Furthermore, the Bureau counties had statistically significantly more production per acre and a higher average number of acres planted with alfalfa, while the production per acre and the average number of acres planted with cotton were statistically significantly larger in non-Bureau counties. 8The U.S. Census of Agriculture data are available for every ten years during this time period. 9Due to the change of boundaries over the years, the 1900 shapefile is chosen as a base year, and 1910 and 1920 weighted average values are calculated applying the Geographic Information System (GIS). 10Sugar beet, Cotton, Wheat, Alfalfa 11All dollar values are in 1926 constant dollars. 20

1.3.2 Major Dams

Information on dams comes from the National Inventory of Dams, Water Control Infrastructure for all the major dams constructed in the U.S. from 1800 to 2003. The data include information on the name, national ID, latitude, longitude, owner name, type of owner, year of completion, purposes and the primary purpose, capacity, height, and some other characteristics for the major dams in the U.S. The dataset includes 8,121 major dams. A major dam is 50 feet or more in height, has a normal storage capacity of 5,000 acre-feet or more, or has a maximum storage capacity of 25,000 acre-feet or more. The total number of dams constructed in each decade by diﬀerent owners is shown in Figure 1.1. There are ﬁve types of owners:

1. Federal: The dam is owned by a federal agency.

2. Local: The dam is owned by a county, city, regional, or other similar local government or government agency.

3. Private: The dam is owned by an individual or individuals, or by a private company.

4. State: The dam is owned by a state or by a state agency.

5. Public utilities: The dam is owned by a public utility, such as Southern Cali- fornia Edison Company, Paciﬁc Gas and Electric Company.

Figure 1.1 indicates that between 1900 and 1920, the number of dams constructed by the federal government increased, but that the number of private dams increased more. Figure 1.3 and Figure 1.4 show two characteristics of dams used to compare their size. Figure 1.3 compares the height of the dams constructed by various owners from 1880 to 2000. The tallest dams belong to the public utilities. Another char- acteristic of size is the maximum storage capacity, which is the total storage space in a reservoir below the maximum attainable water surface elevation, including any 21 surcharge storage. Clearly, federal dams have significantly larger maximum storages compared to the dams constructed by private entities and other type of owners. The summary statistics of the federal dams are presented in Table 1.4. Between 1900 and 1920, there were 66 major federal dams throughout the U.S., 54 of them constructed by the Bureau. Most of the non-Bureau dams at that time had been constructed by the Bureau of Indian Affairs. Idaho, followed by California, Montana and Wyoming, were the states with the most Bureau dams. The primary purposes of dam construction include flood control, debris control, fish and wildlife protection, hydroelectric generation, irrigation, navigation, fire protection, recreation, water supply enhancement, and tailings control. Table 1.5 shows the frequency of the primary purpose of the dam construction. Clearly, most of the dams built by the Bureau were intended for irrigation and water supply. Dams can be constructed for either single purpose or multiple purposes, with different geographical and topographical preferences. For irrigation dams, the river gradient should be neither steep nor flat, but dams for hydroelectric power need a higher river gradient (Cech(2010)). According to a study by Duflo and Pande (2007), “Low (but nonzero) river gradient areas are most suitable for irrigation dams while very steep river gradient areas are suitable for hydroelectric dams.”

1.3.3 Presidential Elections

The political data come from the ICPSR United States Historical Election Re- turns database. The data include the state-level percentage of votes for Republicans in presidential elections. The summary statistics for Western states in Table 1.6 show a strong relationship between Republican votes in the presidential elections of 1900, 1904, 1908 and the location of dams. We choose these elections since this time period was crucial, as the majority of the authorization dates of the early projects were during 1903-1908. Columns (2) through (4) show the percentages of Republican votes for the president in 1900, 1904 and 1908, and column (5) shows the average of the three elections. 22

The dam column is an indicator of whether or not a dam was constructed in a state during the relevant time period. South Dakota, Washington, Wyoming, Oregon and California remained Republican during the ﬁrst decade of the 20th century, and the Bureau constructed there.

1.3.4 Climate and Geographical Data

The climate data are obtained from the U.S. Historical Climatology Network for each weather station. The number of stations is limited, and the Geographic Information System (GIS) software is applied to interpolate the climate data and calculate values at the county level. Then, the average of the climate variables is calculated per decade. Speciﬁcally, the data include averages of extreme events: hot days that exceed 100 degrees Fahrenheit; hot days that exceed 90 degrees Fahrenheit; cold days below 32 degrees Fahrenheit; cold days below 0 degrees Fahrenheit; and total rainfall per year. Looking at Table 1.1, on average, Bureau counties compared to non-Bureau counties were located in higher altitudes and in mountainous areas and also had less annual precipitation, which is statistically signiﬁcant at the one-percent level.

1.3.5 Soil Data

The soil data are from the Web Soil Survey (WSS), which uses information from the National Cooperative Soil Survey. The soil survey was developed for polygons of areas with similar soil characteristics. The data are available for each state, and a special program, STATSGO, is needed to open and observe the data. The data are interpolated by the GIS software for the 1900 shapefile. The data include the fraction of the land prone to floods, the soil erodibility factor (K-Factor), slope steepness (S factor), wind erosion, the fraction of the land occupied by wetland, salinity, permeability, moisture capacity, clay content, and sand content.12 12The fraction of the land prone to floods is reported as a frequency variable, as none, very rare, rare, occasional, frequent, and very frequent, and it is not included in my analysis. 23

1.4 Empirical Strategy

Initially, the Bureau’s ambition was to boost agricultural activities and help the local economy by constructing water projects and delivering water to arid areas. These were the areas with poor soil and unsuitable climate condition where the Bureau expected to have a substantial effect on the agricultural activities. The treatment effect can be defined as whether or not the arid county received a dam. I estimate the average gain in agricultural outcomes treatment in those places that were treated (i.e., treatment on treated) (Imbens and Lancaster(1994), rather than estimating the effect of dam construction randomly (i.e., average treatment effect) (Holland(1986)). Estimating the average treatment effect is not applicable, as the dams are more likely to be constructed in places with geographical and topographical prerequisites and higher potential need for water. To the best of my knowledge, no other study estimates the effect of Bureau Reclamation dam construction on agricultural outcomes.

1.4.1 Empirical Model

I estimate the eﬀect of dam construction on farmland values and crop production using county-level data in the following regression:

Yit = β0 + β1Damit + β2Xit + δi + wst + εit (1.1)

The main outcome variable is the log of the value of the farm per acre in county i in census year t. The Dam variable is an indicator of whether the Bureau constructed a dam in county i in decade t, and β1 is the coefficient of interest. Xit, is a vector of control variables. Rainfall, and, hot and cold extreme weather events are included to control for climate conditions, and they are interacted with soil characteristics to control for the combined effect of soil and climate on outcomes. It is plausible that the value of farmland may have increased because of land speculation rather than improvement in agricultural activities. Therefore, I consider 24 the production per acre and the average number of acres of important crops planted, the value of livestock per acre, dairy value per acre, and the share of improved acres to reveal whether there was an effect on agricultural activities. I also include county fixed effects to capture the time-invariant unobserved characteristics related to each county. Year-state fixed effects are included to control for the shocks that occurred in the states in each year. εit is the unobserved error component. The identification of the effect of the dams comes from changes over time when a dam is added within the same county after controlling for the factors listed above. Figure 1.5 and Figure 1.6 show the variation of farm value growth rates at the county level for the first and second time period, respectively. The geographical figures also illustrate dams constructed by the Bureau in each period. Figure 1.5 illustrates the percent change in the farm value per acre from 1900 to 1910. The value of farms per acre in the Central and Western U.S. increased significantly during this period. The counties that received Bureau dams experienced increases in the value of the farmland per acre. Figure 1.6 shows the percent change in the farm value per acre during 1910 to 1920. Due to the effects of World War I and high inflation, the value of the land per acre did not increase in most of the U.S. during this period.

Endogeneity

To obtain an unbiased estimate when applying a fixed effects model, the unobserved error term must be uncorrelated with dam construction in each county. There are a few scenarios that could lead to the violation of this identification assumption. Construction of a dam required the building of roads and railroads, as most of the dam sites were in remote areas.13 This effect will be captured by the dam coefficient since the prerequisite constructions were part of the projects, and, therefore, I consider these changes to be part of the treatment effect.

13Such as Coulee Railroad between Coulee City and Coulee Dam in Washington, and road construction along Buﬀalo Bill Reservoir near Cody, Wyoming. 25

It is possible that dams were located in counties with high potential for agricultural activity, therefore not controlling for climate conditions leads to an over- estimation of the coefficient of interest. Including more controls, such as climate variables, helps to reduce the endogeneity problem. Educated and up-to-date farmers likely had access to newer and superior technology. At the same time, they could better lobby to bring dams to their areas. Since farmers’ education and knowledge have a positive effect on the agricultural outcome and are positively associated with having a dam constructed nearby, the coefficient of interest would be biased upwards. In order to address this issue I can control for general county education. Since these were rural areas, I find it is unlikely that farmers’ education differed much from the general measure of education. Some farms had access to groundwater to irrigate their lands. It is likely that dams were constructed to irrigate the areas in which construction of wells and access to groundwater were not possible. Thus, not having access to ground water is positively correlated with dam construction, while it is negatively correlated with the productivity of the land. This leads to an underestimate of the coefficient of interest. There is no solid evidence of any new developments in groundwater pumping technology between 1900 and 1920 (Hornbeck and Keskin(2012)). Therefore, the groundwater variable is time-invariant and can be captured by the county fixed effect. Furthermore, some might think that federal spending on other industries in a county might have attracted other industrial construction besides dams. However, the federal government was not spending much in the West at this time on anything other than dam projects. So it is unlikely that such factors would bias my estimates. To ensure that no other factors were contributing to endogeneity, I assess whether there were differential pre-trends in economic activities in Bureau counties versus non-Bureau counties before federal construction. The pre-trend test in a standard difference-in-difference procedure was applied for some of the important variables, such as population, population density, farm value, and farm value per acre, for the 1890-1900 period. 26

Table 1.3 shows the result of these comparisons: there are no statistically significant differences between the mean changes in the population, farm value per acre, and total farm value variables in non-Bureau and Bureau counties. There is a statistically significant difference between the mean growth trends of population density in Bureau and non-Bureau counties, but the mean growth trend is higher in non-Bureau counties. This would bias my coefficients towards finding no positive effect of Bureau dams. When looking at the growth rates (average of the changes), the null hypothesis of identical trends cannot be rejected for population and population density. For farm value per acre and total farm value, there is a statistically significant difference between the means. However, the Bureau counties had a more negative farm value per acre growth rate compared to non-Bureau counties. Also, the total farm value growth rate in non-Bureau counties was higher in Bureau counties. Hence, I find no evidence that the Bureau constructed dams in the counties with more economic activities prior to dam construction.

1.4.2 IV strategy

As a way to further reduce endogeneity, I have developed an instrumental variable approach. The study of the political economy of dam locations, Mirghasemi(2013) shows that the locations of the dams were strongly associated with the states’ average Republican vote share in presidential elections. According to the Bureau of Reclamation records, President Theodore Roosevelt strongly supported the ﬂour- ishing of the West and about 24 out of 27 projects were approved immediately after the Reclamation Act was passed.14 The average of the percentage of votes for Re- publicans in the 1896 and 1900 presidential elections can be used as a measure of the relative political power of the Republicans, who held the Presidency and had a majority in both houses of Congress for 18 consecutive years (from 1896 till 1912). The instrument is constructed based on geographical and political factors. The geographical factor involves the potential places where the Bureau could construct a

14The authorization dates of the projects were during 1903-1908 27

dam, while the political factor captures political strength. More speciﬁcally, the instrument is deﬁned as follows:

  (%Repub1896s+%Repub1900s) 1I{Dam1960i} ∗ 2 if t = 1910, 1920 Zi = 0 if t = 1900

The ﬁrst part of the instrument is a geographical factor for places with potential to have a dam. Dam1960i shows whether the Bureau constructed a dam in county i by 1960.15 The measure captures all of the locations where the Bureau might have expected to locate a dam, given the technology available through 1960. The second part of instrument captures the political strength of the Republicans for the two presentational elections before the Act was passed. It is the average state Republican vote share in the presidential elections of 1896 and 1900. Republican states were rewarded more when the President was Republican and Congress was

dominated by Republicans. The Zi value is interacted with the years 1910 and 1920 as all the projects had been authorized in the ﬁrst few years after the passage of the Reclamation Act. The identiﬁcation assumption is that the instrument is not correlated with the error term. This assumption makes sense, as the political component of the instrument is used for the period before the dam projects were authorized. The measure is based on voting information from periods of ten to 15 years before the impact of the dams. It is unlikely that there is serial correlation that stretches so many years back in time.

The following is the ﬁrst-stage equation:

Damit = β0 + β1Xit + β2Zit + δi + wst + εit (1.2)

X is the vector of the economic and geographical characteristics from Equation

15The year 1960 is chosen since, after this decade, the number of federal dams constructed started decreasing. Changing the year still keeps the IV as an valid instrument and does not alter the results. 28

1. Z is the instrumental variable. δi is the vector of county fixed effects to capture the time-invariant unobserved characteristics related to each county. wst is the year-state interaction fixed effects to control for the shocks that happened within the states, and εist is the unobserved error component.

1.5 Results

Table 1.7 displays the regression results for the log of the value of the farm per acre as an independent variable. The dam variable in the model is an indicator of whether the Bureau constructed a dam in county i during the decade before census year t.16 The first column of the results shows the baseline model controlling for county and time fixed effects. The dam coefficient is positive but not statistically significant. The coefficient indicates that for each newly constructed dam in a county, there is an increase in farm value by roughly six percent of the mean farmland value in the same county. Adding climate controls in the second column increases the size of the dam coefficient, but it remains statistically insignificant. Column 3 indicates the results of the estimation after adding interactions between soil characteristics and climate variables to the model. The dam coefficient increases to 15 percent of the mean farmland value, and it becomes statistically significant at the one-percent level. The change in the coefficient shows that both the dam location and the farmland values are influenced by the interaction between weather and soil, in ways that lead to negative omitted variable bias for coefficients in specifications (1) and (2). When the state year fixed effects are added to the model in the last column, the dam coefficient remains statistically significant and increases to 19 percent. The results of the IV estimation are shown in Table 1.8. The first column is the same as the fourth column of Table 1.7. The Kleibergen-Paap F statistic of 17.3 for the instrument in the first stage shows that the instrument is strong. Applying the IV in the second column makes the dam coefficient larger, but it is not statistically significant.

16I have reestimated the model if the dam variable is the number of new dams constructed in each county instead of a binary variable. The results are robust. 29

Table 1.9 shows the Fixed Effects and IV estimation results after adding the lag of the dam variable and non-Bureau dams separately to the model. The first and second columns display the same specification as the fourth column of Table 1.7, except that they control for the impact of the dam constructed in both the current period and previous period by including the Lag Bureau in the model. The fixed effect estimation results in column 1 show a statistically significant effect of Bureau dams on the farmland value; however, the lag of Bureau dams variable does not have a statistically positive effect on the lag of the value of the farm. The sign and magnitude of the dam coefficient are similar to those in the last column of Table 1.7. Applying the instrument in the second column increases the dam coefficient in size, but it is no longer statistically significant. The third and fourth columns show the same specification as the fourth column of Table 1.7, except that they control for the impact of the dams constructed by other group, including other federal dams, state dams, and private dams. The third column shows the result of the Fixed Effect estimation, and the sign and magnitude of the Bureau dam coefficient are similar to those in the last column of Table 1.7. The new non-Bureau dams do not seem to have a significant impact on the value of the land. When I apply an instrument to control for endogeneity, the effect of Bureau dams on farmland value increases in magnitude, although larger standard errors mean that it is no longer statistically significant at the 10-percent level. This is reasonable, as the Bureau projects were vast federal investments compared to the dam construction by other entities. The narrative shows that land speculation occurred in some cases and led to increases in the price of the land without it being developed.17 However, the IV results here cannot reject the hy-

17Belle Fourche Project report (McCune(2001)): “Land speculation was practiced by several project settlers, who would buy, but not develop lands in anticipation of selling them for greater profit in the future. This led to even more instances of farmers going broke and selling out. Newell, in his report remarked that “few newcomers can handle effectively even forty acres. If they obtain a larger area, they must struggle to pay taxes and water charges with the products of less than forty acres or with the scanty returns of lands ineffectively tilled.” Rio Grande Project report (Autobee(1994)): “Concurrent with the rise of the Elephant Butte Dam, prices for unimproved land shot up. At the beginning of construction in 1906, land averaged 30 pothesis of zero effect. Table 1.10 shows the total acres irrigated and the capital invested, as well as Bureau’s share for 1900, 1920, 1930, and 1940.18 The Bureau had 18.6, 21.7, and 25.8 percent of capital investments, respectively, in 1920, 1930, and 1940. However, the acres irrigated by the Bureau of Reclamation were only 2.7, 6.5, 7.6, and 8.7 percent in 1910, 1920, 1930, and 1940, respectively. To examine whether Bureau dam construction affected agricultural activities, I estimate models with the production per acre and the average number of acres of the important crops planted, the value of livestock and dairy, and the number of improved acres as outcome variables. The results of the estimation of the production per acre and the average number of acres of important crops planted are displayed in Tables 1.11 and 1.11. The fixed effect estimation of the dam coefficient in Table 1.11 is positive and statistically significant only for the alfalfa crops. Constructing a Bureau dam in a county increases the alfalfa production per acre and the average number of acres of alfalfa, respectively, by roughly six percent and one percent. Using IV estimation in Table 1.11, the dam coefficient remains statistically significant only when the outcome variable is alfalfa production per acre. Alfalfa was a crop that was actively produced in Bureau counties before dam construction as seen in Table 1.2, and dam construction, led only to increases in the alfalfa production per acre in these counties. These findings are consistent with the narratives for some of the projects of this period. Christopher J. McCune, in the Belle Fourche Project Bureau of Reclamation report, states:

Beginning in 1915, farmers increasingly turned to stock operations, mostly sheep, to try to and turn a proﬁt, as alfalfa became the primary crop of the project....One of the ﬁrst reports given on the project lands stated that Grain, hay, alfalfa, and perhaps small fruits will constitute

$17.50 an acre. Seven years later, the value of the same unimproved ground was $50 to $75 an acre. A few years later, developed orchard and garden tracts within 10 miles of El Paso sold for $650 to $1,200 an acre.” 18Unfortunately, the census data for the capital invested are not available for the 1900 census year. 31

the main crops, which was not much diﬀerent than what had already been grown in the region for several years (McCune(2001)).

Table 1.11 indicates that Bureau dam construction did not have a statistically significant effect on other crops, such as cotton, sugar beets, and wheat. Estimation results applying the instrument in Table 1.11 show that Bureau dam construction did not have statistically significant effect on sugar beets and wheat. Even though the dam coefficient for cotton, the average number of acres planted, is statistically significant, not much cotton was planted in the Bureau counties as shown in Table 1.2. Furthermore, the results of the Fixed Effect and IV estimations in Tables 1.13 and 1.13 for the value of livestock per acre, dairy value per acre, and the share of improved acres show that none of the estimations of the dam coefficients are statistically significant.

1.6 Conclusion

Did the construction of the Bureau dams in the early 20th century cause the desert to flower and increase the gold that could be earned from the land? In this paper, I develop a new county-level panel dataset from 1890 to 1920, including information on geography, climate, politics, and agriculture, and on the Bureau of Reclamation dams and other major dams. I use the data to evaluate the effect of the Bureau dams on the value of farms and crop productivity. The results indicate that for each newly constructed dam in a county, there is an increase in the value of the farm by 19.3 percent of the mean farmland value in the same county (approximately 6.4 percent). When I apply an instrument to control for potential endogeneity, the effect of Bureau dams on farmland value increases in size; however, the estimates also become noisier and are no longer statistically significant. The estimation results indicate that the new dams constructed by agencies other than the Bureau and the already constructed dams by the Bureau did not have a 32 statistically significant effect on the farm value. Furthermore, I estimate that the only crop that the dams affected was alfalfa, which had been actively produced before. In this study, I focus specifically on the Bureau investments, as they represent a turning point in water investments that shifted the source of dam funding from the private sector to the federal government in the West. The Bureau was created after the passage of the Reclamation Act to take further action to enable the federal government to build larger projects due to the of lack of financing and engineering skills in the private sector. My results support the fact that dams had a positive but limited effect on the local economy. 33

Table 1.1: Summary Statistics, 1900 - 1920

Bureau counties Non-Bureau counties Mean Std. Dev. Mean Std. Dev. Diﬀerence Population 11,293.5 8,772.4 16,414.2 34,060.3 -5,120.7 ∗∗∗ Population density 4.5 6.1 27.7 314.8 -23.2 ∗∗∗

Farm (number) 1,021.1 782.9 1,394.9 1,267.9 -373.8 ∗∗∗ Farm (acre) 399,798.8 301,466.8 416,488.8 307,142.4 -16,690.0 Acres improved 122,609.8 116,088.4 172,211.3 162,201.5 -49,601.4 ∗∗∗

Value of farmland 15,233,854.8 12,352,326.5 15,868,580.4 15,191,183.1 -634,725.6 Value of farmland/acre 49.0 40.6 45.7 55.9 3.3 Value of dairy 125,258.6 119,570.7 170,270.1 244,429.3 -45,011.5 ∗∗∗ Value of livestcok 3,050,205.22 1,819,528.67 2,262,063.35 1,623,395.70 788,141.87 ∗∗∗

Sugar beet (bushel/acre) 1,071.0 4,663.6 1,450.9 12,691.1 -380.0 Sugar beet (acre) 5,667.8 40,211.0 2,439.8 19,838.4 3,228.0 Cotton (bushel/acre) 43.1 364.6 3,319.1 9,792.1 -3,276.0 ∗∗∗ Cotton (acre) 86.0 710.7 11,724.5 32,699.1 -11,638.5 ∗∗∗ Wheat (bushel/acre) 354,318.1 872,999.4 437,975.6 864,168.6 -83,657.5 Wheat (acre) 20,078.2 43,132.2 33,462.9 61,269.9 -13,384.7 ∗∗∗ Alfalfa (bushel/acre) 33,140.3 45,309.6 10,700.1 22,141.0 22,440.2 ∗∗∗ Alfalfa (acre) 11,452.1 12,582.3 4,640.3 8,659.7 6,811.8 ∗∗∗

Elevation 3,232.5 1,648.7 2,272.3 1,811.9 960.2 ∗∗∗ Precipitation 3.0 3.8 9.4 9.9 -6.3 ∗∗∗ Days exceeds 100 F 0.8 1.6 2.9 5.9 -2.2 ∗∗∗ Days exceeds 90 F 5.1 6.5 20.7 29.1 -15.6 ∗∗∗ Days below 32 F 6.4 12.8 6.5 11.4 -0.2 Days below 0 F 0.5 1.6 0.3 1.2 0.2 Number of observation 78 2556 *** p<0.01 ** p<0.05 * p<0.1 34

Table 1.2: Summary Statistics - 1900

Bureau counties Non-Bureau counties Mean Std. Dev. Mean Std. Dev. Diﬀerence Population 7,196.3 5,093.9 12,444.9 19,357.0 -5,248.7 ∗∗∗ Population density 2.8 3.0 22.2 252.1 -19.4 ∗∗∗

Farm (number) 604.8 370.2 1,206.0 1,217.6 -601.3 ∗∗∗ Farm (acre) 340,019.0 306,937.1 384,294.4 314,680.0 -44,275.5 Acres improved 84,515.4 100,784.8 133,998.8 143,430.6 -49,483.3 ∗∗

Value of farmland 8,884,533.8 6,736,996.4 10,202,826.5 10,909,325.7 -1,318,292.6 Value of farmland/acre 33.8 17.8 33.2 36.7 0.7 Value of dairy 99,011.5 83,020.5 154,162.0 181,670.5 -55,150.5 ∗∗∗ Value of livestcok 2,993,432.6 1,983,555.7 2,213,805.9 1,666,527.1 779,626.7 ∗

Sugar beet (bushel/acre) 357.2 1,276.0 619.0 5,763.8 -261.8 Sugar beet (acre) 71.7 273.6 73.3 613.9 -1.6 Cotton (bushel/acre) 0.0 0.0 3,024.6 9,894.5 -3,024.6 ∗∗∗ Cotton (acre) 0.0 0.0 8,452.2 25,720.0 -8,452.2 ∗∗∗ Wheat (bushel/acre) 346,158.5 877,050.2 324,205.0 734,053.6 21,953.4 Wheat (acre) 19,115.8 47,253.8 25,961.8 53,120.3 -6,846.0 Alfalfa (bushel/acre) 13,558.7 15,169.5 5,629.5 14,430.9 7,929.1 ∗∗ Alfalfa (acre) 5,117.6 4,523.6 2,254.4 5,838.2 2,863.2 ∗∗∗

Elevation 3,232.5 1,670.5 2,272.3 1,812.6 960.2 ∗∗∗ Precipitation 1.1 2.1 5.2 7.0 -4.0 ∗∗∗ Days exceeds 100 F 0.3 0.7 0.6 1.8 -0.3 ∗∗ Days exceeds 90 F 1.8 3.5 7.2 11.6 -5.3 ∗∗∗ Days below 32 F 1.6 3.4 2.9 6.0 -1.3 ∗ Days below 0 F 0.09 0.31 0.04 0.18 0.06 Number of observation 26 852 *** p<0.01 ** p<0.05 * p<0.1 35

Table 1.3: Pre Trend Test, 1890-1900

Bureau counties Non-Bureau counties Mean SE Mean SE Diﬀerence t-test

Growth trend (change) Population density .5933242 .3931483 3.609915 1.168711 -3.016591∗∗ 2.446 Population 2036.63 687.0943 2448.817 187.8532 -412.1879 0.578 Farm value per acre -5.147041 1.337172 -3.656948 .466467 -1.490094 1.052 Farm value -288794.6 715602.7 522189.5 70920.24 -810984.1 1.127

Growth rate (% change) Population density .6177756 .1767042 .6273785 .0774735 -.0096029 0.049 Population .606753 .1771024 .6181378 .0770102 -.0113847 0.059 Farm value per acre -.3225601 .0685111 -.186341 .0149981 -.1362191∗ 1.942 Farm value .7791656 .2556491 1.444999 .1927996 -.6658336∗∗ 2.079 *** p<0.01 ** p<0.05 * p<0.1 36

Table 1.4: Federal Major Dams in the West

Federal dams Total Bureau Others 1902 - 1910 31 30 1 1910 - 1920 35 24 11 total 66 54 12

Note: Although Bureau is the major federal dam builder, most of the other dam construction was led by The Bureau of Indian Aﬀairs. The major dams are deﬁned based on the standard criterion by the National Inventory of Dams. These are the major dams, which are higher than 50 feet or have a normal capacity of at least 5,000 acre-feet or a maximum storage capacity of 25,000 or more.

Table 1.5: Primary Purpose

Primary purpose Total Bureau Irrigation or water supply 60 51 Flood control or navigation 2 1 Hydroelectric 1 1 Recreation 1 - Others 2 1

Note: The table shows the primary purposes of the dams constructed by Bureau during 1902-1920. The Others category includes debris control, ﬁsh and wildlife ponds, tailings and ﬁre protection, stock, or small farm pond purposes. 37

Table 1.6: Association of the % of Votes for Republican in Presidential Elections and Dam Construction

State 1900 1904 1908 Average Dam North Dakota 62.1 75.1 61 66.1 - South Dakota 56.8 71.1 58.8 62.2 yes Washington 53.4 70 57.8 60.4 yes Wyoming 58.6 66.9 55.4 60.3 yes Oregon 55.5 67.3 56.5 59.8 yes California 54.5 61.9 55.5 57.3 yes Kansas 52.6 64.9 52.5 56.7 - Utah 50.6 61.5 56.2 56.1 - Idaho 46.9 65.8 54.1 55.6 yes Nebraska 50.5 61.4 47.6 53.2 - Colorado 42 55.3 46.9 48.1 - Montana 39.8 53.5 46.9 46.7 yes Nevada 37.8 56.7 43.9 46.1 - Texas 30.9 22 22.4 25.1 - Arizona - - - - - New Mexico - - - - yes Oklahoma - - 43.5 - -

Arizona, New Mexico and Oklahoma were territories that did not have voting representation in the U.S. Congress and, therefore, had no electoral votes. 38

Table 1.7: Fixed Eﬀect Results: Impact of a Newly Constructed Dam on the Natural Log of the Value of the Land per Acre 1900-1920

(1) (2) (3) (4)

Dam 0.059 0.081 0.155∗ 0.193∗ (0.104) (0.107) (0.092) (0.105)

Time fixed effect Yes Yes Yes Yes County fixed effect Yes Yes Yes Yes Controls No Yes Yes Yes Soil*climate No No Yes Yes State*year fixed effect No No No Yes R2 0.25 0.26 0.31 0.40 N 2610 2610 2610 2610 *** p<0.01 ** p<0.05 * p<0.1 Standard errors are in parentheses, clustered at county level. Controls: Extreme cold and hot events, and precipitation

Table 1.8: Fixed Eﬀect and Instrumental Variable Results, Impact of a Newly Constructed Dam on the Natural Log of the Value of the Land per Acre 1900-1920

(FE) (IV)

Dam 0.193∗ 0.818 (0.105) (0.632)

Controls Yes Yes Time fixed effect Yes Yes County fixed effect Yes Yes State*year fixed effect Yes Yes R2 0.40 0.39 N 2610 2609 Kleibergen-Paap F statistic: 17.319 *** p<0.01 ** p<0.05 * p<0.1 Standard errors are in parentheses, clustered at county level. Controls: Extreme cold and hot events, precipitation and soil-climate interactions 39

Table 1.9: Fixed Eﬀect and Instrumental Variable Results, Impact of a Newly Con- structed Dam on the Natural Log of the Value of the Land per Acre 1900-1920

(lag Bureau) (lag Bureau) (Non-Bureau ) (Non-Bureau ) (FE) (IV) (EF) (IV)

Dam 0.211∗∗ 0.554 0.189∗∗* 0.716 (0.102) (0.361) (0.092) (0.474) Lag Dam 0.122 0.248 (0.145) (0.211) Non-Bureau Dam -0.097 -0.099 (0.091) (0.094)

Controls Yes Yes Yes Yes Time fixed effect Yes Yes Yes Yes County fixed effect Yes Yes Yes Yes State*year fixed effect Yes Yes Yes Yes R2 0.40 0.39 0.40 0.39 N 2610 2609 2610 2609 Kleibergen-Paap F statistic 19.730 19.545 *** p<0.01 ** p<0.05 * p<0.1 Standard errors are in parentheses, clustered at county level. Controls: Extreme cold and hot events, precipitation and soil-climate interactions Lag Dam variable that equals one if Bureau constructed dam in county i in period t − 1 Non-Bureau Dam variable that equals one if other agencies constructed dam in county i in period t

Table 1.10: Area Irrigated, Capital Invested

Year 1910 1920 1930 1940

Total acres irrigated 14,433,285 19,191,716 19,547,544 21,003,739

Bureau share (%) 2.7 6.5 7.6 8.7

Total capital invested - 697,657,328 1,062,049,201 1,052,049,201

Bureau share (%) - 18.6 21.7 25.8 40

Table 1.11: Fixed Eﬀect Results: Impact of a Newly Constructed Dam on Bushel per Acre and Acres Planted of Major Crops 1900-1920

Alfalfa Cotton Sugar Beet Wheat bushel acre bushel acre bushel acre bushel acre

Dam 0.057∗∗ 0.013∗∗ -0.000 -0.001 -0.002 0.033 -0.229 -0.006 (0.024) (0.006) (0.000) (0.001) (0.003) (0.040) (0.168) (0.008)

Controls Yes Yes Yes Yes Yes Yes Yes Yes Time fixed effect Yes Yes Yes Yes Yes Yes Yes Yes County fixed effect Yes Yes Yes Yes Yes Yes Yes Yes State*year fixed effect Yes Yes Yes Yes Yes Yes Yes Yes R2 0.24 0.33 0.20 0.34 0.16 0.16 0.47 0.49 N 2610 2610 2610 2610 2610 2610 2610 2610 *** p<0.01 ** p<0.05 * p<0.1 Standard errors are in parentheses, clustered at county level. Dependent variables are defined as “bushel per acre” and “total acre divided by farm land”. Controls: Extreme cold and hot events, precipitation and soil-climate interactions

Table 1.12: Instrumental Variable Results, Impact of a Newly Constructed Dam on Bushel per Acre and Acres Planted of Major Crops 1900-1920

Alfalfa Cotton Sugar Beet Wheat bushel acre bushel acre bushel acre bushel acre

Dam 0.079∗∗ 0.000 0.000 -0.005∗ 0.051 0.111 -0.488 -0.003 (0.038) (0.021) (0.001) (0.003) (0.063) (0.087) (0.369) (0.021)

Controls Yes Yes Yes Yes Yes Yes Yes Yes Time fixed effect Yes Yes Yes Yes Yes Yes Yes Yes County fixed effect Yes Yes Yes Yes Yes Yes Yes Yes State*year fixed effect Yes Yes Yes Yes Yes Yes Yes Yes R2 0.23 0.32 0.20 0.34 0.12 0.12 0.47 0.48 N 2609 2609 2609 2609 2609 2609 2609 2609 *** p<0.01 ** p<0.05 * p<0.1 Kleibergen-Paap F statistic: 19.548 Standard errors are in parentheses, clustered at county level. Dependent variables are defined as “bushel per acre” and “total acres divided by farmland”. Controls: Extreme cold and hot events, precipitation and soil-climate interactions 41

Table 1.13: Fixed Eﬀect and Instrumental Variable Results, Impact of a Newly Constructed Dam on the Share of Acre Improved, Log of the Livestock per Acre, and Log of the Dairy Value per Acre 1900-1920

Acre improved livestock Dairy

Dam 0.021 0.079 0.134 (0.020) (0.092) (0.153)

Controls Yes Yes Yes Time fixed effect Yes Yes Yes County fixed effect Yes Yes Yes State*year fixed effect Yes Yes Yes N 2599 2599 2599 R2 0.37 0.56 0.27 *** p<0.01 ** p<0.05 * p<0.1 Standard errors are in parentheses, clustered at county level. Controls: Extreme cold and hot events, precipitation and soil-climate interactions

Table 1.14: Instrumental Variable Results, Impact of a Newly Constructed Dam on the Share of Acre Improved, Log of the Livestock per Acre, and Log of the Dairy Value per Acre 1900-1920

Acre improved livestock Dairy

Dam -0.003 0.060 0.202 (0.062) (0.339) (0.438)

Controls Yes Yes Yes Time fixed effect Yes Yes Yes County fixed effect Yes Yes Yes State*year fixed effect Yes Yes Yes N 2597 2597 2597 R2 0.37 0.56 0.27 *** p<0.01 ** p<0.05 * p<0.1 Standard errors are in parentheses, clustered at county level. Controls: Extreme cold and hot events, precipitation and soil-climate interactions 42

Figure 1.1: Number of Dams Constructed by Diﬀerent Types of Owner

Note: Only Western states are included.

Figure 1.2: Percentage of Dams Constructed by Diﬀerent Types of Owner

Note: Only Western states are included. 43

Figure 1.3: Mean of the Height of Dams Constructed by Diﬀerent Types of Owner in Each Decade

Note: Only Western states are included.

Figure 1.4: Mean of the Maximum Storage of Dams Constructed by Diﬀerent Types of Owner in Each Decade

Note: Only Western states are included. 44

Figure 1.5: Percent Change in Farm Value per Acre (1900-1910)

Figure 1.6: Percent Change in Farm Value per Acre (1910-1920) 45

CHAPTER 2

Politics and Dam Construction: Historical Evidence from the Western U.S.

2.1 Introduction

At the end of the nineteenth century private enterprise had reclaimed some of the arid west, however there was a demand for larger irrigation projects to provide water to the area. In 1902, Congress passed the Newlands or Reclamation Act to build irrigation projects. According to the ﬁrst annual report of the Department of the Interior ( Newell(1905)):

The act provides for the entry of the lands reclaimed in accordance with the provisions of the homestead law.

Although the locations for potential water projects were investigated by the Bureau of Reclamation (hereafter referred to as Bureau) commissioners, engineers, and other experts, political pressure had significant influence on location decision for constructing the dams. Since authorizing, initializing and completing of the major national projects had to be passed through Congress, political power and the aim of the congressmen and senators could be quite influential on the location and the number of the dams in each district. According to the history of the Bureau of Reclamation, Michael Robinson1 states:

1The son-in-law of a Commissioner of Reclamation 46

estate interests, and congressman were convinced their areas were ideal for reclamation development( Reisner(1993)).

During the New Deal the Flood Control Act of 1937 authorized another federal agency, the Army Corps of Engineers (hereafter referred to as Corps) to be in charge of the controlling flooding on the west. This created a competition between the two agencies in the regions with navigable rivers and led to couple of agreements to solve the conflict. Acquisition of the projects by each constructor was related to their political power in the region, which was related to the support of the President, congressmen on the related committees, and legislators because of these projects potentially brought a great deal benefit to the district. 2 These major water projects had important effects on the local economy and agriculture, therefore it is worth investigating the important factors affecting the location and the timing of the major water projects. A study by Hansen et al. (2011), identified the major factors in the construction of the water infrastructure using a state-level data set. Their results show that there is a strong correlation between the House committee session-seat representation and the number of dams constructed in a state. This effect is more pronounced in the western states compared to an all state sample. In this paper, we investigate the effect of geographic, economic and political factors on the locational Bureau dams in the early 1900s. Then, the dam construction purposes are compared before and after 1937 when the Corps also started to construct the dams in the West. Finally, a hypothetical estimation of the Corps proportion in dam construction will be conducted.

2.2 Historical Background

2.2.1 National Irrigation Congress Role

The ﬁrst National Irrigation Congress was organized in Salt Lake City, Utah in 1891 and made irrigation a substantial national issue. The key people in the

2Pork-barrel projects 47

Congress were Senator Francis E. Warren (Republican from Wyoming); William E. Smythe, the editor of the “Irrigation Age” ; Elwood Mead, an irrigation engineer from Wyoming. After that, all Congresses were held in the Western States until 1900. The ninth annual National Irrigation Congress organized in Chicago on Nov. 24, 1900 gave the reasonable hope that national irrigation was not a dead issue and would be solved soon by the support of the Federal government and the help of engineers and scientists. This was the first National Irrigation Congress held in the East and the Congress proposed that the irrigation problem was not only an arid States problem but also a national concern. During the meeting, Congressman Newlands of Nevada gave a speech about a great friendship for the irrigation issue and he appreciated the amazing ideas proposed by George H. Maxwell. George H. Maxwell was a lobbyist, publicist, and journalist and one of the most influential people in water development in the arid Western States. He joined the fifth Irrigation Congress in Kansas in 1899 and by the end of the Irrigation Congress he became a national leader. Maxwell was not satisfied with the activity of the Irri- gation Congress in promoting legislation. He and two others founded the National Irrigation Association (later the National Reclamation Association) in the eighth National Irrigation Congress in Kansas in 1899 and he became executive director of the Association. He was sent to Washington, D.C., in 1900 by the Reclamation Asso- ciation along with Nevada Representative Francis G. Newlands, Frederick H. Newell of the United States Geological Survey and director of the Reclamation Service, Sen- ator William M. Stewart of Nevada, and Gifford Pinchot, well-known supporter of forestry management, to promote the legislations. Maxwell, Newlands, and Newell prepared legislation that was proposed in the House by Newlands. Maxwell also offered suggestions on Theodore Roosevelt’s 1901 congressional speech on forestry and irrigation written by Newell and Pinchot (Maxwell(accessed Feb 3, 2014)). At the end of the ninth Congress, the committee reported the following resolution:

We hale with satisfaction the fact that both the great political parties in their platforms in the last campaign, declared in favor of the recla- 48

mation of “Arid America” that settlers might build homes on the public domain and to that end we urge open Congress that national opportunities commensurate with the magnitude on the problem should be made for the preservation of the forest and the reforestation of the de- nuded area as natural storage reservoirs, and for the construction by the national government as a part of its policy of internal improvement of storage reservoirs and other works for flood protection and to save us in aid of navigation and irrigation, the water which now runs to waste, and for the development of artesian and subterranean sources of water supplies. The water of all streams should for ever remain subject to the public control and the rights to the use of water for irrigation should inhere in the land irrigated and beneficial use be the basis the measure and the limit of the right. We commend the efficient work of the various bureaus of the national Government in the investigation of the physical and legal problems and other conditions relating to irrigation, and in promoting the adoption of more effective laws, customs and methods of irrigated agriculture and urge upon Congress the necessity of providing liberal appropriations for this important work (Los Angeles Times (1900)).

Furthermore, the industrial and commercial importance of irrigation was dis- cussed in the evening session by Tom L. Cannon, secretary of the St. Louis Manu- facturers Association, Elliot Durand and B G. Chandler. MR. Cannon stated that:

I believe in internal improvement as I do in the forms of Republican government. I believe in the Federal government improving its own property for the beneﬁt of the people composing the government....If it is right for the Federal government to build harbors along the sea coast and great waterway channels in diﬀerent sections, it is right for the Federal government to improve that great American Desert and reclaim arid America through irrigation. I believe in making this country not only the greatest 49

agricultural country in the world, and the greatest manufacturing country in the world, but I believe in making it the seat of a ﬁnancial empire and becoming a creditor of all nations instead of a debtor... These things are closely allied with the work of this convention today. If we build storage reservoirs in the mountains of the West and control the water supply for irrigation purpose...(Los Angeles Times(1900)).

2.2.2 Passage of Irrigation Bill

The irrigation Bill was first introduced to the Congress by the Nevada Representa- tive, MR. Newlands. After the campaign of 1900, however, even the supporters of irrigations considered it as a dream. Newland believed that “The U.S. should cease its irrigation of foreign lands and begin the irrigation of its arid lands”, and had an influential role in the passage of the Act. The bill faced the opposition by the House leaders in the fifty-sixth Congress. During this time Theodore Roosevelt became the president, and one of his first priorities was to improve the irrigation in the West. On Dec. 3, 1901, President Roosevelt made some important recommendations to the Senate and House of Rep- resentatives regarding different issues. One of his concerns was national irrigation. In his recommendations he stated that:

Great storage works are necessary to equalize the flow of streams and to save the flood waters. Their construction has been conclusively shown to be an undertaking too vast for private effort. Nor can it be best ac- complished by the individual State acting alone. Far-reaching interstate problems are involved and the recourses of single States would often be inadequate. It is properly a national function, at least in some of its features. It is as right for national government to make the streams and rivers of the arid region useful by engineering works for water storage as to make useful the rivers and harbors of the humid region by engineering works of another kind .... The government should construct and 50

maintain these reservoirs as it does other public works. Where their purpose is to regulate the flow of streams, the water should be turned freely into the channels in the dry season to take the same course under the same laws as the natural flow... The reclamation of the unsettled arid public lands presents a different problem. Here it is not enough to regulate the flow of streams. The object of the government is to dispose of the land to settlers who build homes upon it. To accomplish this object water must be brought within their reach.... The pioneer settlers on the arid public domain chose their hoes along streams from which they could themselves divert the water to reclaim their holdings. Such opportunities are practically gone. There remain, however, vast areas of public land, which can be made available for homestead settlement, but only by reservoirs an main-line canals impracticable for private enterprise. These irrigation works should be built by the government for actual settlers, and the cost of construction should so far as possible be repaid be the land reclaimed. The distribution of the water, the division of the streams among irrigators, should be left to the settlers themselves in conformity with State laws and without interference with those laws or with vested rights. The policy of the national government should be to aid to irrigation in the several States and Territories in such manner as will enable the people in the local communities to help themselves, and as well stimulate needed reforms in the State laws and regulations governing irrigation (The Washington Post(1901)).

The second time the Bill was passed by the majority of votes in Senate. In the House, however, there were some strong opposers, such as Representative Joseph Gurney Cannon (leader of the Republican Party, Illinois), Representative John Daizell (Republican from Pennsylvania), Representative William Peters Hepburn (Republican from Iowa), Representative Frank Smith Payne (Republican from Iowa) and some others. Mr. Cannon voted against the bill although he was in favor of the pending amendment. He stated that: 51

The law appropriating $25,000 a year for each of the agricultural colleges out of the proceeds of the sale of public lands was a perpetual charge that fund.

Representative James Robinson (Democrat, Indiana) opposed the bill as he believed that the majority proportion of the fund from the sale of public lands would be depleted for irrigation purposes. Therefore, the agricultural and mechanical colleges would have to rely on Public Treasury. Mr. Daizell did not agree with the Bill as he thought that the benefits only went to the arid States while the other States were paying the costs (New York Times(1901)). Supporters of the Bill were led by Western Republican Frank Wheeler Mondell (Republican, Wyoming), William Augustus Reeder (Republican, Texas), as well as Newlands, John F. Shafroth (Silver Republican (1897-1903), Democratic (1903- 1922), Colorado). Representative Newlands had secured the vote by the Democratic Congressional Campaign Committee and therefore, passing the bill needed a sufficient number of Western Republicans. At the end of the session, Mr. Cowherd (Democrat, Missouri) gave a speech about the benefits of irrigation to the Western states from the passage of the bill (Los Angeles Times(1902)). Eventually, in June 13, 1902 the House passed the irrigation bill by a vote of 16 to 55. The bill provided that the receipts from the sale of the public lands in Arizona, California, Colorado, Idaho, Kansas, Montana, Nebraska, Nevada, New Mexico, North Dakota, Oklahoma, Oregon, South Dakota, Utah, Washington, and Wyoming for all time would be devoted to works of irrigation. The arid land reclamation fund was to be placed in the Treasury for the construction of irrigation works and providing the water available to the settlers. The Secretary had the authority for letting of contracts for the irrigation works whenever the money necessary for a project was available in the Reclamation Fund.

Provision is made for the payment out of the Treasury of any deﬁciencies in the allowances to agricultural colleges owing to this disposition of public lands. The secretory of the Interior is authorized to examine, 52

survey, and construct the irrigation works, and report the cost thereof to Congress at each session.(New York Times(1901)).

The law authorizes the Secretary of the Interior to withdraw from entry the lands capable of being reclaimed and provides that each project shall be self-compensatory, compelling the settlers to pay back into the fund in ten equal annual installments their proportionate part of the cost of each construction. Also to prevent land monopoly, section 5 of the bill provides that:

No right to the use of water for land in private ownership shall be sold for a tract exceeding 160 acres to any one land owner, and no such right shall permanently attach until all payments therefore are made to any land owner unless he be an actual bona ﬁde resident on such land, or occupant thereof. residing in the neighborhood of such land (New York Times(1901)).

2.2.3 Bureau of Reclamation

In 1902, the Reclamation Act created the Bureau to help provide irrigation in the Western States (Miller and Miller(1992)). The bill was designated to convert arid federal lands into suitable places for agriculture. The irrigation projects included construction of dams, power plants, canals, and other water facilities. They were to be ﬁnanced through a Reclamation Fund which was provided from selling the Federal land and later on by selling the water to the irrigators. To determine the feasibility of the water projects, geological surveys were prepared by the Bureau that considered all related factors for dam construction such as the amount of the water ﬂow in the river, elevation of the surface, the streams, and their catchment areas (Newell(1905)). The Bureau’s primary purpose was to help improve the agriculture in the western U.S. However, because of the political pressure by the congressmen, and state governments to acquire the water projects, dams might have been constructed in the districts without having enough potential for agriculture. The water projects 53

mostly had been authorized by Congress; however, the Presidents were able to veto the bill. According to the history of the Bureau of Reclamation, Michael Robinson3 states:

The financing of the Reclamation projects obligated the farmers to meet their repayment obligation in ten years. This proved to be an unrealistic estimate as sixty percent of the farmers delayed their payments. In some cases the delays stretched beyond twenty years from the passage of the first Reclamation law and the repayment period was extended to forty or fifty years.

2.3 Data

To understand the critical factors aﬀecting the decisions of the location of dams, a new dataset is assembled from several diﬀerent datasets. The following section includes the summary statistics and the sources of the combined datasets.

2.3.1 Major Dams

The dam dataset 4 comes from the National Atlas of the United State. The data include information on the name, national ID, latitude, longitude, owner name, type of owner, year of completion, all purposes, primary purpose, capacity, height, and

3The son-in-law of a Commissioner of Reclamation 4Source: National Atlas of the United States, 200603, Major Dams of the United States: Na- tional Atlas of the United States, Reston, VA. 54 some other characteristics for the major dams in the U.S. The dataset includes 8,121 dams considered to be “major”. A major dam is 50 feet or more in height or has a normal storage capacity of 5,000 acre-feet or more, or with a maximum storage capacity of 25,000 acre-feet or more. The summary statistics of the Federal dams are presented in Table 2.1. There are 1,279 major Federal dams throughout the U.S. of which 371 were constructed by the Bureau (Figure 2.4) and 615 by the Corps (Figure 2.5). The shares built by other agencies- the Tennessee Valley Authority (TVA), The Bureau of Indian Affairs, U.S. Forest Service, and U.S. Fish and Wildlife Service- are not large. Furthermore, Figure 2.1 illustrates the histogram of the dams’ completion years for the Bureau and the Corps. The primary purposes of dam construction include flood control, debris control, fish and wildlife protection, hydroelectric generation, irrigation, navigation, fire protection, recreation, water supply enhancement, and tailings control. Table 2.2 shows the frequency of the primary purpose of the dams constructed by the Bureau and the Corps in the West. Clearly, most of the dams constructed by the Bureau were based on the primary purpose of irrigation and water supply. Furthermore, flood control and navigation were main primary purposes of dam construction by the Corps in the American West. Table 2.3 illustrates the primary purposes of the dams constructed by the Bureau and the Corps before and after the competition. Started in 1936, in the Pre and Post periods, the Bureau has the majority of the dams with a primary purpose of irrigation and water supply and the Corps has the majority of the dams with primary purpose of flood control and navigation. However, in some cases both agencies constructed dams with primary purposes that were not related to their original mission. Furthermore, the number of dams built just for one purpose of flood control or irrigation decreases over time; however, the number of the dams constructed with multi purposes and primary purposes of flood control or irrigation increases during 1950 and 1960. This is illustrated in Figures 2.2 and 2.3. 55

2.3.2 Presidential Election

The political data come from the ICPSR’s United States Historical Election Returns (ICPSR, 1824-1968). The data include the percentage of votes for Republicans in presidential elections at the state level. Since we are interested in the water infras- tructures constructed during 1902 to 1910, the percentage of votes for Republicans of the 1900, 1904, 1908 elections are chosen. The summary statistics of the elections for the western states are shown in Table 2.4. The first three columns are the percentages of the republicans vote, and the average for the three elections is shown in the fourth column. The Dam column is an indicator whether a dam has been constructed in a state or not during the mentioned time period. South Dakota, Washington, Wyoming, Oregon, and California had remained Republican during the first decade of the twentieth century and dam projects were constructed by the Bureau there. Idaho and Montana, which were Democrat, had no dams constructed until they had voted more than 50 percent Republican in 1904 (Table 2.5). The comparisons are suggestive that the political power of the Republican party contributed to the construction of major dams in Republican states.5 Therefore, it is reasonable to consider the percentages of votes for Republican Presidential candidate as a political variable that had a significant influence on the location of the dams.

2.3.3 Geographic Characteristics

The climate data are obtained from the U.S Historical Climatology Network for each weather station and Geographic Information Systems software is used to interpolate and obtain the county level climate data. Speciﬁcally, the data include the average of extreme events (hot days that exceed 90 degrees Fahrenheit and cold days, below 32 degrees Fahrenheit for each year) and total rainfall is calculated for the 1900-1910 time period. To have a proxy for the size of the rivers, an indicator is constructed from the

5Arizona, New Mexico and Oklahoma were territories which did not have voting representation in the U.S. Congress, and therefore did not have electoral votes for presidential election. 56

U.S. Geological Survey Geographic Names Information System. The indicators are as follows: A small river is deﬁned as passing through 5 to 10 counties. A medium river passes through 11 to 50 counties, and ﬁnally a large river passes through more than 50 counties. These are three variables that show the number of rivers of each size that pass through or along the county.

2.4 Empirical Strategies and Results

2.4.1 Bureau in 1910

Using the county level dataset, I estimate the effect of the geographic, economic, and political factors on the location of the dams constructed by the Bureau of Recla- mation between 1902 and 1910. The sample includes the counties on the seventeen Western States and the dependent variables is whether the dam has been located in the county during 1902-1910. We restrict our analysis to this time period as we aim to find out whether the Bureau paid appropriate attention to the agricultural economics, soil science, climate condition, and other relevant factors in placing the water projects in a county. Historical narratives suggest that the dams were constructed with little consideration for the need of areas irrigation for agriculture as the political pressures led the proposed projects to have superficial conception of the regions agriculture, soil quality, and market access (Reisner(1993)). We estimate the following model by two logistic regressions, in one of which we

control for the political variable. The DAMist variable is an indicator of whether the dam is constructed in state s, county i during the time period 1900-1910.

Damist = β0 + β1Economicist + β2Geographicist + β3P oliticalst + ist (2.1)

The economic variables include the population density of the year 1900. The geographic variables are the yearly average of the following over 1900 to 1910: the days that the temperature exceeds 90 degrees Fahrenheit, the days that the temperature decreases below 32 degrees Fahrenheit, and total precipitation. Other 57

geographic variables include the elevation of the surface and presence of rivers of diﬀerent sizes in the county. The political factor is the average of the percentage of votes for Republicans in the presidential elections of 1900, 1904, and 1908. Finally,

the unobserved error component is it.

2.4.2 Results

The results of the Logit estimations of Equation 1 are presented in Table 2.6. The first and second specifications are similar except model 2 incorporates the Republi- can voting measure. The coefficient estimates in the two models have the expected signs. The population density has a negative correlation with the dams construction in the county. In the beginning of the Bureau operation most of the dam were constructed in unpop- ulated areas where private dams are unlikely to locate. Furthermore, the counties with large rivers were more likely to have dams in contrast with the counties with small rivers. Extreme weather also has negative effect on the water projects’ construction.Total precipitation has a positive correlation with dam construction and raises the probability that a county has a dam. Finally, the results show that the political variable is statistically significant at the 5 percent level and has a positive effect. This indicates that on average, a one unit increase in the percentage of votes for Republicans increases the probability of having a dam by 0.3 percent. Therefore the results support the hypothesis that support for the Republican administration had a strong impact on the location of major water projects at the beginning of the operation of the Bureau.

Concerns

Since the major national projects in the United States had to be passed through a process involving the Congress, the Senate, and the House of Representatives, thus being inﬂuenced by the political power of the mentioned members, it is necessary to consider this political power in our model as the number of seats in the Senate and 58 the House or the seniority of the Congressman and Senators in relevant committees. Although data are available for the mentioned factors individually, processing and matching of the appropriate variables was not in the context of this paper within the current time frame and will be added to the model in future editions. Applying the average of the percentage of votes for Republican in presidential elections is not a precise variable to evaluate the political power of the lower house and upper house of the Congress; however, we suggest that it can be used as a representative of the relative power balance in the era mainly due to the fact that the majority of the Congress were Republican for consecutive years in the beginning of the 20th century and they were not opposed to the president. Considering this fact, the average of the percentage of the presidential votes in each state can be a representative of the power of the congress and subcommittees related to the same state.

2.4.3 Army Corps of Engineers

Organized in 1815, the Corps was exclusively in charge of the maintenance of the na- tion’s rivers and flood control. However, after the colossal failure of the “levees only” policy and construction of a flood wall to control flooding on the river in the 1927 Mississippi flood, Congress authorized the Corps and the Federal Power Commission to prepare a survey called “308 Reports” about possible improvements to the navigable streams for water power, flood control, and irrigation, in the River and Harbor Act of 1927 (United States House of Representatives(1926)). After the disastrous northeastern floods of March 1936, Congress passed the Flood Control Act of 1937 which made flood control an official activity of the Federal Government. There- fore, the Corps officially became a nationwide planning and construction agency for flood control projects. This was a turning point in the history of the Corps as it dramatically increased the agency’s scope by empowering it to construct flood control reservoirs upon approval of Congress (Reuss and Walker(1983)). The Act authorized the development of the dams for flood-control purposes alone; however, anticipation of the potential possibility of hydro-power was important and included 59

in the bill (Billington et al.(2005)). The act set up a rivalry between the Bureau and the Corps in the areas that had navigable river basins since the Corps had the authority to improve the river for flood control and navigation. The rivalry between the Corps and the Bureau took place in three river basins:The Colombia, Missouri, and California Central Valley. There was also some competition on non-navigable rivers such as the Colorado River, the Slat, and the Gila Rivers. In general, a dam constructed by the Corps was more favorable in the eyes of the farmers since the water was free and the Federal government would have subsidized the dams with the purpose of flood control. (Reisner(1993)) Several agreements were negotiated to prevent the conflict between the Corps and Bureau in the mentioned areas. The first agreement was the Pick-Sloan Missouri Basin Program 6 in 1944. 7 Five years after the establishment of the Bureau in 1902, nine projects were constructed on the river in response to political pressure from the Missouri Basin states. 8 However, the repayments were delayed to as late as 1944. As a solution to overcome the financial difficulties, it was suggested to build high dams along upper tributaries to create hydropower revenue. On the other hand, in March and May of 1943 two major floods occurred between Sioux City and Kansas City and between Jefferson City and the mouth of the Missouri. As a result, the Corps proposed construction of five monstrous dams in the mainstream of the Basin in order to control the flooding. When Congress failed to come up with a solution to reconcile the two proposals, President Roosevelt proposed that a single regional authority such as the Tennessee Valley Authority solve the conflict between the agencies (Billington et al.(2005)). The threat of losing both projects led to the Sloan Plan and the Pick Plan, which established that the Corps would construct the dam on the mainstream of the river and focus on the lower tributaries while the Bureau would be in charge of upper tributaries. Any irrigation development by the Corps was to be done under Reclamation law.

6Flood Control Act 1944 7The Missouri River is the longest river in the American West and the longest tributary in the United States. 8Cadillac Desert, pg 163 60

The second reconciliation was the Newell-Weaver Plan of 1949. The Corps and Bureau agreed to take the development of the Colombia River Basin out of the Colombia Valley Authority 9 in a way similar to the Pick-Sloan Plan that tried to block out the Missouri Valley Authority. Under this agreement the Corps controlled water development of the main stream of the Colombia river and the lower part of the Snake River. The Bureau took control of the upper Snake River basin along the Oregon-Idaho state lines and everything from there to the Continental Divide. The same competition occurred in California. On November 1935 the secretary of the Interior submitted a report to the President for the approval of the development of the Central Valley as a Federal reclamation project. The report was approved by the President on December 2, 1935. The Bureau had just started the investigation phase of their projects on the King and Kern River when the Corps asked the House Flood Control and Appropriation Committees for permission to do similar investigations on the same rivers (Reisner(1993)). Consequently, by 1940 Congress received two plans from the two agencies. Four years later, the 1944 Flood Control Act allowed the Corps to construct ﬂood control purpose dams on the American, Kern, and Kings Rivers. They started constructing the Folsom Dam on the American River near the Sacramento, California in 1948. In 1949 the Folsom Formula settled the competition between the two competitors as follows:

Multiple-purpose dams are the responsibility of the Bureau of Recla- mation, and dams and the works exclusively for ﬂood control are the responsibility of the Corps of Engineers (Graham(1950)).

Under the Folsom Formula the Folsom Dam was transfered to the Bureau for coordinated operation as an integral part of the Central Valley Project. Dam operations for the Pine Flat Dam constructed by the Corps on the Kings River in the Southern San Joaquin Valley for the primary purpose of ﬂood control were taken over by Reclamation.

9The Pittsburgh Press, Jun 9, 1949 61

2.4.4 Corps versus Bureau

The Flood Control Act of 1936 gave the authority to Corps to be in charge of flood control throughout the U.S. Involving the Corps in the construction of the dams in the West resulted in a fierce competition between the Bureau and the Corps over dam locations and the appropriate construction sites in the areas where both agencies could have water projects. The Corps was constructing the dams more for the purpose of flood control at the beginning but by this time, the number of single purpose dams had declined and the number of multi-purpose dams had increased. This is illustrated in Figure 2.5. The multi purpose dams with the primary purpose of flood control dramatically increased during the 1950s and 1960s. Dams with recreation and hydropower purposes were potential sources of generating revenue. We are interested in finding out hypothetically what would be the Corps dam construction proportion if the Corps were just allowed to construct dams for flood control purposes. In order to find that, we estimate the probability that the Corps constructed a dam, in a sample including only the counties where Bureau or Corps constructed dams after 1937, while controlling for different construction purposes.

Corpsist = β0 + β1F lood P urposeist + β2Irrigation P urposeist + β3Economicist

+β4Geographicist + αst + ist (2.2)

The Corpsist variable is an indicator whether the dam was constructed by the Corps in state s and county i during time period t. The purposes of the construction of the dam are categorized as: 1) Flood control and navigation, 2) Irrigation and water supply, 3) Others . The Flood Purpose is an indicator if the county has a dam with flood control and navigation purposes at time t. The Irrigation Purpose is an indicator if the county has a dam with irrigation and water supply at time t. Therefore, β1 and β2 are the coefficients of interest, which respectively measures how much flood control and irrigation purposes an influence on choosing dam construction agents between Corps and Bureau. αst is year state fixed effects to capture the shocks that 62 happened within the states in each year.

2.5 Conclusion

After establishing the Bureau of Reclamation in the early twentieth century, many major federal dams were constructed to improve the arid land and agriculture in the western U.S. Geological surveys were prepared to locate areas with the potential for planting valuable corps, appropriate soil quality, and accessibility to markets. However, several histories claim that these factors were frequently disregarded as a results of the political pressure by the powerful constituencies. Assembling several different datasets for the beginning of the Bureau of Reclama- tion’s operation, this study investigates the extent to which geographic, economic, and political factors had influence on the locations of dam construction. The estimation results show that the percentage of votes for Republicans in presidential elections had a positive and statistically significant effect on the dam locations. On average, a one unit increase in the percentage of votes for Republicans increases the probability of having a dam by 0.3 percent. Furthermore, entering the Army Corps of Engineers as another federal agency in constructing dams in the western U.S. led to a rivalry between the Corps and the Bureau and changed the primary purpose of constructing dams by the two agencies. It is worthwhile to expand the datasets and perform the analysis beyond the first ten years of the Bureau’s operation and include the 1920s and 1930s . Furthermore, it might be interesting to investigate the effect of competition in the competitive areas that reconciliation between two agencies occurred. However, this requires identifying the competitive dam sites more precisely. 63

Table 2.1: Federal Dams in the West

Total Bureau Corps (West) Corps TVA Others Federal dams 1279 371 190 615 60 233

Note: The Corps and the Bureau were the major federal dam builders. The shares by another agencies, The Tennessee Valley Authority (TVA), The Bureau of Indian Aﬀairs, U.S. Forest Service, and U.S. Fish and Wildlife Service are not considerable. Major dams: height greater that 50 feet or a normal capacity of at least 5,000 acre-feet or a maximum storage capacity of 25,000 or more: “http://nationalatlas.gov/mld/dams00x.html” 64

Table 2.2: Primary Purposes

Primary Purpose Total Bureau Corps (West) Irrigation or Water Supply 353 309 19 Flood control or Navigation 537 29 141 Hydroelectric 51 24 13 Recreation 18 1 17 Others 22 4 6

Note: The table shows primary purposes of dams constructed by the Bureau and the Corps. Others category includes debris control, ﬁsh and wildlife pond, tailings and ﬁre protection, stock, or small farm pond purposes. 65

Table 2.3: Primary Purposes - Pre and Post 1936

Primary purposes Bureau-Pre Bureau-Post Corps-Pre Corps-Post Irrigation or water supply 89 216 1 18 Flood Control or navigation 2 25 1 140 Hydroelectric or recreation 2 23 - 24

Note: The table includes number of dams constructed with diﬀerent primary purposes by the Bureau and the Corps before and after 1936, the year that Congress passed the Flood Control Act. 66

Table 2.4: Association of the % of Votes for Republicans in Presidential Election and Dam Construction

Note: Arizona, New Mexico, and Oklahoma were territories which did not have voting representation in the U.S. Congress, and therefore did not have electoral votes for presidential election. 67

Table 2.5: Montana and Idaho Dams

Dam Name State Starting Completion Como MT 1908 1910 Willow Creek Bor MT MT 1907 1911 Boise River Diversion ID 1906 1908 Deer Flat Lower ID 1906 1908 Deer Flat Upper ID 1905 1908 Reservoir A ID 1907 1907 Minidoka ID 1904 1906

Note: Dams were constructed by the Bureau during 1902 to 1910 in Montana and Idaho. These states were Democratic and became Republican in 1904. 68

Table 2.6: Logit Estimation

Model 1 Model 2 Variables Coef. Std. Error Coef. Std. Error Mar. eﬀ. (%) Population Density -0.124*** 0.061 -0.107** 0.061 -0.26 Small River -1.211** 0.647 -1.250** 0.651 -3.01 Large River 0.410 1.177 0.648 1.258 1.56 Hot days -0.031*** 0.011 -0.013 0.013 -0.03 Cold days -0.026** 0.015 -0.039*** 0.017 -0.09 Rainfall 0.017 0.041 0.004 0.041 0.01 % of Votes for Republicans - - 0.125 *** 0.059 0.30 Constant -1.214 1.142 -8.732*** 3.725 -21.03

Log Likelihood -75.353 -71.268 Pseudo R2 0.128 0.175 Total # of obs. 813 813 LR test 8.171*** p-value 0.004

Note: *** signiﬁcant at 5% ; ** signiﬁcant at 10% 69

Table 2.7: Purposes - Pre and Post 1936

Purposes Bureau-Pre % Bureau-Post % Corps-Post % Irrigation 98 97 226 84 41 22 Water supply 16 16 88 33 58 31 Flood control 18 18 115 43 155 82 Navigation - - 4 1 20 11 Hydroelectric 24 24 83 31 50 27 Recreation 29 29 114 42 123 65 Total 101 270 188

Note: The table includes the number of dams constructed by the Bureau and the Corps and the percentage of dams constructed with diﬀerent purposes before and after 1936. 70

Figure 2.1: Dams - Bureau and Corps

Note: The horizontal axis is the year of completion of dams. The right hand side graph shows the number of dams constructed in each year by the Corps in the West. Except for two dams that were completed in 1916 and 1923, the completion year starts in 1937. 71

Figure 2.2: Dams Constructed by the Bureau: (a) One Purpose: Irrigation, (b) Multi Purposes: Irrigation - Hydroelectric, (c) Multi Purposes: Irrigation - Recre- ation

(a) (b)

Figure 2.3: Dams Constructed by the Corps: (a) One Purpose: Flood Control, (b) Multi Purposes: Flood Control - Hydroelectric, (c) Multi Purposes: Flood Control - Recreation

(a) (b)

Figure 2.4: Dams Constructed by the Bureau and Corps

Note: Total number of Federal dams constructed by the Bureau and the Corps. The black dots are dams constructed by the Corps and the Red dots are are the dams constructed by the Bureau. Major rivers in the USA are illustrated with the blue color. 74

Figure 2.5: Dams Constructed by the Corps

Note: Total number of Federal dams constructed by the Corps. Major rivers in the U.S. are illustrated with blue color. 75

Figure 2.6: Dams Constructed by the Bureau

Note: Total number of Federal dams constructed by the Bureau. Major rivers in the U.S. are illustrated with blue color. 76

CHAPTER 3

The Impact of Climate Change on Agriculture: Accounting for Climate Zones in the Ricardian Approach

Soudeh Mirghasemi 1

Sandy Dall’erba2

Francina Dominguez3

3.1 Introduction

Although small in terms of employment, agriculture is one of the most important sectors of the U.S. economy. The market value of agricultural products sold was about $395 billion in 2012.4 Roughly 52 percent of this value comes from just nine states: California, Iowa, Texas, Nebraska, Minnesota, Kansas, Illinois, North Car- olina, and Wisconsin. The United States controls almost half of the world’s grain exports and is the world’s largest producer and exporter of agricultural goods. In 2011, the U.S. produced 18.5% of the total grain production of the world.5 Further- more, agriculture is linked to other sectors of the economy such as transportation, manufacturing, marketing, and utilities through its supply and purchase linkages. Consequently, understanding the eﬀect of climate change on agriculture plays a major role on the price of the agricultural products in the U.S. and worldwide.

1University of Arizona, Department of Economics, University of Arizona, AZ 85721, USA. E-mail: [email protected] 2Department of Agricultural and Consumer Economics and Regional Economics Appli- cations Laborator, University of Illinois at Urbana-Champaign, IL 61801, USA. E-mail: [email protected] 3Department of Atmospheric Sciences, University of Illinois at Urbana-Champaign, IL 61801, USA. E-mail: [email protected] 4This was $97 billion, or 33 percent, more than in 2007, at the time of the last agriculture census. Source: http://www.agcensus.usda.gov 540.5% of corn, 35% of soybean, 12% of cotton, and 8% of wheat 77

Since agriculture uses over 42% of the land of this country, one should not expect the effect of climate on agriculture to be the same throughout the U.S. One example is the drought of the summer of 2012 which was the most severe drought after the 1950s. Indeed, its effect was more severe in some regions such as the Corn Belt compared to the other parts of the country. The U.S. export prices for corn increased roughly about 130 percent above the historical average 6, as a consequence of drought-related crop damage. The drought led also to an increase in the price of the products derived from corn, such as ethanol (Adonizio et al.(2012)). The attention generated by the impact of the summer of 2012 drought on the Corn Belt illustrates how vital it is for the U.S. agricultural sector to understand climate change in different regions and how to mitigate and/or adapt to it. Beyond the case of extreme events, the slow process of climate change is also very likely to bring more dramatic transformations in some parts of the country than others. For instance, a recent report by Jardine et al.(2013) highlights how specifically the Southwestern U.S. region is likely to be challenged by future climate conditions. In addition to already being the hottest7 and driest region of the country, the Southwest climate zone is warming and experiencing more drought than in the past century. It is also encountering a reduction in streamflows from its four major drainage basins. The projected climate conditions compiled in this report indicate more frequent heat waves in summer, a decrease in precipitation, more frequent precipitation extremes in winter, a decline in river flows and soil moisture, and more severe extremes (droughts and/or floods) in parts of the Southwest. Measuring the impact of future climate conditions on agriculture has attracted the attention of many scholars over the last two decades. According to previous studies, some regions of the US will be winners and others losers, but it is still unclear whether climate change will bring a net gain or a net loss for U.S. agriculture as a whole. Deschenes and Greenstone(2007) found that U.S. agriculture will benefit from the climate change and the annual profits will increase by $1.3 billion. Mendelsohn et al.(1994)

6In the last 20 years: October 1992 to September 2012. Source: http://www.bls.gov 7Based on July maximum temperatures 78

estimated that global warming may have benefits of about $1 − $2 billion per year, even without CO2 fertilization. Massetti and Mendelsohn, 2011 showed that the U.S. will benefit from climate change by 14.8-15.2 billion. In contrast, some other studies such as Schlenker et al.(2006) and Schlenker et al.(2005) lsuggest that there will be an annual loss of $3.1−$7.2 billion and $5−$5.3 billion respectively. So there is significant uncertainty about the impact of future climate changes on agriculture. There are several reasons why previous studies generate contradictory results. Some use a cross-sectional approach while some use panel data estimation with different sets of geographic fixed effects; there are differences in the discount factor used for actualization; and different specifications of temperature and precipitation data are used. The way that heterogeneity is defined is also different in previous studies. For example, Deschenes and Greenstone(2007) divided counties into irrigated vs. non-irrigated counties. Schlenker et al.(2006) and Schlenker and Roberts(2009) simply focus on the counties east of the 100th meridian, the historical boundary of non-irrigated agriculture.8 Yet, the econometric literature has highlighted numerous times that an improper consideration of heterogeneity in coefficients across cross- sectional observations may lead to biased and inconsistent estimates and hence to improper conclusions on the marginal effects of the estimates (Anselin and Le Gallo (2006); Durlauf and Johnson(1995)). Furthermore, preceding studies may suffer from an omitted variable bias for their lack of consideration of extreme climate events, which have increased in frequency and intensity over the last decades and are expected to increase even further in the future (Trenberth(1999)). Schlenker and Roberts(2009) consider the extreme climate events in their analysis and state that:

Holding the average temperature constant, days with more variation will include more exposure to extreme outcomes, which can critically inﬂuence yields.

Researchers have taken diﬀerent approaches to studying the eﬀect of climate

8100th meridian is one hundred degrees of longitude west of Greenwich. To the east of the 100th Meridian, average annual precipitation is in excess of twenty inches. 79 change on agriculture and crop production. These approaches can be summarized in three categories. The first one is the crop growth simulation model used by Nord- haus(1991), Richard(1995), Neumann and Mendelsohn(1999), Reilly et al.(2003), and Nelson et al.(2009). Based on agronomic (biophysical) models, their strength lies in their capacity to simulate crop growth over the life cycle of a plant exposed to the full range of weather outcomes, including extreme events. They also simulate how changes in climate modify a crop’s input requirements such as fertilizers and irrigation. However, this approach has been criticized for the large number of calibrated parameters it relies on, its lack of validity on a global basis, and for not accounting for the farmers’ capacity to adapt their crop choice to climate (Antle and Capalbo(2010); Hertel et al.(2010)). It should be noted that the Intergovernmental Panel on Climate Change (IPCC) uses the crop model approach in conjunction with the Basic Linked System (BLS) of National Agricultural Policy models, a world-level equilibrium model system (Fisher(2001)), to estimate the impact of climate change on food production (Parry et al.(2004)). This approach is very complex and requires coordination among several groups with large computational capacity and is beyond the goals of the present study. The second type of model uses regression analysis to estimate the impact of climate and other exogenous inputs (such as soil quality) on one type of crop (Adams (1989), Lobell and Burke(2008), and Schlenker and Roberts(2009)). 9 This approach requires significantly less data and computational resources than the simulated crop growth models, but does not consider the ability of farmers to adapt their choice of inputs and crops to changing climate conditions. This framework is particularly relevant in developing countries where it is difficult to assume that farmers have the private capital or government support necessary to adapt their farming practices (Lobell and Field(2007); Parry et al.(2004)). However, when it comes to the U.S., empirical evidence clearly demonstrates that adaptation at the farm level is already taking place. The 1996 report of Schim- melpfennig et al. indicates that:

9Production function approach 80

Some of the alternatives considered are adoption of later maturing cul- tivars, change in crop mix, and a timing shift of ﬁeld operations to take advantage of longer growing seasons.

Adaptation does not limit itself to crop-producers. Schimmelpfennig et al.(1996) also reports that:

The growth of dairy in the South is a testament to the creativity of farmers in ﬁnding ways to cool animals in hot climates 10. Other adaptations include herd reduction in dry years, shifting to heat-resistant breeds, and replacing cattle with sheep.

Another reason that makes it difficult to justify a crop-production approach in the U.S. is that the regression framework used in this literature treats each crop individually. By definition it does not account for the fact that crops are mutually dependent through factors such as crop rotation11 (Padgitt et al.(2000) or access to inputs (land, water, fertilizers, and government subsidies) whether they share the latter or compete for them. Finally, we believe that the prospect of an uncertain climate drives the desire to make public and private investment decisions that mitigate the impact of climate and increase the chance of adaptation. For instance, Antle and Capalbo(2010) indicate that mitigation policies, such as cap and trade on CO2 emissions linked with agricultural offsets, are already under consideration in the U.S. Based on these observations, a third theoretical framework called the Ricar- dian approach was initiated by Mendelsohn et al.(1994). Their regression model approach relies on the assumption that landowners, well aware of their local production conditions, allocate their land to the most rewarding use. This framework has attracted much attention when analyzing the US agricultural sector (e.g. Schlenker et al.(2006), Schlenker et al.(2005), and Deschenes and Greenstone(2007)), partly because decisions on which crop to plant, how much of each input (fertilizers, herbi- cides, etc.) to use and what tillage/management technique to adopt are determined

10For example, shading, wetting, circulating air, and air conditioning. 11Practiced for 85 percent of the corn and 75 percent of the wheat of the U.S. over 1990-1997. 81

endogenously and will be reflected in the value of farmland and/or agricultural profits, the usual dependent variables in a Ricardian regression framework (Kelly et al. (2005)). The estimated impact of climate change on agriculture by applying the Ricardian approach is found to be smaller than the former estimates as a result of taking adaptation into account.12 The main hypothesis in the Ricardian approach is that as the climate conditions change, the farmer adapts to the new conditions and adjusts the input or output. In principle, the farmer, in response to a change in weather, might modify the amount of the irrigated water or fertilizer, modify combination of the crops, or switch to grow another crop in order to maximize his profit. Therefore, in a well-functioning market, land prices are representative of the discounted value of land rent. While we also adopt a Ricardian framework in this paper, our work departs from previous contributions on a number of important points. First, we pay particular attention to the spatial heterogeneity among the climate zones. Previous contributions about the U.S. have overlooked the fact that the role of climate conditions on agriculture is expected to vary across climate zones, which raises concerns about the accuracy of coefficient estimates measured on the entire sample of U.S. counties. Even within one climate zone such as the Southwest, the large variety of the Southwest’s landscapes (mountains, valleys, plateaus, canyons, and plains) and associated elevation leads to diverse climates.13 Waldinger(2014) examines the economic effects of long term climate change during the Little Ice Age in Europe. She investigates economic and climatological heterogeneities in the effect of temperature on the size of city. Her findings show that cities that were larger and had better access to trade were less affected by climate change. Further, temperature changes affected cities differently depending on the climate zone where the city was located. While relatively warm cities were negatively affected by increases in temperature, relatively cold cities benefited from temperature increases.

12The ﬁrst two approaches overestimate the damage from climate change. 13Pinal, AZ, is 713 meters above sea level while Hinsdale, CO, is 3311 meters high. 82

We aim to obtain the effect of climate change on the U.S. agricultural farm value in years 1997, 2002, 2007 for the nine different climate zones in Figure 3.1 throughout the U.S. Secondly, we include extreme events (heat and cold waves as well as heavy precipitation) because several global and regional climate models suggest that extreme events will occur more often in the future (Tebaldi et al.(2006) and Dominguez et al.(2012)). Unlike work using global climate models (GCMs) data or statistically downscaled GCMs, we use dynamically downscaled data that allow us to explicitly account for changes in the intensity and frequency of extreme events at the local scale (a spatial resolution of 35-50km while GCMs used in Schlenker et al.(2006) have a spatial resolution of about 200-300km). This level of detail is applied to future climate projections, which in our case come from seven regional climate model (RCM) simulations. Such a variety of projections allow us to account for model uncertainty for future climate projections and will allow us to provide an envelope of likely future farmland values. This approach improves upon the usual projections based on a single climate model. The next section extends the Ricardian framework to a setting allowing for cross- sectional heterogeneity. Section 3 presents the dataset and data sources which report climate data at a much finer scale than in any of the previous Ricardian papers. Sec- tion 4 displays and discusses the estimation results first and then carries on with a set of predictions on future agricultural land values considering cross-sectional heterogeneity. Finally, section 5 summarizes our results and provides some concluding remarks.

3.2 The Ricardian Setting

In the traditional Ricardian setting a single farmer puts his land to its most prof- itable use given a set of local conditions. The farmer responds to local changes in weather by modifying the amount of irrigated water, fertilizer, land planted for the crop or switching to another crop. In a well-functioning market land prices are representative of the discounted value of land rent. This assumption allows us 83

to measure the eﬀect of the climate change on farmland value. In the absence of data at the individual farm level, a Ricardian model is estimated on a sample of geographical units and its reduced form is as follows:

Yizt = Xiztβ + εit (3.1)

where the dependent variable Y, farmland value per acre, is a function of climate, land and human variables. One caveat of Equation (3.1) is that it does not consider the spatial heterogeneity across the climate zones. The parameter estimate equals the average marginal eﬀect of the zth climate variable on the dependent variable y as if there is no diﬀerent impact across the climate regions.

∂y izt = β (3.2) ∂xizt

However, we believe that the marginal effect of the climate variable on the land value is different in each climate zone. Therefore, we estimate the effect of climate on farm land values using county level data:

9 X Yizt = Xizktβzk + wst + δi + εit (3.3) k=1

The main outcome variable is the log of the value of the farm per acre in county i,

in climate zone z, and in census year t. Xizt includes measures of income per capita, income per capita squares, population density, and population density squares as a proxy for demand. It also includes climate variables, rainfall, and hot and cold

extreme events to control for climate conditions. βz1 to βz9 are the coefficients estimates that are assumed to be different in each of the nine climate zones. The coefficient of interest are the Hot and Cold Events, which are the new factors we incorporate. We also include county fixed effects to capture the time-invariant unobserved characteristics related to each county. Year-state fixed effects are included to control 84

for the shocks that happened within the states in each year. εit is the unobserved error component. The identiﬁcation of the eﬀect of extreme events comes from changes over time within the same county after controlling for shocks common to the counties in each state-year and for the factors listed above.

3.3 Data

We follow the traditional Ricardian (Hedonic) approach and perform our analysis at the county level.14 The new county level dataset is assembled from several sources for the 1992 through 2007 agricultural census years for the whole United States. 15 The dependent variable is the log of the estimated value of the land and build- ings per acre provided by the US department of Agriculture (USDA). Independent variables can be divided into three different categories: climate data, soil data and socio-economic characteristics. When it comes to climate data, we rely on a finer scale to avoid any errors (noise) created by coarse measurements commonly used in both past observations and future projections. Unlike work using global climate models (GCMs) data or statistically downscaled GCMs, the use of dynamically downscaled data will allow us to explicitly account for changes in the intensity and frequency of extreme events at the local scale. This technique is applied to both past observations and to projections. While current approaches use projected climate variables derived from global climate models (GCMs) with a spatial resolution of about 200-300km (as in Schlenker et al.(2006)), we use data provided by regional climate model (RCM) simulations, driven by GCMs at their boundaries, with a spatial resolution of 35-50km. Our projections are performed over the 2038-2080 period defined by NARCCAP. This ensemble of RCM-GCM simulations provides us with an envelope of future climate scenarios that is then used in the econometric model. This helps to compare the results in the historical period with those obtained with gridded

14Speciﬁcally, we focus on the 3,076 counties of the conterminous U.S. States. Alaska and remote islands such as Hawaii and Puerto Rico are thus excluded. 15The U.S. Census of Agriculture data are available for every 5 years in this time period. 85

observations from the North American Regional Reanalysis (NARR, Mesinger et al. (2006)), available from 1979 to the present. To control for the climate conditions, we collected temperature and precipitation data from the North American Regional Reanalysis (NARR). The NARR assimi- lated dataset covers the contiguous United States and provides data at a 32km spatial resolution and 3-hourly temporal resolution from 1979 to the present. Thus, we obtain accurate estimates of climate variables based only on agricultural plots within each county for all the U.S. counties. We prefer NARR data over the pop- ular PRISM data (Deschenes and Greenstone(2007), Schlenker et al.(2006) and Schlenker and Roberts(2009)) since the latter is a monthly dataset that will not provide information on extreme events and cannot capture how the precipitation or Hot/Cold Events are being experienced by the plants. In contrast, NARR is available on a daily time scale permitting us to include extreme precipitation as well as Hot and Cold Events.16 Speciﬁcally, the climate variables that will be put in the model are:

1. Monthly temperature of January, April, July, and October.

2. Monthly average precipitation of January, April, July, and October.

3. Extreme Hot Event: number of events per year.

4. Extreme Cold Event : number of events per year.

We calculate extreme high and cold temperatures (extreme events) based on two thresholds, below 8◦C as a Cold threshold, above 32◦C as a Hot threshold. The Hot and Cold Events are deﬁned as the number of times that temperature was below or above the threshold for more than three consecutive days in a year.

16Since the climate data are provided as points, Inverse Distance Weighting (IDW) interpolation method was used to make continuous climate variation on continental U.S. counties by using Geographical Information System (GIS) software. Zonal Sum analysis with respect to counties was also used to calibrate each county’s climate conditions. The IDW (Inverse Distance Weighted) tool uses a method of interpolation that estimates cell values by averaging the values of sample data points in the neighborhood of each processing cell. The closer a point is to the center of the cell being estimated, the more inﬂuence, or weight, it has in the averaging process. Special thanks to Dongwoo Kang for helping us with this. 86

In future work when we develop projections, while current approaches use projected climate variables derived from coarse resolution GCMs, we will use data at a much finer resolution by using dynamically downscaled simulations.17 A dynamically downscaled model is used to capture the future impact of the climate change on yield production. These models provide more realistic precipitation and temperature data with high resolution.18 To date, all of the Ricardian studies have used only the coarse global climate models (GCM) with a spatial resolution of about 200-300 km. While GCMs generally do not represent climate variables due to their coarse spatial resolution and physical parameterizations, using these downscaled simulations datasets will help to address the issue of uncertainty in the future climate. Figure 3.2 shows the difference in aggregation between the two sets of spatial models.19 The variables capturing human intervention include population density, which acts as a proxy for demand and for the potential effect of development upon farmland value as well as per capita income. They come from the Regional Economic Accounts developed by the Bureau of Economic Analysis.

17Downscaled simulations were generated by Francina Dominguez and Yolande Serra at the Uni- versity of Arizona, department of atmospheric science using the Weather Research and Forecasting (WRF) model driven by the GCM : Hadley Centre coupled model, version 3 , HadCM3. 18Dominguez et al.(2012): “GCMs generally do not realistically represent precipitation due to their coarse spatial resolution and physical parameterizations, especially in complex terrain. Higher spatial resolution, improved representation of orography and land-surface heterogeneity, and hence a better representation of precipitation, are most practically achieved with the use of regional climate models (RCMs), as GCMs are presently too computationally expensive for multidecade climate change projection-type integrations with equivalent resolution. RCMs generally better capture mean and extreme precipitation at the regional scale (Diﬀenbaugh et al.(2005); Leung and Qian(2009)). We refer to the process of using RCMs forced at their lateral boundaries by GCMs as Dynamical Downscaling.” 19Since soil properties of land have an important role in agricultural production, soil properties of counties were also considered as an explanatory variables in the model. The Natural Resource Conservation Service of U.S. Department of Agriculture (USDA) provides U.S. General Soil Map (STATGO2) containing various soil properties including ﬂood frequency, soil erodibility factor (K factor), representative slope, wind erodibility index, hydric rating, electrical conductivity, permeability(K sat), salinity, available water capacity, percent clay, and percent sand. These properties are provided as polygon data which do not match with counties boundaries. By using Intersect analysis of GIS software, each counties’ weighted average soil properties were recalculated. The Flood Frequency ratio, Slope Steepness, Moisture Capacity, and Clay Content are used as a control for soil characteristics. 87

3.4 Results

3.4.1 Climate Regions

To test if we can pool some of the climate zones into one climate zone, we perform Chow tests for each climate zone with its neighbors. Table 3.10 illustrates the F test results. Check marks indicate that coefficient estimates of the two zones in the pooled model do not statistically differ from each other. Notice that we did the comparison only for the climate zones that are adjacent to each other. Table 3.10 shows the null hypothesis is rejected for climate regions such as South, Southeast, Northeast, Center, and East North Central. However, looking at the west side, the West climate zone has similar effect on the land values as the Northwest and Southwest. Furthermore, the Northwest region has similar effect on land value as the West, Southwest, and West North Center. After doing the pairwise comparison, we merge two of the similar zones and again compare them with their neighbors. We combine the West and Northwest regions and compare them with Southwest region. The F test indicates that we can combine these three regions and consider them as one. Finally, we combine the West, Northwest, and Southwest and compare them with the Northwest Center region. The Chow test shows that we can merge all these four regions to one climate zone which we call “West 4” accordingly. Therefore, the number of climate zones is reduced to six instead of nine. We do the rest of the analysis based on six climate regions.

3.4.2 The Results for the Past (1997-2007)

Table 3.11 shows the within R2 estimates when estimating the log of the farmland values as a function of different types of fixed effects. This is done to examine whether there is still much variation to be explained after controlling for fixed effects. The first column only controls for the climate zone fixed effects and has a low within R2 of zero. Controlling for the year and year-state fixed effects in the second and third columns increases the within R2 to 0.38 and 0.56. 88

Table 3.12 illustrates the effect of various measures of climate on the log of the farmland value per acre. For the moment, we focus on the Hot and Cold Events, which are the new factors that we incorporated. In the first column, the extreme event parameters are identified from variation within counties across time, across counties at the same time, and across time and space. The Hot and Cold Event estimates both have an expected negative impact on the farmland values; however, the Hot Event coefficient is not statistically significant. The second specification’s identifying variation is within county across time. A one unit increase in Cold Events decreases the farmland value per acre by 0.86 percent. The Hot Event’s coefficient is positive and statistically significant. We examine the variation within the same state in the same year in the third column using county fixed effects. The extreme weather parameters both have the expected sign, but only the Cold Event coefficient is statistically significant. In the last column, the extreme events parameters are identified from within county comparisons across time controlling for shocks in each state-year. The parameters both have statistically negative impact on the land values. A one unit increase in Hot Events and Cold Events reduce the land values per acre by 0.22 and 0.16 percent. The Hot Event estimate is the most negative compared to the other specifications in the table. The last specification has the smallest effect of the Cold Event on the land value. We estimate separate marginal effects of the explanatory variables across climate zones to be able to distinguish each zone’s marginal effects on the farm value per acre. We interact the zone dummies with all the control variables and run Equation 3 again. Table 3.13 indicates the estimation result. The statistical significance and magnitude of the marginal effects of the control variables are different among climate zones. For example, an increase in population density in the West4 has larger effect on the land value compared to the rest of the U.S. Further, the average temperature and precipitation in July have statistically negative effects on the land values in the West4, South, and Southeast regions. Also, although in a pooled model the Hot and Cold Events have statistically negative effects on farmland value, here we see that the coefficients are statistically positive in the Center region. Finally, to see 89 if the structural differences between the marginal effects are statistically significant across climate zones, we run different regressions considering each climate zone as a base group in each regression. In the first six sets of regressions, we control only for state year fixed effects (Tables 15A to 20A). In the second six sets of regressions we control for both county and state year fixed effects (Tables 15B to 20B). The reference group is changing in each table. The stars show the level of the significance. The important point is that no matter which variation we apply, some of the coefficients are statistically different from the one in the reference group. This shows how important it is to consider the differences in marginal effects for each climate zone separately instead of treating them as identical across the country.

3.5 Conclusion

This study shows the importance of considering spatial heterogeneity in estimating the impact of climate change on U.S. agriculture based on the Ricardian approach. We investigates the effect of climate change on U.S. agriculture using county-level data for the 1997, 2002, and 2007 census years. Compared to previous contributions, we pay particular attention to the spatial heterogeneity among the climate zones and include the presence of extreme weather events. We test if climate has different marginal effects on the land values in the various climate zones. Furthermore, while current approaches use projected climate variables derived from coarse resolution generation global climate models (GCMs), we use data at a much finer resolution by using dynamically downscaled simulations which are the combination of the GSMs and Regional climate models (RCMs).

We estimate separate marginal effects of the explanatory variables across climate zones to be able to distinguish each zone’s marginal effects on the farm value per acre. The statistical significance and magnitude of the marginal effects of the control variables are different among climate zones. Although in a pooled model the Hot and Cold Events have statistically negative effects on farmland value, here 90 we see that the coefficients are statistically positive in the Center region.

Finally, to examine whether the marginal effects vary in a statistically significant fashion across climate zones, we run different regressions considering each climate zone as a base group in each regression. In the first six sets of regressions, we control only for state year fixed effects . In the second six sets of regressions we control for both county and state year fixed effects. The important point is that no matter which variation we apply, some of the coefficients are statistically different from the one in the reference group. This shows how important it is to consider the marginal effect for each climate zone separately instead of assuming the same marginal effect throughout the U.S. 91

Table 3.1: Summary Statistics - Northwest

Variable Mean Std. Dev. Population 59339.605 94845.217 Population density 44.281 72.266 January temperature 180.733 128.637 January precipitation 3.914 3.592 April temperature 189.059 128.809 April precipitation 2.177 1.439 July temperature 202.508 128.715 July precipitation 0.728 0.663 October temperature 191.268 129.159 October precipitation 2.294 2.506 Hot event frequency 0.98 1.369 Cold event frequency 10.83 2.979 Farmland (acre) 387708.723 378476.139 Farmland value 670855.782 566314.092 Farmland value per acre 3264.336 3076.488 Log of farmland value per acre 7.702 0.883 N 339 Note: values in 2007 constant dollar

Table 3.2: Summary Statistics - West

Variable Mean Std. Dev. N Population 206855.6 382482.391 185 Population density 69.946 86.256 185 January temperature 187.012 129.085 185 January precipitation 2.742 3.076 185 April temperature 194.501 128.682 185 April precipitation 0.901 0.798 185 July temperature 207.07 128.352 185 July precipitation 0.194 0.19 185 October temperature 197.227 128.678 185 October precipitation 0.664 1.007 185 Hot event frequency 3.188 2.508 185 Cold event frequency 9.898 2.151 185 Farmland (acre) 513319.148 545410.853 176 Farmland value 1865789.746 2494569.288 185 Farmland value per acre 4213.164 4869.8 183 Log of farmland value per acre 7.913 0.971 183 Note: values in 2007 constant dollar 92

Table 3.3: Summary Statistics - Southwest

Variable Mean Std. Dev. N Population 73311.188 242059.756 393 Population density 29.829 67.77 393 January temperature 179.706 128.852 393 January precipitation 0.807 0.691 393 April temperature 190.58 127.786 393 April precipitation 1.093 0.857 393 July temperature 204.744 128.693 393 July precipitation 1.478 0.746 393 October temperature 192.674 129.972 393 October precipitation 0.921 0.559 393 Hot event frequency 2.466 2.66 393 Cold event frequency 7.59 1.91 393 Farmland (acre) 815238.53 848288.351 372 Farmland value 562565.136 503668.13 389 Farmland value per acre 1503.412 3489.254 386 Log of farmland value per acre 6.819 0.916 386 Note: Values in 2007 constant dollar

Table 3.4: Summary Statistics - West North Center

Variable Mean Std. Dev. N Population 46835.14 98146.979 8509 Population density 65.413 77.941 8509 January temperature 183.665 127.988 8509 January precipitation 2.129 1.686 8509 April temperature 194.085 127.467 8509 April precipitation 2.638 1.413 8509 July temperature 207.298 128.34 8509 July precipitation 2.811 1.643 8509 October temperature 197.088 129.774 8509 October precipitation 2.648 1.527 8509 Hot event frequency 3.009 2.73 8509 Cold event frequency 10.402 3.157 8509 Farmland (acre) 318911.791 390442.044 8461 Farm value 507217.709 558760.413 8501 Farmland value per acre 2455.807 2104.989 8494 Log of farmland value per acre 7.553 0.751 8494 Note: Values in 2007 constant dollar 93

Table 3.5: Summary Statistics - South

Variable Mean Std. Dev. N Population 34915.899 55283.007 1907 Population density 44.102 58.897 1907 January temperature 188.706 127.703 1907 January precipitation 2.257 1.83 1907 April temperature 198.723 126.889 1907 April precipitation 3.123 1.655 1907 July temperature 209.853 128.582 1907 July precipitation 3.448 2.087 1907 October temperature 200.913 129.407 1907 October precipitation 3.366 1.868 1907 Hot event frequency 5.489 2.683 1907 Cold event frequency 10.784 2.884 1907 Farmland (acre) 380163.144 292894.274 1901 Farmland value 435447.563 273935.256 1906 Farmland value per acre 1505.015 842.513 1904 Log of farmland value per acre 7.148 0.618 1904 Note: Values in 2007 constant dollar

Table 3.6: Summary Statistics - Southeast

Variable Mean Std. Dev. N Population 52111.258 66746.392 1461 Population density 93.515 81.986 1461 January temperature 190.312 128.867 1461 January precipitation 3.09 1.095 1461 April temperature 198.045 127.152 1461 April precipitation 2.63 1.148 1461 July temperature 208.694 128.655 1461 July precipitation 3.952 1.228 1461 October temperature 201.385 128.617 1461 October precipitation 3.208 1.115 1461 Hot event frequency 3.294 2.473 1461 Cold event frequency 12.137 3.57 1461 Farmland (acre) 101119.077 78966.049 1455 Farmland value 316950.833 310144.239 1459 Farmland value per acre 3368.493 1867.119 1459 Log of farmland value per acre 8.012 0.453 1459 Note: Values in 2007 constant dollar 94

Table 3.7: Summary Statistics - Center

Variable Mean Std. Dev. N Population 39964.056 49727.896 1879 Population density 85.321 79.172 1879 January temperature 183.231 127.872 1879 January precipitation 2.494 0.947 1879 April temperature 193.887 127.29 1879 April precipitation 3.125 1.218 1879 July temperature 207.062 128.339 1879 July precipitation 2.635 0.98 1879 October temperature 197.175 129.896 1879 October precipitation 2.617 1.082 1879 Hot event frequency 1.731 1.575 1879 Cold event frequency 12.216 2.353 1879 Farmland (acre) 177588.156 109108.827 1877 Farmland value 495801.603 376332.839 1879 Farmland value per acre 2750.748 1067.162 1879 Log of farmland value per acre 7.846 0.389 1879 Note: Values in 2007 constant dollar

Table 3.8: Summary Statistics - East North Center

Variable Mean Std. Dev. N Population 38898.946 45601.236 961 Population density 59.049 67.987 961 January temperature 175.638 126.969 961 January precipitation 0.906 0.716 961 April temperature 189.324 128.08 961 April precipitation 2.412 1.023 961 July temperature 205.389 128.004 961 July precipitation 2.993 1.407 961 October temperature 192.865 130.973 961 October precipitation 2.531 0.851 961 Hot event frequency 1.468 1.82 961 Cold event frequency 7.414 1.353 961 Farmland (acre) 258554.965 155865.322 958 Farmland value 671052.257 429349.405 961 Farmland value per acre 2646.751 941.158 961 Log of farmland value per acre 7.810 0.397 961 Note: Values in 2007 constant dollar 95

Table 3.9: Summary Statistics - Northeast

Variable Mean Std. Dev. N Population 85632.151 66156.289 517 Population density 133.068 101.092 517 January temperature 180.25 128.02 517 January precipitation 2.149 0.638 517 April temperature 189.723 127.472 517 April precipitation 3.129 1.36 517 July temperature 203.446 128.124 517 July precipitation 2.397 0.887 517 October temperature 194.109 130.032 517 October precipitation 2.565 1.268 517 Hot event frequency 0.274 0.774 517 Cold event frequency 9.355 1.924 517 Farmland 110903.417 73976.955 516 Farmland value 371702.686 300972.764 516 Farmland value per acre 4038.252 3953.638 516 Log of farmland value per acre 8.1 0.588 516 Note: Values in 2007 constant dollar 96

Table 3.10: Chow Test

Climate Regions Nw W SW WNCen S SE ENCen C NE North West XXX West XX South West XX West North Center X South South East East North Center Center North East The Chow test is done for only the climate regions that are adjacent. Check marks: Cannot reject the null hypothesis, and so we can pool the two regions. 97

Table 3.11: Control Only for Fixed Eﬀects in the Model

Fixed eﬀects (1) (2) (3) (4) (5) (6) climate zone yes yes Year yes yes Year-state yes yes yes County yes yes N 8494 8494 8494 8494 8494 8494 R2 (within) 0.00 0.38 0.56 0.38 0.56 0.56 Dependent variable: log of the value of the land per acre Standard errors are in parentheses 98

Table 3.12: Eﬀect of Climate on Land Values

(1) (2) (3) (4) Population density .0108911∗∗∗ .009503∗∗∗ .007176∗∗∗ -.0018484 (.0003499) (.0012022) (.0003279) (.0015375) Population density square -.0000205∗∗∗ -.0000159∗∗∗ -.0000139∗∗∗ 2.15e-07 (1.03e-06) (2.21e-06) (9.19e-07) (2.66e-06) January temperature .0207079∗∗∗ .0246165∗∗∗ .0073757∗∗ -.0005569 (.0015374) (.0017524) (.0032901) (.0039477) January precipitation .0070436∗∗ -.0066635∗ .0170788∗∗∗ -.0028808 (.0031685) (.0034151) (.0036607) (.0038558) April temperature -.0285335∗∗∗ -.0120193∗∗∗ -.0118179∗ -.0190122∗∗∗ (.0039409) (.0038125) (.0060563) (.0062151) April precipitation -.0062709∗∗∗ -.006612∗∗∗ .0059995∗∗ -.0018096 (.0020935) (.0020823) (.002759) (.002645) July temperature -.0200398∗∗∗ -.0331554∗∗∗ -.0566639∗∗∗ -.0144192∗∗ (.0028481) (.0029059) (.005609) (.0063906) July precipitation -.0052232∗∗ -.008637∗∗∗ -.0055112∗∗ -.0035824 (.0021624) (.0021606) (.0024381) (.0023981) October temperature .0274608∗∗∗ .0203432∗∗∗ .0209097∗∗∗ .0099965∗ (.001675) (.0015695) (.0046691) (.0052864) October precipitation .0146347∗∗∗ .002956 .0215846∗∗∗ .005547∗ (.0022229) (.0022659) (.0028795) (.0029546) Hot Event -.0010818 .0025622∗∗∗ -.00105 -.0025514∗∗ (.0008374) (.0008274) (.0011571) (.0010826) Cold Event -.0061158∗∗∗ -.0086834∗∗∗ -.0023965∗∗∗ -.0016772∗∗∗ (.0011172) (.001797) (.0006508) (.0005798) Constant 7.613559∗∗∗ 7.94954∗∗∗ 19.41571∗∗∗ 12.10962∗∗∗ (.0509266) (.0833759) (1.871941) (1.972285) County FE Y Y Year-state FE Y Y N 8494 8494 8494 8494 R2 0.18 0.20 0.54 0.57 Dependent variable: log of the value of the land per acre Standard errors are in parentheses 99 0.55 8494 West4 South South East Center East North Center North East (0.0118)(0.0108) (0.0155) (0.00429) (0.0219) (0.00717) (0.0157)(0.0103) (0.00634) (0.00809) (0.0132) (0.00913) (0.0134) (0.0363) (0.0199) (0.0103) (0.0145) (0.0744) (0.00148) (0.000695) (0.000417)(0.00546) (0.000373)(0.00881) (0.0100) (0.00720) (0.000687) (0.0139) (0.00770) (0.00107) (0.00767)(0.00968) (0.00645)(0.00866) (0.0156) (0.0128) (0.0180) (0.00407) (0.0145)(0.00652) (0.0205) (0.00652) (0.0239) (0.0141)(0.00625) (0.00514) (0.00563) (0.00230) (0.00766) (0.0145) (0.00597) (0.00218) (0.00575) (0.00160) (0.0633) (0.0218) (0.0116) (0.00341) (0.0174) (0.0151) (3.75e-06) (2.32e-06) (1.30e-06) (1.05e-06) (2.27e-06) (2.73e-06) (0.000595) (0.00231) (0.00980) (0.00280) (0.00116) (0.00430) Table 3.13: Effect of Climate on Land Values - Including State-Year Fixed Effects 2 R Population densityPopulation density square -2.78e-05***January temperature 0.0140*** -1.83e-05*** -8.13e-06***January 0.00817*** precipitation -9.62e-06*** 0.00463*** -0.00424April temperature -1.14e-05*** 0.00494*** 0.0357***April 0.0619*** precipitation 0.0411*** 6.02e-07 0.00537***July temperature 0.0251* 0.0229* -0.0195**July precipitation -0.0212*** 0.0221** 0.00152 -0.0269*** -0.101***October temperature -0.0840*** 0.00165 -0.0233* -0.0217 -0.0962***October precipitation -0.0835*** -0.0312*** 0.00226 0.0171*Hot -0.0575*** Event -0.00451 -0.0254 0.00936 0.00577 0.0557***Cold -0.0285*** Event 0.0658*** -0.0145 -0.0256*** 0.0130**N -0.0758*** 0.0384*** -0.00874 0.00686The -0.00330 0.0100 table -0.0193 shows estimationDependent result variable: 0.0910** for 0.000178 log one of regression. Standard the errors -0.00452 value are of 0.0171*** in -0.00150** -0.00427 theEstimation parentheses, land includes clustered 0.0487*** per state on -0.00444* acre year county*** fixed -0.0708 level. p effects. -0.00498** <0.01, ** p<0.05, * p<0.1 0.00669 0.000902 0.0113 0.00296 0.0875 0.00417*** -0.00470* 0.0472*** -0.0121*** 0.00128 -0.00165 0.000931 100 0.59 8494 West4 South South East Center East North Center North East (0.0112) (0.0202)(0.0104) (0.0316) (0.0111)(0.0104) (0.0196) (0.0130) (0.0203) (0.0123) (0.0265)(0.0117) (0.0125) (0.0221) (0.0146) (0.0101) (0.0194) (0.0363) (0.0201) (0.0170) (0.0205) (0.0731) (0.0196) (0.0812) (0.00327) (0.00625) (0.00351)(0.00538) (0.00360) (0.0124)(0.00934) (0.00293) (0.0156) (0.0120) (0.0117) (0.0120) (0.00979) (0.0120) (0.0143)(0.00810) (0.0204) (0.00894) (0.0251) (0.0106)(0.00697) (0.0256) (0.00984) (0.00887)(0.00555) (0.0105) (0.00605) (0.00581) (0.00908) (0.00608) (0.00580) (0.0211) (0.0122) (0.00364) (0.0157) (0.0127) (6.57e-06) (1.15e-05) (6.95e-06) (7.22e-06) (5.71e-06) (1.57e-05) (0.000694) (0.00245) (0.0103) (0.00285) (0.00108) (0.00424) 2 Table 3.14: Effect of Climate on Land Values - Including County and State-Year Fixed Effects January temperature 0.0157*** -0.0250** -0.0147 -0.0394*** -0.0773*** -0.0501** Population density square 9.45e-07 -7.22e-06January precipitation 2.51e-07April temperature -1.16e-06 0.00867April precipitation -0.0196July temperature -4.12e-06 -0.00674 -0.0223*July -0.0420** precipitation 0.00628 -0.0347*** 2.55e-05 October -0.0484 temperature -0.00818 -0.0189* -0.0959***October -0.0568*** precipitation -0.0188 -0.0204** 0.00106 0.0335*** -0.0177 0.0172* -0.0288** 0.00994 -0.0193 0.0175** 0.0362*** 0.0600***Cold Event -0.00347 -0.0264*** -0.0990*** 0.0185* 0.0231** -0.0291* -0.0112 -0.00455 -0.0206 0.00567 -0.000604 -0.00840 -0.0171 -0.00157** -0.0211 0.00459* 0.0173 0.0749 -0.0136 0.00174 0.0416*** 0.000267 0.00114 -0.000640 Population density 0.000389 0.00173 -0.00296 -0.00289 -0.000928 -0.0176* Hot EventN R The table shows estimation -0.0177***Dependent result variable: for log one of regression. Standard the 0.0154** errors value are of in theEstimation parentheses, land includes clustered per state on acre 0.0219*** year county*** fixed level. p effects <0.01, and ** county p<0.05, fixed 0.0208*** * effects. p<0.1 -0.00387 0.00370 101

Table 15A: Eﬀect of Climate on Land Values - Including State-Year Fixed Eﬀects - Reference Group: West4

Variables Base S SE C ENCen NE Population density *** *** *** *** *** *** Population density square *** ** *** *** *** *** January temperature *** ** * January precipitation *** *** *** *** April temperature * *** * ** * April precipitation ** * *** July temperature *** *** *** July precipitation *** *** ** ** October temperature * *** *** * October precipitation *** *** *** *** *** Hot event * Cold event ** ** The table shows estimation results of one regression. West4 region is dropped. Estimation includes state year ﬁxed eﬀects. *** p <0.01, ** p<0.05, * p<0.1

Table 15B: Eﬀect of Climate on Land Values - Including County and State-Year Fixed Eﬀects - Reference Group: West4

Variables Base S SE C ENCen NE Population density * Population density square January temperature *** ** *** *** ** January precipitation * *** *** April temperature ** ** April precipitation ** July temperature * *** *** July precipitation ** * ** October temperature *** *** * October precipitation ** *** Hot event *** ** *** *** ** Cold event ** * ** The table shows estimation results for one regression. West4 region is dropped. Estimation includes state year fixed effects and county fixed effects. *** p <0.01, ** p<0.05, * p<0.1 102

Table 16A: Eﬀect of Climate on Land Values - Including State-Year Fixed Eﬀects - Reference Group: South

Variables Base West4 SE C ENCen NE Population density *** *** *** *** *** *** Population density square *** ** *** *** ** *** January temperature *** *** ** *** *** ** January precipitation *** *** *** *** April temperature *** *** *** *** *** *** April precipitation * *** July temperature *** *** *** July precipitation *** *** October temperature *** *** *** *** October precipitation ** *** * Hot event * * *** * Cold event ** ** The table shows estimation results of one regression. South region is dropped. Estimation includes state year ﬁxed eﬀects. *** p <0.01, ** p<0.05, * p<0.1

Table 16B: Eﬀect of Climate on Land Values - Including County and State-Year Fixed Eﬀects - Reference Group: South

Variables Base West4 SE C ENCen NE Population density * Population density square * January temperature ** *** January precipitation *** April temperature *** ** *** April precipitation *** * July temperature *** *** *** ** July precipitation * *** October temperature *** October precipitation *** *** ** *** Hot event ** * * Cold event * The table shows estimation results of one regression. South region is dropped. Estimation includes state year fixed effects and county fixed effects. *** p <0.01, ** p<0.05, * p<0.1 103

Table 17A: Eﬀect of Climate on Land Values - Including State-Year Fixed Eﬀects - Reference Group: Southeast

Variables Base West4 S C ENCen NE Population density *** *** *** *** Population density square *** *** *** *** January temperature * ** ** *** ** January precipitation ** *** *** *** April temperature * *** ** *** April precipitation *** July temperature *** ** * July precipitation *** *** * ** October temperature *** *** *** *** *** ** October precipitation *** * Hot event * *** Cold event The table shows estimation results of one regression. Southeast region is dropped. Estimation includes state year ﬁxed eﬀects. *** p <0.01, ** p<0.05, * p<0.1

Table 17B: Eﬀect of Climate on Land Values - Including County and State-Year Fixed Eﬀects - Reference Group: Southeast

Variables Base West4 S C ENCen NE Population density ** Population density square January temperature *** * January precipitation * * *** April temperature * *** April precipitation *** July temperature ** July precipitation *** *** *** * October temperature *** *** *** *** *** * October precipitation * ** Hot event * *** * * Cold event The table shows estimation results of one regression. Southeast region is dropped. Estimation includes state year fixed effects and county fixed effects. *** p <0.01, ** p<0.05, * p<0.1 104

Table 18A: Eﬀect of Climate on Land Values - Including State-Year Fixed Eﬀects - Reference Group: Center

Variables Base West4 S SE ENCen NE Population density *** *** *** *** Population density square *** *** *** *** January temperature *** * *** *** January precipitation *** *** *** *** April temperature ** *** *** *** April precipitation *** *** *** *** *** July temperature *** *** ** July precipitation ** * October temperature *** *** *** October precipitation *** *** Hot event *** *** *** Cold event * ** The table shows estimation results of one regression. Center region is dropped. Estimation includes state year ﬁxed eﬀects. *** p <0.01, ** p<0.05, * p<0.1

Table 18B: Eﬀect of Climate on Land Values - Including County and State-Year Fixed Eﬀects - Reference Group: Center

Variables Base West4 S SE ENCen NE Population density *** *** *** *** Population density square January temperature ** *** *** January precipitation *** *** *** April temperature *** April precipitation *** ** *** *** *** July temperature ** *** *** ** ** July precipitation ** *** October temperature * *** October precipitation *** *** * Hot event * *** * * Cold event The table shows estimation results of one regression. Center region is dropped. Estimation includes state year fixed effects and county fixed effects. *** p <0.01, ** p<0.05, * p<0.1 105

Table 19A: Eﬀect of Climate on Land Values - Including State-Year Fixed Eﬀects - Reference Group: East North Center

Variables Base West4 S SE C NE Population density *** *** *** *** Population density square *** *** ** *** January temperature * *** ** January precipitation *** *** *** *** *** *** April temperature *** *** ** *** April precipitation *** July temperature *** *** * July precipitation ** ** October temperature *** * *** *** October precipitation *** * Hot event *** * * *** *** Cold event ** ** ** The table shows estimation results of one regression. ENCen region is dropped. Estimation includes state year ﬁxed eﬀects. *** p <0.01, ** p<0.05, * p<0.1

Table 19B: Eﬀect of Climate on Land Values - Including County and State-Year Fixed Eﬀects - Reference Group: East North Center

Variables Base West4 S SE C NE Population density Population density square * January temperature *** *** *** *** *** January precipitation *** *** *** *** *** *** April temperature *** ** *** *** *** April precipitation * * *** July temperature ** ** July precipitation * October temperature *** October precipitation ** Hot event ** * * Cold event ** The table shows estimation results of one regression. ENCen region is dropped. Estimation includes state year fixed effects and county fixed effects. *** p <0.01, ** p<0.05, * p<0.1 106

Table 20A: Eﬀect of Climate on Land Values - Including State-Year Fixed Eﬀects - Reference Group: Northeast

Variables Base West4 S SE C ENCen Population density *** *** *** *** *** Population density square *** *** *** *** *** January temperature ** January precipitation *** April temperature ** * *** *** *** April precipitation July temperature July precipitation October temperature ** October precipitation *** * * * Hot event Cold event The whole table shows estimation results of one regression. NE region is dropped. Estimation includes state year ﬁxed eﬀects. *** p <0.01, ** p<0.05, * p<0.1

Table 20B: Eﬀect of Climate on Land Values - Including County and State-Year Fixed Eﬀects - Reference Group: Northeast

Variables Base West4 S SE C ENCen Population density * * * Population density square * * January temperature ** ** * January precipitation *** April temperature April precipitation July temperature July precipitation October temperature * October precipitation *** *** ** Hot event Cold event The table shows estimation results of one regression. NE region is dropped. Estimation includes state year fixed effects and county fixed effects. *** p <0.01, ** p<0.05, * p<0.1 107

Figure 3.1: Climate Regions

Note: Through climate analysis, National Climatic Data Center scien- tist have identiﬁed nine climatically consistents regions within the contiguous United States. Source: http://www.ncdc.noaa.gov/monitoring-references/maps/us- climate-regions.php 108

Figure 3.2: Climate Data

Top: Coarse global climate model precipitation simulated over the U.S. (GCMs with 200 - 300 km spatial resolution) Bottom: Dynamically downscaling global ﬁelds using a regional climate model will allow us to explicitly account for changes in the intensity and frequency of extreme events at the local scale. (Spatial resolution of 35km - Dominguez et al., 2012) 109

REFERENCES

Adams, R. M. (1989). Global climate change and agriculture: an economic perspective. American Journal of Agricultural Economics, 71(5), pp. 1272–1279.

Adonizio, W., K. Nancy., and S. Royales (2012). Impact of the drought on corn exports: paying the price. Beyond the Numbers: Global Economy, 1(17), pp. 477–489.

Aleseyed, M., T. Rephann, and A. Isserman (1998). The Local Eﬀects Of Large Dam Reservoirs: U.S. Experience, 1975-1995. Review of Urban and Regional Development Studies, 10, pp. 91–108.

Anselin, L. and J. Le Gallo (2006). Interpolation of air quality measures in hedonic house price models: spatial aspects. Spatial Economic Analysis, 1(1), pp. 31–52.

Antle, J. M. and S. M. Capalbo (2010). Adaptation of agricultural and food systems to climate change: an economic and policy perspective. Applied Economic Perspectives and Policy, p. ppq015.

Atack, J. and R. A. Margo (2011). The Impact of Access to Rail Transportation on Agricultural Improvement: The American Midwest as a Test Case, 1850-1860. Journal of Transport and Land Use, 4(2).

Autobee, R. (1994). Rio Grande Project report. Technical report, Bureau of Recla- mation.

Barnard, C. H. (2000). Urbanization aﬀects a large share of farmland. Rural Con- ditions and Trends, 10(2), pp. 57–63.

Billington, D. P., D. C. Jackson, and M. V. Melosi (2005). The history of large federal dams: planning, design, and construction in the era of big dams. Government Printing Oﬃce.

Burt, O. R. (1986). Econometric modeling of the capitalization formula for farmland prices. American journal of agricultural economics, 68(1), pp. 10–26.

Castle, E. N. and I. Hoch (1982). Farm real estate price components, 1920–78. American Journal of Agricultural Economics, 64(1), pp. 8–18.

Cech, T. V. (2010). Principles of water resources: history, development, management, and policy. John Wiley & Sons. 110

Deschenes, O. and M. Greenstone (2007). The economic impacts of climate change: evidence from agricultural output and random ﬂuctuations in weather. The Amer- ican Economic Review, pp. 354–385.

Diﬀenbaugh, N. S., J. S. Pal, R. J. Trapp, F. Giorgi, and S. H. Schneider (2005). Fine-scale processes regulate the response of extreme events to global climate change. Proceedings of the National Academy of Sciences of the United States of America, 102(44), pp. 15774–15778.

Dominguez, F., E. Rivera, D. Lettenmaier, and C. Castro (2012). Changes in winter precipitation extremes for the western United States under a warmer climate as simulated by regional climate models. Geophysical Research Letters, 39(5).

Donaldson, D. and R. Hornbeck (2013). Railroads and American Economic Growth: A “Market Access” Approach. Technical report, National Bureau of Economic Research.

Duﬂo, E. and R. Pande (2007). Dams. The Quarterly Journal of Economics, 122(2), pp. 601–646.

Durlauf, S. N. and P. A. Johnson (1995). Multiple regimes and cross-country growth behaviour. Journal of applied econometrics, 10(4), pp. 365–384.

Eckstein, O. (1958). Water resource development-the economics of project evalua- tion. Cambridge, Mass.: Harvard Univ. Pr.

Featherstone, A. M. and T. G. Baker (1987). An examination of farm sector real asset dynamics: 1910–85. American Journal of Agricultural Economics, 69(3), pp. 532–546.

Fisher, A. C. (2001). Uncertainty, irreversibility, and the timing of climate policy. Pew Center on Global Climate Change working paper, Arlington, VA.

Fogel, R. W. (1994). Railroads and American economic growth. Cambridge Univ Press.

Graham, L. O. (1950). The Central Valley Project: Resource Development of a Natural Basin. California Law Review, pp. 588–637.

Hansen, Z., G. Libecap, and S. Lowe (2011). The Political Economy of Major Water Infrastructure Investments in the Western United States and the Impact on Agriculture: An Historical Analysis. Technical report, Working Paper.

Hertel, T. W., M. B. Burke, and D. B. Lobell (2010). The poverty implications of climate-induced crop yield changes by 2030. Global Environmental Change, 20(4), pp. 577–585. 111

Holland, P. W. (1986). Statistics and causal inference. Journal of the American statistical Association, 81(396), pp. 945–960.

Hornbeck, R. and P. Keskin (2012). Does agriculture generate local economic spillovers? short-run and long-run evidence from the ogallala aquifer. Techni- cal report, National Bureau of Economic Research.

Howe, C. W. (1968). Water resources and regional economic growth in the United States, 1950-1960. Southern Economic Journal, pp. 477–489.

Imbens, G. W. and T. Lancaster (1994). Combining micro and macro data in microeconometric models. The Review of Economic Studies, 61(4), pp. 655–680.

Jardine, A., R. Merideth, M. Black, and S. LeRoy (2013). Assessment of climate change in the southwest United States: a report prepared for the National Climate Assessment. Island press.

Kelly, D. L., C. D. Kolstad, and G. T. Mitchell (2005). Adjustment costs from environmental change. Journal of Environmental Economics and Management, 50(3), pp. 468–495.

Kitchens, C. (2014). The Role of Publicly Provided Electricity in Economic De- velopment: The Experience of the Tennessee Valley Authority, 1929–1955. The Journal of Economic History, 74(02), pp. 389–419.

Leung, L. R. and Y. Qian (2009). Atmospheric rivers induced heavy precipitation and ﬂooding in the western US simulated by the WRF regional climate model. Geophysical research letters, 36(3).

Livanis, G., C. B. Moss, V. E. Breneman, and R. F. Nehring (2006). Urban sprawl and farmland prices. American Journal of Agricultural Economics, 88(4), pp. 915–929.

Lobell, D. B. and M. B. Burke (2008). Why are agricultural impacts of climate change so uncertain? The importance of temperature relative to precipitation. Environmental Research Letters, 3(3), p. 034007.

Lobell, D. B. and C. B. Field (2007). Global scale climate–crop yield relation- ships and the impacts of recent warming. Environmental research letters, 2(1), p. 014002.

Los Angeles Times (1900). Not a Thing of The Past: National Irrigation not a Dead Issue, Nov 25, 1900; pg. 2.

Los Angeles Times (1902). The struggle for National Irrigation.: How a Congres- sional Battle Was Fought and Won. Aug 17, 1902; pg. C10. 112

Maxwell, G. H. (accessed July 29, 2015). The George Hebard Maxwell Papers, MG 1, 1903-1905. Url- http://www.azarchivesonline.org/xtf/view?docId=ead/asl/MG1GeorgeMaxwell.xml.

McCune, C. J. (2001). Belle Fourche Project. Historic Reclamation Projects. Denver: Bureau of Reclamation.

Mendelsohn, R., W. D. Nordhaus, and D. Shaw (1994). The impact of global warming on agriculture: a Ricardian analysis. The American economic review, pp. 753–771.

Mesinger, F., G. DiMego, E. Kalnay, K. Mitchell, P. C. Shafran, W. Ebisuzaki, D. Jovic, J. Woollen, E. Rogers, E. H. Berbery, et al. (2006). North American regional reanalysis. Bulletin of the American Meteorological Society, 87(3), pp. 343–360.

Miller, E. W. and R. M. Miller (1992). Water quality and availability: a reference handbook. ABC-CLIO.

Mirghasemi, S. (2013). Politics and Dam Construction: Historical Evidence from the Western U.S. Working paper.

Nelson, G. C., M. W. Rosegrant, J. Koo, R. Robertson, T. Sulser, T. Zhu, C. Ringler, S. Msangi, A. Palazzo, M. Batka, et al. (2009). Climate change: Impact on agriculture and costs of adaptation, volume 21. Intl Food Policy Res Inst.

Neumann, J. E. and R. Mendelsohn (1999). The Impact of Climate Change on the United States Economy. Cambridge University Press.

New York Times (1901). Irrigation Bill Passed: The House Adopts It by Vote of 146 to 55. Jun 14, 1902; pg. 8.

Newell, F. H. (1905). Annual Report of the Reclamation Service, Volume 3, Issue 1. Geological Survey (U.S.). Reclamation Service.

Nordhaus, W. D. (1991). To slow or not to slow: the economics of the greenhouse eﬀect. The economic journal, pp. 920–937.

Padgitt, M., D. Newton, R. Penn, and C. Sandretto (2000). Production practices for major crops in US agriculture, 1990–97. Stat. Bull. no. SB969. Econ. Res. Serv., Washington, DC.

Parry, M. L., C. Rosenzweig, A. Iglesias, M. Livermore, and G. Fischer (2004). Eﬀects of climate change on global food production under SRES emissions and socio-economic scenarios. Global Environmental Change, 14(1), pp. 53–67. 113

Plantinga, A. J., R. N. Lubowski, and R. N. Stavins (2002). The eﬀects of potential land development on agricultural land prices. Journal of Urban Economics, 52(3), pp. 561–581.

Reilly, J., F. Tubiello, B. McCarl, D. Abler, R. Darwin, K. Fuglie, S. Hollinger, C. Izaurralde, S. Jagtap, J. Jones, et al. (2003). US agriculture and climate change: new results. Climatic Change, 57(1-2), pp. 43–67.

Reisner, M. (1993). Cadillac desert: The American West and its disappearing water. Penguin.

Reuss, M. and P. K. Walker (1983). Financing water resources development. Wash- ington, DC: US Army Corps of Engineers (EP 870-1-13).

Richard, S. T. (1995). The damage costs of climate change toward more comprehensive calculations. Environmental and Resource Economics, 5(4), pp. 353–374.

Schimmelpfennig, D., J. Lewandrowski, J. Reilly, M. Tsigas, I. Parry, R. Mendelsohn, and T. Mount (1996). Agricultural adaptation to climate change. Economic Research Service, USDA, Agricultural Economic Report. URL http://www. ers. usda. gov/media/490977/aer740a_1_. pdf, (740).

Schlenker, W., W. M. Hanemann, and A. C. Fisher (2005). Will US agriculture really beneﬁt from global warming? Accounting for irrigation in the hedonic approach. American Economic Review, pp. 395–406.

Schlenker, W., W. M. Hanemann, and A. C. Fisher (2006). The impact of global warming on US agriculture: an econometric analysis of optimal growing conditions. Review of Economics and Statistics, 88(1), pp. 113–125.

Schlenker, W. and M. J. Roberts (2009). Nonlinear temperature eﬀects indicate severe damages to US crop yields under climate change. Proceedings of the National Academy of sciences, 106(37), pp. 15594–15598.

Severnini, E. (2014). The power of hydroelectric dams: Agglomeration spillovers. IZA Discussion Paper, 8082, pp. 1–69.

Tebaldi, C., K. Hayhoe, J. M. Arblaster, and G. A. Meehl (2006). Going to the extremes. Climatic change, 79(3-4), pp. 185–211.

The Washington Post (1901). President Roosevelt’s Message To Congress: Tribute To Mr. Mkinley,´ Dec. 3, 1901. pg. 13.

Trenberth, K. E. (1999). Conceptual framework for changes of extremes of the hydrological cycle with climate change. In Weather and Climate Extremes, pp. 327–339. Springer. 114

United States House of Representatives (1926). Executive Document 308, 69th Congress, 1st session.

Waldinger, M. (2014). The Economic Eﬀects of Long-Term Climate Change: Ev- idence from the Little Ice Age, 1500-1750. Technical report, London School of Economics - Working Paper.

Widtsoe, J. (1928). Success on irrigation projects.