An Analysis of the Impact of Drought

on Agriculture, Local Economies,

Public Health, and Crime Across the
Western United States
Dale Manning, Jesse Burkhardt,
Chris Goemans, Alex Maas

1
 An Analysis of the Impact of Drought on Agriculture,
Local Economies, Public Health, and Crime Across
the Western United States

Dale Manning Jesse Burkhardt

Associate Professor Assistant Professor
Department of Agricultural and Department of Agricultural and
Resource Economics Resource Economics
Colorado State University Colorado State University

Chris Goemans Alex Maas

Associate Professor Assistant Professor
Department of Agricultural and Department of Agricultural
Resource Economics Economics and Rural Sociology
Colorado State University University of Idaho

Executive Summary
Overview

We estimate the average impact of drought on a wide-variety of economic and social indicators over a
forty-year period across the western U.S. Estimated impacts are then used to simulate the specific
effects of particular droughts. Importantly, our approach distinguishes between the effect of
prolonged drought conditions (as measured by PDSI) and the impact of contemporaneous variation in
temperature and precipitation. Thus, simulated impacts reflect the combined effect associated with
both prolonged drought conditions and short-term weather conditions within a given year.

Our analysis utilizes county level data from the last 40 years to generate estimates of the average
impact that monthly drought conditions over the course of a year have on each of the outcome
indicators considered. In addition to generating average estimates, we also explore how the timing of
drought conditions within a year, as well as whether a county is groundwater dependent, impacts the
average effect.

Estimates generated from the statistical analysis are then used to simulate the effects of specific
drought years on specific counties. While this report examines the impact of periods of extreme
drought and extreme wet, the following summary of results focuses on extreme drought which was
the original motivation for this study. References to positive or negative effects in the following
summary are only made if the estimated impact was statistically different from zero.

1
Summary of Results

Agricultural Productivity Impacts

• Prolonged drought has a negative and statistically significant impact on corn, hay,
sorghum, and wheat production.1 Each additional month of extreme drought conditions
(PDSI<-3) is associated with a decrease in total corn and wheat production of 3.2 and 3.6
percent, respectively.
• All else constant, higher temperatures and precipitation have a positive and statistically
significant impact on total production; however, these effects are diminishing with higher
temperatures and become negative for conditions well above the mean.
• Heterogeneity exists in terms of the average effect of an additional month of drought on
crop production. Not surprisingly drought conditions that occur during the growing
season have a much larger negative impact on total production than those that occur
during the non-growing season.
• For corn and wheat, the reduction in total production is largely driven by a decrease in
harvested acres, whereas for hay the reduction in total production is primarily due to a
decrease in yields per acre harvested.
• Although less clear, the impacts on agricultural production appear to be lower in counties
with higher median household incomes.
• In the Appendix, we include estimates of the impact of watershed-level drought (drought
measures at the HUC level). We find that the impact of a watershed-level drought is
qualitatively similar to the impact of a county specific drought, which is unsurprising given
that the two are correlated. However, the watershed-level drought impacts are muted in
comparison to the county specific drought.

Impacts to Wages, Employment and Number of Establishments

• The effect of extreme drought conditions on total wages paid to employees across all sectors
is not statistically different from zero, as total wages decrease in some industries and increase
in others.
o For agriculture supply and agricultural, fishing, and hunting, each additional month of
extreme drought within a given year is associated with a reduction in total wages paid
to employees in that sector of approximately 1.2 percent and 0.5 percent, respectively.
o Extreme drought has a positive association with wages paid to employees in the
public utilities.

1
We chose these crops because they are the most abundant in the USDA QuickStats data. Data on other crops
are sparsely available.

2
 o An additional month of extreme drought reduces wages in the broader recreation and
entertainment industry by 0.59 percent.
• Similar to wages, the effect of extreme drought on total employment is not significantly
different from zero, however, the result is mixed across different sectors.
o Each additional month of extreme drought within a given year is associated with a
reduction in total employment in the agricultural supply, livestock production, and
agricultural, fishing, and hunting sectors of approximately 1.2 percent, 0.6 percent,
and 0.5 percent respectively.
o Each additional month of extreme drought also has a negative impact on the
recreation and entertainment sectors.
• Severe drought leads to a modest, but statistically significant decrease in the overall number
of businesses. The negative effect is largely due to decreases in the recreation and
entertainment, service, agricultural supply, and trade sectors.
o The percent decrease in the number of agricultural establishments is largest in the
recreation and entertainment (0.5%) and agricultural supplier sectors (0.7%).

Health Outcomes

• Drought is generally uncorrelated with annual, county-level health outcomes, with the
exception that an additional month of extreme drought is associated with a 0.5 percent
reduction in heart disease mortality.

Criminal Activity

• Drought is generally uncorrelated with annual, county-level criminal activity, with the
exception that an additional month of moderate drought is associated with small
reductions in simple assaults and property crimes.

Insurance Outcomes

• An additional month of extreme drought is associated with a 10 percent and 7 percent

increase in indemnity amounts and insurance policy counts, respectively. An additional
month of moderate drought is associated with a 6.5 percent and 4.7 percent increase in
indemnity amounts and insurance policy counts, respectively.
• Drought is uncorrelated with liability amounts.

Total Taxes and Total Revenue

• An additional month of extreme drought is associated with a 0.4 percent increase in total
revenue, while an additional month of moderate drought is associated with a 0.3 percent
increase in total taxes.

3
Introduction
Severe, extreme, or exceptional drought conditions have become increasingly common throughout
the western United States over the past 20 years. Abnormally dry conditions can lead to a wide range
of negative economic impacts across a wide range of sectors that either directly or indirectly depend
on weather-dependent ecosystem services. Examples include agricultural production, where yields
are negatively impacted by inadequate water supplies, and tourism, where lower stream flows and
snowpack can change the attractiveness of both winter and summer tourism activities. In both of
these instances, the impacts are likely to extend beyond just those sectors directly impacted and
beyond traditional economic indicators. The goal of this study is twofold. First, utilizing approximately
40 years of county level data from across the western United States2, we attempt to empirically
identify the impact of drought on society. This includes characterizing how drought impacts
agricultural productivity, employment, wages paid, business openings/closings, criminal activity, and
health outcomes. Our second objective is to develop a low-cost, spatially and temporally
generalizable means of simulating the impact of specific droughts on society. This information not
only helps in terms of responding to the current drought, but also aims to inform planners and
policymakers of the potential changes that will result as drought becomes more frequent.

Water provides a critical input to many resource-based activities that make up a large part of the
economy throughout the western US. For example, water shortages in agriculture can lead to
reductions in yield and revenue in the immediate term, while lower stream flows can impact demand
for boating and fishing activities and extreme heat can deter potential tourists from visiting a region.
As drought becomes more frequent, we will likely see changes throughout the entire economy. For
example, lower snowpack attracts fewer skiers and winter sports enthusiasts, leading to decreased
demand for hotels, restaurants and other services utilized by tourists. As tourism-oriented sectors are
directly affected, input suppliers may also be impacted, creating economy-wide ripples that lead to
lower incomes and government revenue for the provision of public services.

Previous studies examining drought in Colorado have largely focused on the direct impacts to
agriculture and the indirect effects associated with sectors related to agriculture. They have also
typically focused on the impacts to a single region stemming from a particular drought. Examples
include Goemans et al. (2013), Pritchett et al. (2013), and Bauman et al. (2014). These studies have
attempted to estimate the magnitude of the impacts on agriculture resulting from drier than normal
conditions while also detailing the economy-wide implications of lost agricultural activity. Direct
economic impacts to other sectors (and their corresponding indirect effects) have largely been
ignored given the focus on agriculture. The results presented herein represent the cumulative, general

2
Specifically, our analysis includes data from the following states: Arizona, California, Colorado, Idaho, Kansas,
Montana, Nebraska, New Mexico, Oklahoma, Oregon, Texas, Utah, Washington, and Wyoming.

4
equilibrium impacts associated with the combined direct and indirect impacts of drought to all
economic sectors.

Furthermore, despite anecdotal evidence suggesting that drought’s impacts extend beyond
traditional economic indicators3, little work has examined how drought impacts health and crime
outcomes. This study aims to assess the impacts of the 2018 drought on agriculture, providing a point
of comparison to previous droughts, but also to broaden the scope of indicators examined. Unlike
previous efforts, this study uses an empirical approach (multivariate regression) that assesses impacts
using actual experiences with drought over time. We utilize this approach as it requires few
assumptions relative to the structural models traditionally used in conducting drought impact
analysis and is easily transferable across regions/droughts.

Methodological Approach and Data

Estimating Marginal Impacts

In the simplest of terms, drought corresponds to the absence of water.4 Available water supply and
demand are the cumulative result of historic weather conditions combined with contemporaneous
weather conditions.5 In this sense drought conditions can be characterized in terms of a stock and
flow. Weather events leading up to any given year result in a “stock” of available water to which
contemporaneous weather events in that year build off of (i.e., the “flow”). To understand the
importance of making this distinction consider the impact of a period of excessive rainfall (flow) that
occurs following a prolonged period with low soil moisture (stock) as compared to one that occurs
when soils are extremely wet.

With this distinction in mind, we utilize the following base statistical model to capture the short-term
impact of prolonged drought (i.e. the stock of available water) and contemporaneous weather (i.e. the
flow) on a range of outcomes (e.g., corn acres harvested).

𝑙 𝑛(𝑌𝑖𝑡𝑗 ) 𝑘
= 𝜃𝑊 𝑊𝑒𝑎𝑡ℎ𝑒𝑟𝑖𝑡 + ∑ 𝛽𝑘 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑖𝑡 𝑗
+ 𝛼𝑖 + 𝛾𝑡 + 𝜖𝑖𝑡 . (1)
𝑘=1

3
https://www.denverpost.com/2018/10/03/colorado-farmers-mental-health/
4
Multiple definitions of drought exist, some specific to the impacted area (e.g., agricultural drought, hydrologic
drought, etc.). For a more comprehensive discussion of the different types and causes of drought see:
https://drought.unl.edu/Education/DroughtIn-depth/TypesofDrought.aspx
5
Contemporaneous demand for water depends on contemporaneous weather conditions in so far as
households and farms require water for irrigation. For example, if there is less precipitation in a given year, the
demand for municipal and irrigation water will increase.

5
𝑌𝑖𝑡𝑗 is outcome 𝑗 in county 𝑖 in year 𝑡. A complete list of outcomes is provided below in Table 1.
𝑊𝑒𝑎𝑡ℎ𝑒𝑟𝑖𝑡 is a vector of contemporaneous weather variables including annual precipitation, annual
precipitation squared, annual average temperature, and annual average temperature squared.
𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑖𝑡 𝑘
is intended to reflect the stock effect and corresponds to the number of months in a given
year that fall into drought category 𝑘.

Given this specification, 𝜃𝑊 provides an estimate of the impact of contemporaneous weather

conditions (the flow effect) on each outcome whereas 𝛽𝑘 provides an estimate of the impact of an
additional month of 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦 𝑘 on a particular outcome.

One obvious concern is that agricultural and economic outcomes (𝑌𝑖𝑡𝑗 ) are likely correlated with other
non-drought variables, and if those variables are not accounted for, our drought impact estimates will
be biased. To address such concerns, we include additional variables as controls for potentially
confounding factors between outcomes 𝑌𝑖𝑡𝑗 and drought. Specifically, we include county fixed effects
(𝛼𝑖 ) which control for any time-invariant county specific unobserved variables such as average soil
type, elevation, and time-invariant management practices. We also include year fixed effects (𝛾𝑡 ) to
account for any time-varying unobserved variables such as macroeconomic trends, commodity price
trends, and geographically broad production trends. All standard errors are clustered at the county
level.

Simulating Drought Impacts

The statistical methodology outlined in Equation 1 produces the average marginal effect of changes in
weather and the PDSI over a 40-year period across the entire West. In this section, we use our model
to produce disaggregated estimates for particular counties in particular years. To do so, we combine
county specific drought, temperature, and precipitation data with the output from our model in
Equation 1. The procedure proceeds in two steps. First, we estimate the model in Equation 1 to
produce estimates of the model parameters including 𝜃̂𝑊 and 𝛽̂𝑘 .6 Second, we plug the county-year
specific data into the model output to generate county-year specific drought impact estimates. For
instance, for Weld County, Colorado in 2012, we could write:

Impact YWeld,2012
= 𝜃̂𝑊 (𝑊𝑒𝑎𝑡ℎ𝑒𝑟𝑊𝑒𝑙𝑑,2012 − 𝑊𝑒𝑎𝑡ℎ𝑒𝑟𝑊𝑒𝑙𝑑 )
4

+ ∑ 𝛽̂𝑘 (𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑊𝑒𝑙𝑑,2012
𝑘 𝑘
− 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑊𝑒𝑙𝑑 ) (2)
𝑘=1

6
Estimated parameters, for example 𝛽, are denoted with hats as in “𝛽̂”.

6
where Impact YWeld,2012 corresponds to the proportional deviation from normal for outcome 𝑌 in Weld
county, CO associated with the temperature, precipitation and 𝑃𝐷𝑆𝐼 conditions for 2012.
𝑊𝑒𝑎𝑡ℎ𝑒𝑟𝑊𝑒𝑙𝑑,2012 includes the data on precipitation, precipitation squared, temperature, and
temperature squared for Weld County in 2012. 𝑊𝑒𝑎𝑡ℎ𝑒𝑟𝑊𝑒𝑙𝑑 represents the mean values for each
variable for Weld county over the period of analysis. Likewise, 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑊𝑒𝑙𝑑,2012
𝑘
includes the PDSI
data for Weld County in 2012 and 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑊𝑒𝑙𝑑
𝑘
corresponds to the mean number of months within a
year in each category for Weld county. The impact from our simulation, when multiplied by 100, is the
estimated percentage change in outcome 𝑌 relative to what could have been expected under normal
conditions.

Data
The variable 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦 is created using the Palmer Drought Severity Index (PDSI) sourced from the
Climate Engine, which uses data from gridMET (Abatzoglou et. al, in press). 7 PDSI is designed to
capture droughts occurring on time scales longer than 12 months, or to capture the presence of
prolonged drought (Dai, 2019). PDSI is a water budget based approach that is calculated using a
variety of water supply and demand variables to develop a cumulative measure of water deficit
excess.

We bin PDSI values into one of 5 categories. Specifically, to calculate the number of months in each
year corresponding to category 𝑘 we sum the number of months within a given county (𝑖) and year (𝑡)
that fall within each of the following categories: less than -3, -1 to -3, -1 to 1, 1 to 3 and greater than 3.
Moving forward we refer to each of these categories as extreme drought, moderate drought, normal
conditions, moderate wet, and extreme wet respectively. For example, 𝐶𝑎𝑡𝑒𝑔𝑜𝑟𝑦𝑖𝑡 1
is equal to the
number of months that county (𝑖) experienced in year (𝑡) that were classified as extreme drought.
When estimating Equation 1, the “normal conditions” bin is omitted so that the coefficients on the
drought conditions and wet conditions bins are interpreted as relative to normal conditions. Thus, the
coefficient estimates, when multiplied by 100, represent the average percent change in the outcome
variable associated with 1 additional month in each particular category. One advantage to binning
PDSI in this way (as opposed to, for example, only counting the number of months for which PDSI<-3)
is that we are able to identify the effect of abnormal conditions, whether dry or wet, on each of the
outcome indicators.

We also estimate a series of models in which we separate the bins into growing season and non-
growing season aggregates. This is done to estimate the impact of drought at different times of the

7
We note that this is the same data that can be downloaded from Google Earth Engine. NOAA also has PDSI data,
however, NOAA’s data is aggregated to the climate division level, which is larger than the county. In email
exchanges with NOAA, we also discovered that the data presented in the maps on their website is not always
accurate.

7
year. The growing season is defined as April-September while the non-growing season is defined as
October-March.

Table 1A: Outcome Indicators and Data Sources

Category Variable Name Data Geographic Temporal Temporal

Source Scale Scale Coverage

Employment, Wages, and Number of

Establishments

Employment QCEW County Monthly 1980-

2016

Total Wages Paid QCEW County Monthly 1980-

2016

Number of QCEW County Monthly 1980-

Establishments 2016

Agricultural

Total Production NASS County Annual 1980-

2016

Acres Harvested NASS County Annual 1980-

2016

Yield per Acre NASS County Annual 1980-

2016

Health

Opioid CDC County Monthly 1980-

2016

Alcohol CDC County Monthly 1980-

2016

Heart Disease CDC County Monthly 1980-

2016

8
 Criminal Activity

Aggravated Assault FBI-UCRS County Monthly 1980-

2016

Simple Assault FBI-UCRS County Monthly 1980-

2016

Property Crime FBI-UCRS County Monthly 1980-

2016

Violent Crime FBI-UCRS County Monthly 1980-

2016

Insurance

Liability Amount RMA County 1989-

2010

Indemnity Amount RMA County 1989-

2010

Count of Indemnity RMA County 1989-

Policies 2010

Total Revenue and Total Taxes

Total Revenue Government County 1980-

Finance 2015
Database8

Total Taxes Government County 1980-

Finance 2015
Database

8
Pierson K., Hand M., and Thompson F. (2015). The Government Finance Database: A Common Resource for
Quantitative Research in Public Financial Analysis. PLoS ONE doi: 10.1371/journal.pone.0130119

9
Table 1B: Right Hand Side Variables and Data Sources

Variable Data Source Geographic Temporal Temporal Coverage

Name Scale Scale

PDSI Climate Engine (GridMET) County Monthly 1980-2020

Temperature PRISM County Daily 1980-2020

Precipitation PRISM County Daily 1980-2020

Table 2: QCEW Sector Mapping

Variable Name NAICS Sector Name NAICS Code

All All Industries

AgFishHunt Agriculture, Forestry, fishing, and hunting 1

Crop Production Crop Production 111

Livestock Animal Production and Aquaculture 112

Ag Supply Support Activities for Agriculture and Forestry

Recreation and Arts, Entertainment, and Recreation 71

Entertainment

Fish Fishing, Hunting and Trapping 114

Travel Traveler Accommodation, Drinking Places, 7211, 7224, 7225,

Accommodation Restaurants and Other Eating Places, Beer, Wine, and 4453, 1026
Liquor Stores, Leisure and Hospitality

Education and Health Educational Services, Health Care and Social 6

Assistance

Manufacturing Manufacturing 3

Other Public Administration 9

10
 Service Information, Finance and Insurance, Real Estate and 50, 81
Rental Leasing, Professional, Scientific and Technical
Services, Management of Companies and Enterprises,
Admin and Support, Other Services

Trade Wholesale Trade, Retail Trade, Transportation and 4

Warehousing

Utilities and Construction Utilities, Mining, Quarrying, Oil and Gas Extraction, 2
and Construction

Note: Sectors contain all sub-sectors that begin with the NAICS code listed above. For example,
EducHealth includes NAICS sectors 61 and 62. Complete definition for all sectors can be found online
at: https://www.census.gov/cgi-bin/sssd/naics/naicsrch?chart=2017

11
Results
We begin by presenting results of estimating Equation 1 for each outcome variable in our data. First,
we present agricultural impacts. Agricultural impacts are presented graphically whereas the full
regression results tables are presented for the other outcomes 9. Second, we present other non-
agricultural economic sector impacts including the impact of drought on wages and number of
establishments in each sector. We conclude by presenting the impacts of drought on public health,
crime, insurance measures, and government revenue and taxes.

Agricultural Impacts
Figure 1 illustrates the impact of an additional month in each PDSI category (the bins defined above)
on total production for each of the four crops analyzed as part of this study. We chose to analyze corn,
hay, wheat, and sorghum because there were too few observations of other crops such as dry beans
and potatoes to provide statistically meaningful results. For each PDSI category, the dot corresponds
to the estimated proportional impact of an additional month in that category, relative to baseline
conditions (−1 ≤ 𝑃𝐷𝑆𝐼 ≤ 1); the lines extending out from the dot cover the 95 percent confidence
interval. When multiplied by 100 the point estimate provides the estimated percentage change in each
outcome indicator. Consistent with expectations, all else equal, a month of extreme or moderate
drought has a negative and statistically significant impact on total production. Each month of extreme
drought (PDSI<-3) reduces total corn production by approximately 3.2 percent, hay production by
approximately 2.2 percent, wheat production by approximately 3.8 percent, and sorghum production
by approximately 2.7 percent. Similarly, each month of moderate drought (-1>PDSI>-3) reduces corn,
hay, wheat, and sorghum production by 0.9 percent, 0.8 percent, 1.9 percent, and 1.7 percent
respectively. In contrast, we find that abnormally wet periods slightly decrease corn, hay and wheat
production, but increase sorghum production.

To better understand the effect of pro-longed drought on total production we separately examine
how yields and acres harvested are impacted by different PDSI conditions.10 These results are
presented in Figures 2 and 3. Figure 2 presents the proportional impacts of an additional month in
each PDSI category on per acre yields for corn, hay, wheat, and sorghum. We find that an additional
month of extreme drought (PDSI<-3) reduces corn, hay, wheat, and sorghum yields by approximately
1.5, 1.9, 1.6, and 2.5 percent respectively. An additional month of moderate drought (-3<PDSI<-1)
decreases hay, wheat, and sorghum yield by approximately 0.5, 0.8, and 0.7 percent respectively.
Finally, an additional month of extreme wet conditions (PDSI>3) decreases corn, hay, wheat, and
sorghum yields by 1, 0.3, 0.8, and 0.9 percent respectively.

9
The full set of results for agriculture are presented in the Appendix.
10
Total production is the production in lbs, bushels, or other physical measurement per county per year. Total
acres harvested is measured in acres per county per year. Yield is the ratio of production per acres harvested.

12
Figure 3 presents the proportional impacts of an additional month in each PDSI category on harvest
for corn, hay, wheat, and sorghum. We find that an additional month of extreme drought (PDSI<-3)
reduces corn, hay, and wheat harvests by approximately 1.6, 0.4, and 2 percent respectively. An
additional month of moderate drought (-3<PDSI<-1) decreases corn, hay, and wheat harvests by
approximately 0.7, 0.4, and 1 percent respectively. An additional month of moderately wet conditions
(1<PDSI<3) does not statistically significantly impact hay, wheat, and sorghum harvests. Sorghum
harvests are positively correlated wet conditions.

The results suggest that, at least for corn and wheat, the reduction in total production is largely driven
by a decrease in harvested acres as opposed to a reduction in yield. For these two crops, the results
suggest (consistent with Manning et al. 2016) that producers respond by adjusting on the extensive
margin (i.e., concentrating limited resources on limited number of acres), as opposed to the intensive
margin (i.e., maintaining the same number of acres, but reducing the quantity of water resources
dedicated to each acre). This has important implications for economy-wide impacts as a reduction in
acres harvested is likely to result in larger impacts to backward-linked industries than those
associated with a reduction in per acre yields. The full set of results for agriculture are presented in
Tables A1-A3 in the Appendix.

13
Figure 1: Proportional Drought Impacts on Corn, Hay, Wheat, and Sorghum Total Production

14
Figure 2: Proportional Drought Impacts on Corn, Hay, Wheat, and Sorghum per Acre Yields

15
Figure 3: Proportional Drought Impacts on Corn, Hay, Wheat, and Sorghum Harvested Acres

16
Heterogeneity in Drought Impacts on Total Production
In addition to estimating average marginal impacts, we also explore whether drought impacts depend
on the within year timing of drought conditions (e.g., growing season versus non-growing season), the
extent to which an area has historically been dependent on groundwater for production, and the
geographic scope of the drought. Figure 4 presents the growing season versus non-growing season
drought impacts on total production for corn and hay. We find that an additional month of extreme
drought (-3<PDSI) during the growing season decreases corn and hay production by 4.8 and 4.2
percent respectively, substantially more than the average annual drought impacts presented above.
Additionally, an additional month of moderate drought (-3<PDSI<-1) during the growing season
decreases hay production by approximately 0.8 percent but does not statistically impact corn
production. In contrast, moderately wet and extreme wet months during the growing season have
opposing effects on corn and hay production: corn production decreases under wet growing season
conditions while hay production is largely unaffected.

Drought conditions in the non-growing season also impacts corn and hay production in interesting
ways. Drought in the non-growing season is negatively correlated with corn production, but the
results are statistically insignificant. Whereas moderate drought in the non-growing season is
statistically negatively correlated with hay production. Moderately wet and extreme wet months
during the non-growing season substantially increase corn production but decrease hay production.

Complete growing season and non-growing season results for all crops, including the impact on total
production, per acre yields, and harvested acres are presented in Appendix A7-A9. Figure 4 highlights
the importance of the timing of drought conditions within a year and how this importance varies
depending on the crop in question. Corn and hay production are clearly more sensitive to drought
during the growing season. Interestingly, wet conditions during the non-growing season substantially
improve corn production but do not appear to improve hay production. These discrepancies highlight
the differences in how each crop utilizes water over the course of the year.

Figure 5 evaluates whether the impacts of drought on corn and hay production depend on whether a
county is dependent on groundwater. The results show that the impact of drought on corn production
is largely unaffected by the dependence on groundwater. In contrast, the impact of drought on hay
production is largely remedied by groundwater. Again, full regression results for these crops and
wheat and sorghum are presented in Tables A4-A6 Appendix.

We also examine the relationship between drought impacts and median household income within a
county. To do so, we interact the PDSI bins with the county level 5-year average median household
income from the from American Consumer Survey (US Census Bureau, 2010). The results are
presented in Table 3 below. The impact of drought on each crop is qualitatively similar to the non-
interacted models presented in the figures above, however the standard errors on the point estimates
tend to increase. The important takeaway from this table is that when statistically significant, the

17
interaction terms are positive indicating that higher incomes tend to reduce the impact of drought
(the positive interaction term is added to the negative drought impact, reducing the overall
magnitude of the impact).

Finally, we explore the impact of drought at larger geographic extents. To do so, we calculate the PDSI
bins at the hydrologic unit code (HUC) subregion level. The HUCs are analogous to watersheds. The
results are presented in Tables A10 and A11 in the Appendix. Table A10 includes both the county level
PDSI bins and the HUC level PDSI bins, whereas Table A11 includes only the HUC level PDSI bins.
Watershed level drought and county specific drought are highly correlated and the results reflect this.
We find that an additional month of a HUC level drought has qualitatively the same effect as a county
level drought, but the magnitude of the effect is smaller. This is likely due to the fact that the county
drought measures are localized to the county, whereas the HUC level drought measures include
drought conditions in other parts of the country, which introduces measurement error into the
drought variables.

Table 3: Agricultural Results with PDSI Interacted with Median Household Income Per County

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

VARIABLES Corn Hay Wheat Sorghum

PDSI<-3 -0.0191 -0.0220** -0.0743*** -0.0669**

(0.0213) (0.0105) (0.0211) (0.0313)

PDSI -3 to-1 -0.0396** 0.00216 -0.0270** -0.0466***

(0.0154) (0.00658) (0.0109) (0.0150)

PDSI 1 to 3 -0.0147 0.00486 -0.0320** -0.00338

(0.0141) (0.00562) (0.0128) (0.0136)

PDSI >3 -0.0630*** -0.0164*** -0.0170 0.0228

(0.0172) (0.00628) (0.0134) (0.0166)

PDSI<-3*Income -0.000302 -3.48e-06 0.000875* 0.000906

(0.000465) (0.000237) (0.000479) (0.000713)

18
 PDSI -3 to-1*Income 0.000714** -0.000230 0.000188 0.000698**

(0.000351) (0.000145) (0.000246) (0.000341)

PDSI 1 to 3*Income 9.55e-05 -0.000124 0.000675** 9.26e-05

(0.000321) (0.000124) (0.000287) (0.000308)

PDSI >3*Income 0.00104*** 0.000260* 0.000169 -0.000179

(0.000381) (0.000134) (0.000305) (0.000385)

Annual average temperature 0.268*** 0.0773*** 0.234*** 0.270***

(0.0447) (0.0210) (0.0303) (0.0607)

Annual average temperature

squared -0.00894*** -0.00292*** -0.00643*** -0.00524***

(0.00123) (0.000599) (0.000856) (0.00146)

Precipitation 0.000667*** 0.000549*** 0.000177* 0.00120***

(0.000129) (4.64e-05) (0.000103) (0.000183)

Precipitation squared -3.57e-07*** -1.30e-07*** -2.30e-07*** -5.76e-07***

(6.80e-08) (1.62e-08) (4.74e-08) (9.44e-08)

Constant 11.11*** 9.892*** 11.41*** 9.053***

(0.432) (0.193) (0.296) (0.711)

Observations 14,488 16,671 18,697 12,908

R-squared 0.334 0.125 0.177 0.224

Number of fips 618 773 808 544

Robust standard errors in parentheses. Income measured in $1,000s. We are using 2010 5-year
median household income estimates.

*** p<0.01, ** p<0.05, * p<0.1

19
Figure 4: Proportional Drought Impacts on Total Production for Corn and Hay: Growing Season v. Non-Growing Season

20
Figure 5: Proportional Drought Impacts on Total Production for Corn and Hay: Groundwater dependent versus non-groundwater
dependent.

21
Economy-Wide impacts
Next, we turn to evaluating the impact of drought on other economic sectors. Importantly, the
impacts presented below reflect the total effect of the drought on each sector. This includes the
direct, as well as those associated with shocks to both backward and forward industries (indirect and
induced). Figure 6 presents drought impacts on wages across all sectors in our study. The coefficient
for moderate drought conditions was statistically significant at the 10% level. The results suggest that
total wages increase during periods of moderate drought. Specifically, each additional month of
moderate drought conditions increases total wages by approximately 0.2 percent. Alternatively, each
month of extreme wet conditions reduces total wages by approximately 0.3 percent. We note
however, that the impact on total wages masks important nuances in that some sectors experience
increased wages while other sectors experience decreased wages. We explore these nuances below.

The impact of drought on total employment across all sectors is presented in Figure 6. We find that an
additional month of moderate drought increases total employment by approximately 0.2 percent. On
the other hand, extreme wet conditions do not statistically significantly impact total employment.
Higher temperatures have a positive (and diminishing) effect on total employment, whereas more
precipitation has a negative effect on total employment.

The impact of drought on the number of establishments present across all sectors is presented in the
right-hand side panel of Figure 7. Again, moderate drought is positive and statistically significantly
correlated with establishment counts. Alternatively, an additional month of extreme wet conditions is
negatively correlated with establishments. Higher temperatures have a positive (and diminishing)
effect on the overall number of establishments, whereas more precipitation has a negative effect on
the overall number of establishments. Full results are presented in Tables 3-5 below.

Figure 6: Proportional Drought Impacts on Total Wages across all sectors of the economy
measured in the NAICS codes.

22
Figure 7: Proportional Drought Impacts on Total Employment and Number of Establishments
across all sectors of the economy measured in the NAICS codes.

Sector-Level Analysis
The results in the previous section present the impacts of drought on wages, number of
establishments, and employment across all economic sectors. However, this aggregate analysis
masks some important nuances. In the following section, we evaluate drought impacts on wages,
number of establishments, and employment for specific sectors of the economy. We find that drought
has negative impacts on some sectors and positive impacts on other sectors.

Total Wages

First, we look at total wages for specific sectors. The results are presented in Table 3. Column 1
presents the aggregate impacts of drought across all economic sectors. The first four coefficients, the
coefficients on the PDSI bins, are the same coefficients that are presented in Figure 6. The table also
displays coefficient estimates on temperature, temperature squared, precipitation and precipitation
squared. Columns 2-9 estimate the model for specific sectors. The results indicate that an additional
month of extreme drought decreases agricultural, fishing, and hunting wages by 0.5 percent and
recreation and entertainment wages by 0.6. In contrast, an additional month of extreme drought
increases utility wages by 0.7 percent. The sum of these coefficients explains why aggregate wages
(Column 1) are not statistically significantly affected by extreme drought conditions.

Several possible explanations exist for these findings. While it is reasonable that drought would
negatively impact agricultural, fishing, and hunting wages and recreation and entertainment wages, it

23
is somewhat surprising that drought would positively impact utility wages. One possible explanation
is that drought conditions cause business owners to use more municipal water, which would result in
greater demand for the services utilities provide.

Next, Table 4 presents the impacts of drought on employment across several economic sectors. The
results are qualitatively similar to the impacts on wages and substantiate the hypothesis that drought
reduces employment in agriculture, fishing, and hunting and recreation and entertainment, which
reduces wages. The finding that wages increases in some sectors during periods of drought is
consistent with Branco and Feres (2020) who find that drought increases employment in non‐
agricultural sectors – albeit in a developing country context.

Table 5 presents the impacts of drought on the number of establishments across several economic
sectors. The impact of drought on establishment numbers is slightly different from the impact of
drought on wages and employment. We find that the number of recreation and entertainment,
service, and trade establishments are negatively correlated with extreme drought conditions, but the
standard errors are relatively large for service and trade. In contrast, moderate drought conditions are
positively correlated with the number of agricultural, fishing, and hunting, other, service, and trade
establishments, while extreme wet conditions are negatively correlated with recreational and
entertainment establishments, manufacturing establishments, service establishments, and trade
establishments.

24
Table 3: Drought Impact on Total Wages: Economy-Wide and by Major Economic Sector

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Ag, Fishing Recreation Education

All Hunting Entertainment Health Manufacturing Other Service Trade Utilities

PDSI <- 3 0.000532 -0.00542** -0.00587** -4.01e-05 0.00186 0.000413 0.000550 3.31e-05 0.00726***

(0.000987) (0.00262) (0.00293) (0.00275) (0.00258) (0.00259) (0.00198) (0.00134) (0.00228)

PDSI between -3 0.00193** 0.00195 -0.00116 0.00258 0.00245 0.0127*** 0.00286* 0.00165* 0.00476**
and -1

(0.000766) (0.00181) (0.00204) (0.00215) (0.00170) (0.00232) (0.00149) (0.000882) (0.00194)

PDSI between 1 0.00118 0.000340 -0.00229 0.00236 0.00439** 0.00846*** -0.00183 -0.000218 -0.000316
and 3

(0.000822) (0.00190) (0.00248) (0.00219) (0.00182) (0.00283) (0.00147) (0.00101) (0.00199)

PDSI > 3 -0.00289*** -0.00147 -0.00490* 0.00326 -0.000846 -0.00976** -0.00697*** -0.00434*** -0.00539**

(0.00101) (0.00220) (0.00295) (0.00262) (0.00226) (0.00413) (0.00175) (0.00125) (0.00233)

Annual average 0.0496*** 0.103*** 0.0494 0.00684 0.000779 -0.0398 0.0336 0.0384*** 0.0597**
temperature

(0.0135) (0.0275) (0.0366) (0.0289) (0.0253) (0.0271) (0.0207) (0.0138) (0.0274)

Annual average -0.00135*** -0.00231*** -0.00128 0.000200 0.000221 0.00115* -0.00150*** -0.000983*** -0.00128*
temperature
squared (0.000345) (0.000696) (0.000908) (0.000741) (0.000659) (0.000681) (0.000540) (0.000364) (0.000688)

25
 Precipitation -5.47e-05*** 0.000104* -6.50e-05 -2.70e-05 -7.19e-06 -0.000133 -6.67e-05* -4.65e-05** -1.75e-05

(1.59e-05) (5.33e-05) (6.43e-05) (5.00e-05) (4.04e-05) (8.32e-05) (3.62e-05) (2.07e-05) (4.77e-05)

Precipitation 1.38e-08*** -5.06e-08*** 3.34e-08 1.07e-08 1.01e-08 5.15e-08 2.39e-08** 1.31e-08** -5.69e-09
squared

(4.94e-09) (1.76e-08) (2.17e-08) (1.39e-08) (1.15e-08) (3.51e-08) (1.10e-08) (6.37e-09) (1.58e-08)

Constant 2.912*** -1.672*** -2.031*** -1.566*** 1.632*** -3.334*** 1.611*** 1.837*** 0.949***

(0.133) (0.282) (0.377) (0.291) (0.253) (0.325) (0.207) (0.131) (0.281)

County fixed yes yes yes yes yes yes yes yes yes
effects

Year fixed effects yes yes yes yes yes yes yes yes yes

Observations 32,148 23,504 19,805 21,441 26,994 16,031 31,354 31,790 29,939

R-squared 0.762 0.476 0.478 0.854 0.209 0.891 0.281 0.392 0.361

Number of 871 860 718 840 836 871 871 870 869
Counties

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and 1. All dependent variables
are logged. PDSI variables are number of months with PDSI in indicated bin. *** p<0.01, ** p<0.05, * p<0.1

26
Table 4: Drought Impact on Employment: Economy-Wide and by Major Economic Sector

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Ag, Fishing Recreation Education

All Hunting Entertainment Health Manufacturi Other Service Trade Utilities

ng

PDSI <- 3 -0.000746 -0.00500* -0.00634** -0.00117 -0.00135 -7.95e-05 -0.000567 0.000280 0.00318

(0.000802) (0.00257) (0.00258) (0.00235) (0.00226) (0.00239) (0.00170) (0.000972) (0.00197)

PDSI between -3 and -1 0.00163*** 0.00160 -0.000915 0.00144 0.00196 0.0101*** 0.00257* 0.00151** 0.00382**

(0.000609) (0.00176) (0.00181) (0.00185) (0.00147) (0.00216) (0.00134) (0.000682) (0.00166)

PDSI between 1 and 3 -0.000112 0.000455 -0.00246 0.00219 0.00210 0.00382 -0.00177 -0.000813 -0.00178

(0.000667) (0.00184) (0.00217) (0.00191) (0.00156) (0.00265) (0.00130) (0.000802) (0.00171)

PDSI > 3 -0.00360*** 0.000813 -0.00385 0.00410* -0.00322* -0.00597 - -0.00363*** -

0.00691*** 0.00704***

(0.000802) (0.00216) (0.00249) (0.00221) (0.00194) (0.00365) (0.00149) (0.000955) (0.00200)

Annual average temperature 0.0543*** 0.0760*** 0.0525 -0.00387 0.0254 -0.0724*** 0.0280 0.0190* 0.0916***

(0.0108) (0.0262) (0.0325) (0.0244) (0.0216) (0.0257) (0.0181) (0.0108) (0.0237)

Annual average temperature squared -0.00152*** -0.00167** -0.00148* 0.000429 -0.000508 0.00185*** - -0.000559* -
0.00136*** 0.00219***

(0.000278) (0.000666) (0.000811) (0.000620) (0.000568) (0.000627) (0.000472) (0.000288) (0.000598)

27
 Precipitation -5.39e-05*** 5.54e-05 -5.08e-05 -6.25e-05 -1.98e-05 -0.000133* -6.24e-05* -5.01e-05*** -1.80e-05

(1.41e-05) (4.92e-05) (5.79e-05) (4.51e-05) (3.59e-05) (7.28e-05) (3.21e-05) (1.67e-05) (3.90e-05)

Precipitation squared 1.40e-08*** -2.88e-08* 2.57e-08 1.99e-08 1.18e-08 3.88e-08 2.09e-08** 1.51e-08*** -4.56e-09

(4.29e-09) (1.56e-08) (1.96e-08) (1.32e-08) (1.00e-08) (2.94e-08) (9.82e-09) (5.32e-09) (1.22e-08)

Constant 9.859*** 5.787*** 5.649*** 6.186*** 8.287*** 4.413*** 8.874*** 9.200*** 7.364***

(0.105) (0.267) (0.330) (0.248) (0.213) (0.312) (0.179) (0.102) (0.242)

County fixed effects yes yes yes yes yes yes yes yes yes

Year fixed effects yes yes yes yes yes yes yes yes yes

Observations 32,148 23,504 19,805 21,441 26,994 16,023 31,354 31,790 29,938

R-squared 0.275 0.171 0.232 0.768 0.038 0.868 0.369 0.317 0.097

Number of Counties 871 860 718 840 836 871 871 870 869

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and 1. All dependent variables
are logged. PDSI variables are number of months with PDSI in indicated bin. *** p<0.01, ** p<0.05, * p<0.1

28
Table 5: Drought Impact on Number of Establishments: Economy-Wide and by Major Economic Sector

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Ag, Fishing Recreation Education

All Hunting Entertainment Health Manufacturing Other Service Trade Utilities

PDSI <- 3 -0.00139* -0.000428 -0.00504*** -0.00152 -0.000859 -0.000452 -0.00214* -0.00170* -0.000333

(0.000748) (0.00177) (0.00165) (0.00182) (0.00134) (0.00185) (0.00112) (0.000866) (0.00130)

PDSI between -3 and -1 0.00190*** 0.00300** -0.000589 -0.00166 0.000510 0.00696*** 0.00196*** 0.00134** 0.00189*

(0.000513) (0.00131) (0.00107) (0.00135) (0.000982) (0.00171) (0.000738) (0.000594) (0.00104)

PDSI between 1 and 3 -0.000778 0.00249* -0.00224* -0.00225 0.000144 0.00670*** -0.000599 -0.000517 -0.00235**

(0.000561) (0.00146) (0.00133) (0.00137) (0.00102) (0.00185) (0.000739) (0.000660) (0.00108)

PDSI > 3 -0.00366*** 0.00587*** -0.00301** 0.00210 -0.00249* 0.00277 -0.00436*** -0.00293*** -0.00789***

(0.000698) (0.00166) (0.00152) (0.00143) (0.00129) (0.00311) (0.000946) (0.000761) (0.00132)

Annual average 0.0728*** 0.0587*** 0.0296* -0.0114 0.0109 -0.114*** 0.0690*** 0.0470*** 0.122***
temperature
(0.00840) (0.0180) (0.0162) (0.0159) (0.0158) (0.0213) (0.0111) (0.00887) (0.0145)

Annual average -0.00220*** -0.000753* -0.00112*** 0.000476 -0.000505 0.00273*** -0.00235*** -0.00139*** -0.00321***
temperature squared
(0.000219) (0.000435) (0.000413) (0.000405) (0.000406) (0.000525) (0.000284) (0.000234) (0.000361)

29
 Precipitation -7.74e-05*** 6.84e-05** -5.75e-05** - -0.000137*** - 2.74e-05 -4.81e-05*** -3.66e-05
0.000149*** 0.000210**
*

(1.21e-05) (3.32e-05) (2.67e-05) (3.47e-05) (2.45e-05) (5.77e-05) (2.04e-05) (1.27e-05) (2.67e-05)

Precipitation squared 2.02e-08*** -9.74e-09 1.12e-08 4.79e-08*** 4.59e-08*** 7.87e- -4.79e-09 1.57e-08*** 5.98e-09
08***

(3.81e-09) (9.28e-09) (7.36e-09) (1.21e-08) (7.67e-09) (2.35e-08) (5.66e-09) (3.53e-09) (9.41e-09)

Constant 6.526*** 2.573*** 2.784*** 3.092*** 4.278*** 3.447*** 5.421*** 5.863*** 4.091***

(0.0822) (0.196) (0.165) (0.164) (0.158) (0.257) (0.112) (0.0849) (0.153)

County fixed effects yes yes yes yes yes yes yes yes yes

Year fixed effects yes yes yes yes yes yes yes yes yes

Observations 32,189 27,039 22,685 25,072 28,783 16,279 31,787 32,061 30,932

R-squared 0.302 0.329 0.260 0.836 0.030 0.828 0.204 0.426 0.236

Number of Counties 871 869 834 867 863 871 871 871 871

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and 1. All dependent variables
are logged. PDSI variables are number of months with PDSI in indicated bin. *** p<0.01, ** p<0.05, * p<0.1

30
Sub-Sector Specific Analysis: Agriculture
In this subsection, we further decompose the impact of drought on wages, employment, and number
of establishments for specific agricultural sectors including livestock, crop production, and
agricultural supply. The results for wages, employment, and number of establishments are presented
in Tables 6-8 respectively. The results overwhelmingly indicate that extreme drought negatively
impacts the agricultural economy. An additional month of extreme drought decreases agricultural
supply wages by 1.3 percent. Correspondingly, an additional month of extreme drought decreases
livestock and agricultural supply employment by 0.6 and 1.2 percent, respectively. Finally, an
additional month of extreme drought decreases the number of agricultural supply establishments by
0.9 percent. However, livestock and crop production establishments are positively affected by
extreme wet conditions. Overall, drought appears to significantly impact most agricultural sectors,
even putting some out of business, with the effects on livestock production being especially important
for states such as Colorado. Unfortunately, we cannot further disaggregate these into small and large
businesses.

Table 6: Drought Impacts on Total Wages: Agricultural

(1) (2) (3)

Crop Agricultural
Livestock Production Supply

PDSI <- 3 -0.00306 -0.00270 -0.0135***

(0.00329) (0.00344) (0.00434)

PDSI between -3 and -1 -0.00283 0.00189 0.000635

(0.00218) (0.00259) (0.00313)

PDSI between 1 and 3 0.000495 0.000598 -0.00572*

(0.00234) (0.00288) (0.00331)

PDSI > 3 -0.00276 -0.00388 -0.00802**

(0.00299) (0.00330) (0.00367)

Annual average temperature 0.0670** 0.183*** 0.200***

(0.0279) (0.0382) (0.0444)

31
 Annual average temperature
squared -0.00137* -0.00391*** -0.00470***

(0.000700) (0.000990) (0.00110)

Precipitation -3.02e-05 6.05e-05 -6.78e-05

(6.16e-05) (7.04e-05) (0.000101)

Precipitation squared 3.25e-08 -6.10e-09 8.90e-09

(2.12e-08) (2.05e-08) (3.80e-08)

Constant -1.865*** -2.967*** -3.304***

(0.304) (0.379) (0.467)

County fixed effects yes yes yes

Year fixed effects yes yes yes

Observations 18,117 13,691 15,351

R-squared 0.395 0.416 0.298

Number of Counties 777 658 745

Standard errors in parentheses clustered at the county level. Impacts are relative to a
month with PDSI between -1 and 1. All dependent variables are logged. PDSI variables are
number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

32
Table 7: Drought Impacts on Employment: Agricultural

(1) (2) (3)

Crop Agricultural
Livestock Production Supply

PDSI <- 3 -0.00641** -0.00357 -0.0122***

(0.00315) (0.00334) (0.00418)

PDSI between -3 and -1 -0.00300 0.000916 7.15e-05

(0.00211) (0.00260) (0.00293)

PDSI between 1 and 3 0.000718 -0.000155 -0.00346

(0.00225) (0.00280) (0.00303)

PDSI > 3 -0.000857 -0.00211 -0.00695*

(0.00292) (0.00324) (0.00358)

Annual average temperature 0.0765*** 0.192*** 0.156***

(0.0272) (0.0372) (0.0417)

Annual average temperature squared -0.00158** -0.00402*** -0.00374***

(0.000680) (0.000958) (0.00103)

Precipitation -5.56e-05 0.000100 -9.83e-05

(6.04e-05) (7.15e-05) (9.10e-05)

Precipitation squared 3.88e-08* -1.41e-08 1.59e-08

(2.11e-08) (2.10e-08) (3.24e-08)

Constant 5.148*** 4.111*** 4.424***

(0.297) (0.372) (0.443)

County fixed effects yes yes yes

33
 Year fixed effects yes yes yes

Observations 18,117 13,691 15,351

R-squared 0.091 0.101 0.235

Number of Counties 777 658 745

Standard errors in parentheses clustered at the county level. Impacts are relative to a month
with PDSI between -1 and 1. All dependent variables are logged. PDSI variables are number of
months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

Table 8: Drought Impacts on Number of Establishments: Agricultural

(1) (2) (3)

Crop Agricultural
Livestock Production Supply

PDSI <- 3 0.000432 0.000119 -0.00931***

(0.00206) (0.00218) (0.00249)

PDSI between -3 and -1 0.00133 0.00277 -0.000889

(0.00152) (0.00182) (0.00178)

PDSI between 1 and 3 0.00372** 0.000837 -0.000992

(0.00169) (0.00205) (0.00200)

PDSI > 3 0.00587*** 0.00615*** -0.00265

(0.00210) (0.00236) (0.00232)

Annual average temperature 0.0173 0.132*** 0.106***

(0.0184) (0.0261) (0.0249)

34
 Annual average temperature
squared -3.81e-05 -0.00226*** -0.00247***

(0.000455) (0.000651) (0.000609)

Precipitation -1.07e-05 0.000122*** -9.99e-05**

(4.08e-05) (4.71e-05) (4.89e-05)

Precipitation squared 3.00e-08** -9.34e-10 3.50e-08**

(1.35e-08) (1.40e-08) (1.57e-08)

Constant 2.093*** 0.776*** 1.992***

(0.206) (0.268) (0.264)

County fixed effects yes yes yes

Year fixed effects yes yes yes

Observations 21,161 17,287 20,272

R-squared 0.308 0.289 0.331

Number of Counties 851 812 825

Standard errors in parentheses clustered at the county level. Impacts are relative to a
month with PDSI between -1 and 1. All dependent variables are logged. PDSI variables
are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

35
Sub-Sector Specific Analysis: Recreation,
Entertainment, and Travel
In this subsection, we continue to further disaggregate the impact of drought on wages, employment,
and number of establishments for specific sectors including recreation, entertainment and travel. The
results for wages, employment, and number of establishments are presented in Tables 9-11
respectively. In contrast to agricultural sectors, we find that drought conditions are statistically
uncorrelated with fishing wages, employment, and number of establishments. However, this is likely
due to the small number of fishing establishments we observe in our data.

On the other hand, all three measures for recreation and entertainment are negatively impacted by
extreme drought conditions, which could be driven by lower snowpack during the winter months, low
river levels during the summer, or increased fire severity during the summer. For example, an
additional month of extreme drought decreases recreation and entertainment wages by 0.59 percent.
Moderate drought is uncorrelated with recreational and entertainment wages, employment, and
establishments.

Finally, an additional month of extreme drought increases travel wages by 0.3 percent while an
additional month of moderate drought increases travel wages by 0.5 percent on average. Results are
similar for travel employment and establishments. Overall, drought appears to negatively impact the
recreation and entertainment sectors, which again is especially important for states such as Colorado
that have large tourism and outdoor industries.

36
Table 9: Drought Impacts on Total Wages: Recreation, Entertainment, and Travel

(1) (3) (3)

Recreation Travel

Fish Entertainment Accommodation

PDSI <- 3 0.00593 -0.00587** 0.00289*

(0.0255) (0.00293) (0.00159)

PDSI between -3 and -1 0.0130 -0.00116 0.00524***

(0.0124) (0.00204) (0.00121)

PDSI between 1 and 3 0.0123 -0.00229 0.000133

(0.0143) (0.00248) (0.00117)

PDSI > 3 -0.0144 -0.00490* -0.00339**

(0.0207) (0.00295) (0.00153)

Annual average temperature 0.249 0.0494 0.0380*

(0.269) (0.0366) (0.0199)

Annual average temperature

squared -0.00459 -0.00128 -0.00133***

(0.00589) (0.000908) (0.000499)

Precipitation 0.000172 -6.50e-05 2.14e-06

(0.000239) (6.43e-05) (3.10e-05)

Precipitation squared -3.40e-08 3.34e-08 8.41e-09

(4.28e-08) (2.17e-08) (9.38e-09)

Constant -3.377 -2.031*** -0.0111

(2.949) (0.377) (0.200)

County fixed effects yes yes yes

37
 Year fixed effects yes yes yes

Observations 951 19,805 27,737

R-squared 0.096 0.478 0.743

Number of Counties 62 718 863

Standard errors in parentheses clustered at the county level. Impacts are relative to a
month with PDSI between -1 and 1. All dependent variables are logged. PDSI variables are
number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

Table 10: Drought Impacts on Number Employed: Recreation, Entertainment, and Travel

(1) (2) (3)

Recreation Travel

Fish Entertainment Accommodation

PDSI <- 3 -0.00668 -0.00634** 0.00324**

(0.0216) (0.00258) (0.00139)

PDSI between -3 and -1 0.0182 -0.000915 0.00443***

(0.0127) (0.00181) (0.00102)

PDSI between 1 and 3 0.0117 -0.00246 -0.000125

(0.0142) (0.00217) (0.00103)

PDSI > 3 -0.00977 -0.00385 -0.00208

(0.0191) (0.00249) (0.00131)

Annual average temperature 0.139 0.0525 0.00464

(0.226) (0.0325) (0.0168)

38
 Annual average temperature
squared -0.00195 -0.00148* -0.000413

(0.00527) (0.000811) (0.000422)

Precipitation 0.000124 -5.08e-05 4.10e-06

(0.000190) (5.79e-05) (2.68e-05)

Precipitation squared -2.06e-08 2.57e-08 5.88e-09

(3.46e-08) (1.96e-08) (8.08e-09)

Constant 4.451* 5.649*** 8.003***

(2.308) (0.330) (0.169)

County fixed effects yes yes yes

Year fixed effects yes yes yes

Observations 951 19,805 27,737

R-squared 0.330 0.232 0.524

Number of Counties 62 718 863

Standard errors in parentheses clustered at the county level. Impacts are relative to a
month with PDSI between -1 and 1. All dependent variables are logged. PDSI variables are
number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

39
Table 11: Drought Impacts on Number of Establishments: Recreation, Entertainment, and Travel

(1) (2) (3)

Recreation Travel

Fish Entertainment Accommodation

PDSI <- 3 -8.90e-05 -0.00504*** 0.00316***

(0.00745) (0.00165) (0.000944)

PDSI between -3 and -1 0.0118*** -0.000589 0.00287***

(0.00454) (0.00107) (0.000669)

PDSI between 1 and 3 0.00978* -0.00224* 0.000267

(0.00511) (0.00133) (0.000746)

PDSI > 3 -0.00682 -0.00301** -0.000401

(0.00814) (0.00152) (0.000927)

Annual average temperature -0.00544 0.0296* -0.0205**

(0.0560) (0.0162) (0.00935)

Annual average temperature

squared 0.000592 -0.00112*** 0.000426*

(0.00119) (0.000413) (0.000235)

Precipitation 0.000146 -5.75e-05** 5.54e-05***

(9.27e-05) (2.67e-05) (1.90e-05)

Precipitation squared -3.72e-08* 1.12e-08 -2.14e-09

(2.06e-08) (7.36e-09) (5.83e-09)

Constant 3.010*** 2.784*** 4.764***

(0.603) (0.165) (0.0976)

40
 County fixed effects yes yes yes

Year fixed effects yes yes yes

Observations 2,320 22,685 28,046

R-squared 0.304 0.260 0.743

Number of Counties 232 834 871

Standard errors in parentheses clustered at the county level. Impacts are relative to a
month with PDSI between -1 and 1. All dependent variables are logged. PDSI variables are
number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

41
Health Impacts and Criminal Activity
In the following section we evaluate the impact of drought on public health and criminal activity. The
results are presented in Tables 12 and 13, respectively. While the mechanisms driving the relationship
between drought and agricultural production and the agricultural economy are clear, the
mechanisms driving the relationship between drought and public health and criminal activity are less
clear. Consequently, we find very little statistical relationship between drought and public health with
the exception that an additional month of extreme wet conditions decreases the annual county-level
incidence of heart disease by 0.5 percent. This could possibly be due to the decrease in agricultural
production associated with drought, which could lower agricultural worker stress.

Similar to the impact on health outcomes, Table 13 indicates that drought and wet conditions are
largely uncorrelated with annual criminal activity. We do find that an additional month of moderate
drought is associated with a 0.4 and 0.3 percent reduction in simple assaults and property crime,
respectively.

Table 12: Drought Impacts on Health Outcomes

(1) (2) (3)

Heart
Opioids Alcohol Disease

PDSI <- 3 0.00430 -0.000818 -0.00504***

(0.00553) (0.0111) (0.00118)

PDSI between -3 and -1 -0.000207 -0.0115 0.000206

(0.00422) (0.00863) (0.000995)

PDSI between 1 and 3 0.00196 -0.0153 -0.00216*

(0.00407) (0.0104) (0.00115)

PDSI > 3 -0.00334 -0.0163 -0.000810

(0.00583) (0.0124) (0.00152)

Annual average temperature -0.0814* -0.130 -0.0187

(0.0460) (0.102) (0.0119)

42
 Annual average temperature
squared 0.00128 0.000550 6.37e-05

(0.00109) (0.00233) (0.000276)

Precipitation 0.000113 -0.000253 -5.71e-05*

(8.76e-05) (0.000201) (3.04e-05)

Precipitation squared -4.58e-08* 8.20e-08 5.06e-09

(2.70e-08) (6.96e-08) (9.16e-09)

County fixed effects yes yes yes

Year fixed effects yes yes yes

Observations 13,266 9,918 15,448

R-squared 0.766

Number of Counties 735 550 868

Standard errors in parentheses clustered at the county level. Impacts are relative to a
month with PDSI between -1 and 1. All dependent variables are logged. PDSI variables
are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

43
Table 13: Drought Impacts on Criminal Activity

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

Aggravated Simple Property Violent

Assault Assault Crime Crime

PDSI <- 3 -0.00396 0.00474 0.00150 -0.00202

(0.00337) (0.00323) (0.00218) (0.00315)

PDSI between -3 and -1 -0.00379 -0.00434* -0.00292* -0.00229

(0.00245) (0.00233) (0.00163) (0.00232)

PDSI between 1 and 3 -0.000999 -0.000750 -0.00141 0.000385

(0.00295) (0.00269) (0.00183) (0.00270)

PDSI > 3 -0.000747 -0.00261 0.00231 0.00101

(0.00324) (0.00311) (0.00212) (0.00304)

Annual average temperature 0.0425 -0.0790*** -0.0254 0.0420

(0.0318) (0.0282) (0.0212) (0.0286)

Annual average temperature

squared -0.000996 0.00243*** 0.000813 -0.000873

(0.000762) (0.000678) (0.000496) (0.000678)

Precipitation -1.54e-05 3.69e-06 1.56e-05 1.55e-05

(5.95e-05) (5.40e-05) (3.53e-05) (5.50e-05)

Precipitation squared 1.06e-09 -2.08e-08 -1.04e-08 -6.53e-09

(1.51e-08) (1.51e-08) (1.02e-08) (1.38e-08)

Constant 0.646* 2.547*** 7.715*** 4.516***

(0.346) (0.307) (0.236) (0.317)

County fixed effects yes yes yes yes

44
 Year fixed effects yes yes yes yes

Observations 17,163 17,015 17,510 17,455

R-squared 0.019 0.168 0.124 0.017

Number of Counties 825 822 824 826

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with
PDSI between -1 and 1. All dependent variables are logged. PDSI variables are number of months
with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

45
Total Revenue, Total Taxes, and Insurance Claims and
Payments
This section presents the relationship between months of drought and total revenue and total taxes
(Tables 14 and 15). We find that an additional month of extreme drought increases total government
revenue by 0.4 percent while an additional month of moderate drought increases total taxes by 0.3
percent. Moderately wet and extreme wet conditions are also positively associated with total
government revenue and taxes.

While drought conditions reduce agricultural crop production, acres harvested, and yields, these
impacts are likely offset by crop insurance payments. Not surprisingly, extreme drought is associated
with increases in indemnity payments and policy counts by 10 and 7 percent, respectively (Table 15).
Moderate drought increases indemnity payments and policy counts by 6.5 and 4.7 percent,
respectively. These payments, which are not reflected in traditional analyses like IMPLAN, likely offset
the negative agricultural production impacts associated with drought and help to explain the positive
impact of drought on local government revenues. Interestingly, indemnity payments are also
positively associated with extreme wet conditions, which may reflect the negative relationship
between some crop yields and other events that are correlated with extreme wet conditions.

Table 14: Drought Impacts on Total Revenue and Total Taxes

(1) (2)

Total Total
Revenue Taxes

PDSI<-3 0.00416* 0.00112

(0.00234) (0.00261)

PDSI -3 to-1 2.42e-05 0.00344**

(0.00179) (0.00161)

PDSI 1 to 3 0.00312* 0.00171

(0.00183) (0.00171)

PDSI >3 0.00357** 0.00299*

(0.00176) (0.00169)

46
 Annual average temperature -0.0325 -0.0779***

(0.0208) (0.0243)

Annual average temperature

squared 0.000675 0.00172**

(0.000581) (0.000665)

Precipitation -1.93e-05 -6.18e-05

(3.67e-05) (4.21e-05)

Precipitation squared -2.94e-09 1.47e-08

(9.34e-09) (1.18e-08)

Constant 8.994*** 8.366***

(0.199) (0.230)

Observations 8,189 8,186

R-squared 0.784 0.814

Number of fips 286 286

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

47
Table 15: Drought Impacts on Insurance Liability and Indemnity Amounts and Counts of
Indemnity Policies

(1) (2) (3)

Liability Indemnity
Amount Amount Policy Count

PDSI<-3 0.000668 0.101*** 0.0705***

(0.00366) (0.00681) (0.00446)

PDSI -3 to-1 0.00265 0.0654*** 0.0466***

(0.00238) (0.00455) (0.00308)

PDSI 1 to 3 0.000905 0.00204 -0.00232

(0.00282) (0.00476) (0.00325)

PDSI >3 -0.00743** 0.0127** 0.00698

(0.00297) (0.00631) (0.00424)

Annual average temperature -0.0440 -0.309*** -0.0643

(0.0404) (0.0717) (0.0456)

Annual average temperature

squared 0.00156 0.0114*** 0.00487***

(0.00101) (0.00193) (0.00123)

Precipitation 0.000657*** 0.000303 0.000188

(0.000190) (0.000290) (0.000193)

Precipitation squared -1.74e-07* 4.12e-07*** 2.99e-07***

(9.72e-08) (1.51e-07) (1.01e-07)

Constant 14.16*** 12.79*** 2.557***

(0.433) (0.750) (0.496)

Observations 14,701 13,984 13,984

48
 R-squared 0.488 0.318 0.250

Number of fips 768 751 751

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

49
Simulations
The estimates presented thus far correspond separate out the effects of marginal changes in
precipitation and temperature, as well as the average impact of an additional month of drought
conditions across a span of approximately 40 years and across the entire western US. In this section,
we apply the methodology outlined in the Methodological Approach and Data Section to simulate the
impact of specific drought years on particular counties. The impacts presented reflect the combined
effect of prolonged drought conditions, as well as contemporaneous weather conditions. We present, as
an example, the impacts for Otero County, Colorado associated with conditions experienced in 2012
and 2018. We chose Otero County as a representative agricultural county in Colorado, but the analysis
could be applied to any county in our sample. Figures 8-10 display the inputs to the simulation model
including the number of months in different drought PDSI categories during the growing and non-
growing season as well as annual temperatures and precipitation for Otero County. For comparison,
Figures 11-14 display the distribution of the moderate and extreme drought conditions across the
western US for 2012 and 2018. We present the drought maps (Figures 11-14) simply as a reference to
the broader drought conditions faced by the entire west during each year.

The results are presented in Tables 16-19. The first column of each table displays the output variable
we are simulating. The second column of each table displays the estimated impact. The third column
displays the t-statistic for the simulated impact. The fourth column indicates whether the estimate is
significant at the 95% confidence level. The final two columns provide the lower and upper 95%
confidence intervals. For example, our simulated results indicate the 2012 drought, precipitation, and
temperature conditions reduced corn production in Otero County, CO by approximately 63 percent
with a 95 percent confidence interval of 53-73 percent. On the other hand, Table 15 shows that the
simulated impact of the 2018 drought did not statistically significantly impact corn production in
Otero County. This is perhaps unsurprising as Figure 8 shows that growing season conditions were
much less extreme in 2018 for Otero County.

While a reduction of 63 percent in corn production may seem large, this effect is in line with what
producers self-reported during this period. For example, survey responses from dryland corn
producers in the Arkansas Valley of Colorado reported an 87 percent reduction in yield (relative to
expected) and a 68 percent reduction in harvested acres (relative to expected) (Goemans et al. (2013)).
Hay, wheat, and sorghum were also dramatically impacted but not as much as corn production.
Again, we also find that a sizable fraction of the production declines can be attributed to reduced
acres harvested, indicating that farmers are responding on both the intensive and extensive margins.

The simulated corn production response to drought could be caused by several factors. First, corn
crops may be more sensitive to changes in water supply than other crops. Second, farms that produce
corn might have water rights that differ from farmers that produce other crops on average. While
there are other possible factors, our analysis does not allow us to disentangle the mechanisms behind
the declines in production.

50
We should also highlight several caveats to these projections. First, the models do not account for
time varying factors other than drought, precipitation, and temperature. In particular, we cannot
account for farm specific practices such as changes in irrigation strategies or crop choices. Second,
the projections do not account for market specific shocks, such as a pandemic or recession. Finally,
the projections cannot account for general equilibrium responses by farmers to past drought
conditions, or to changes in the demand or competition in certain markets. In summary, these results
should be considered the predicted impact of drought conditions, temperature, and precipitation in
Otero County relative to average conditions in that county.

The tables also show the simulated impact of each drought year on the other outcome variables
analyzed in the previous sections. For instance, the 2012 drought was associated with a reduction in
overall employment, number of establishments, and wages, with results varying by sector. Perhaps
most interesting is the 2012 drought is negatively associated with alcohol, opioid, and heart disease
death rates in Otero County, but positively associated with simple assault and property crime rates.
Drought can clearly be a stressor for farmers and workers whose income depends on the presence of
water, which could cause increased aggression in the form of simple assaults.

An important finding is that the 2012 drought did not statistically impact employment in livestock
sector (Table 18). Again, this could be caused by several factors. First, the timing estimates from the
regression analysis show a positive relationship between employment in the livestock sector and
temperature. Second, and more importantly, the results presented herein correspond to the short-
term impacts associated with prolonged drought conditions. Drought impacts on the livestock sector
are dynamic in nature, with substantial negative impacts unfolding over multiple years.

51
Figure 8: Number of months Otero County, CO experienced different drought conditions during
the growing season from 2010-2020.

Figure 9: Number of months Otero County, CO experienced different drought conditions

during the non-growing season from 2010-2020.

52
Figure 10: Average annual precipitation and temperature for Otero County, CO from
2010-2020.

53
 # of Months
10 12
8 10
6 8
4 6
2 4
0 2

Figure 11: Number of months of Moderate Drought (-3<PDSI<-1) by county in 2012.

# of Months
10 12
8 10
6 8
4 6
2 4
0 2

Figure 12: Number of months of Moderate Drought (-3<PDSI<-1) by county in 2018

54
 # of Months
10 12
8 10
6 8
4 6
2 4
0 2

Figure 13: Number of months of Extreme Drought (PDSI<-3) by county in 2012

# of Months
10 12
8 10
6 8
4 6
2 4
0 2

Figure 14: Number of months of Extreme Drought (PDSI<-3) by county in 2018

55
Table 16: Simulation of Proportional Impact of 2012 Drought on Otero County, Colorado

Standard Lower Upper

Impact error t Significant 95% CI 95% CI

Agricultural Outcomes

Corn Production -0.63193 0.052504 -12.0359 Yes -0.73484 -0.52902

Hay Production -0.37348 0.025978 -14.3768 Yes -0.42439 -0.32256

Wheat Production -0.38478 0.037876 -10.1589 Yes -0.45902 -0.31054

Sorghum Production -0.269 0.058801 -4.57472 Yes -0.38425 -0.15375

Corn Acres Harvested -0.23426 0.043438 -5.39294 Yes -0.3194 -0.14912

Hay Acres Harvested -0.06466 0.018946 -3.41279 Yes -0.10179 -0.02752

Wheat Acres Harvested -0.24198 0.032134 -7.53021 Yes -0.30496 -0.179

Sorghum Acres Harvested 0.228137 0.054662 4.173573 Yes 0.120999 0.335275

Employment

Overall -0.02798 0.008568 -3.26516 Yes -0.04477 -0.01118

Ag Fish and Hunt -0.03303 0.024741 -1.33494 No -0.08152 0.015465

Education and Health 0.039157 0.023259 1.683551 No -0.00643 0.084744

Manufacturing -0.00021 0.022149 -0.00932 No -0.04362 0.043206

Recreation and Entertainment -0.05728 0.026004 -2.20287 Yes -0.10825 -0.00632

Services -0.06645 0.016976 -3.91462 Yes -0.09973 -0.03318

Retail and Wholesale 0.001723 0.010343 0.166596 No -0.01855 0.021995

Utilities and Construction 0.025076 0.020081 1.248728 No -0.01428 0.064434

Establishments

Overall -0.05695 0.007636 -7.45752 Yes -0.07192 -0.04198

Ag Fish and Hunt 0.050945 0.015417 3.304443 Yes 0.020727 0.081162

56
Education and Health 0.04069 0.016088 2.529183 Yes 0.009157 0.072222

Manufacturing -0.0148 0.013811 -1.0714 No -0.04187 0.012273

Recreation and Entertainment -0.06934 0.015125 -4.58442 Yes -0.09899 -0.0397

Services -0.10747 0.011125 -9.66069 Yes -0.12928 -0.08567

Retail and Wholesale -0.03756 0.008893 -4.22312 Yes -0.05499 -0.02013

Utilities and Construction -0.03668 0.012197 -3.00699 Yes -0.06058 -0.01277

Wage Bill

Overall -0.0157 0.010562 -1.48649 No -0.0364 0.005001

Ag Fish and Hunt -0.04395 0.025816 -1.70223 No -0.09455 0.006655

Education and Health 0.043677 0.027368 1.595913 No -0.00996 0.097318

Manufacturing 0.031593 0.025321 1.247685 No -0.01804 0.081222

Recreation and Entertainment -0.03405 0.029553 -1.15199 No -0.09197 0.023879

Services -0.05908 0.019872 -2.97275 Yes -0.09802 -0.02013

Retail and Wholesale -0.00026 0.014127 -0.01847 No -0.02795 0.027427

Utilities and Construction 0.071924 0.023176 3.103446 Yes 0.0265 0.117348

Crime

Aggravated Assault -0.03572 0.027321 -1.30727 No -0.08927 0.017833

Simple Assault 0.111016 0.026035 4.264132 Yes 0.059988 0.162045

Property Crime 0.035462 0.017703 2.003138 Yes 0.000764 0.070161

Violent Crime -0.01596 0.024794 -0.64371 No -0.06456 0.032637

Health Outcomes

Opioids -0.06968 0.03265 -2.13415 Yes -0.13368 -0.00569

Alcohol -0.22354 0.08995 -2.48515 Yes -0.39984 -0.04724

Heart Disease -0.05507 0.009795 -5.62257 Yes -0.07427 -0.03588

57
 Insurance

Liability Amount -0.01532 0.042605 -0.35958 No -0.09883 0.068187

Indemnity Amount 1.156458 0.084906 13.62046 Yes 0.990043 1.322874

# Indemnity Policies 0.869058 0.056017 15.51426 Yes 0.759265 0.97885

Taxes and Revenue

Total Revenue 0.004998 0.023048 0.216862 No -0.04018 0.050171

Total Taxes 0.009473 0.025892 0.365845 No -0.04128 0.060221

Table 17: Simulation of Proportional Impact of 2018 Drought on Otero County, Colorado

Lower
Standard 95% Upper
Impact error t Significant CI 95% CI

Agricultural Outcomes

Corn Production -0.03613 0.020376 -1.77325 No -0.07607 0.003805

Hay Production -0.08257 0.007544 -10.9457 Yes -0.09736 -0.06779

Wheat Production -0.14652 0.015873 -9.23085 Yes -0.17764 -0.11541

Sorghum Production -0.07654 0.023931 -3.19818 Yes -0.12344 -0.02963

Corn Acres Harvested 0.057248 0.01852 3.091216 Yes 0.02095 0.093546

Hay Acres Harvested -0.04107 0.006044 -6.79522 Yes -0.05292 -0.02922

Wheat Acres Harvested -0.06885 0.013454 -5.11736 Yes -0.09522 -0.04248

Sorghum Acres Harvested -0.00822 0.022373 -0.36748 No -0.05207 0.035629

Employment

Overall -0.00949 0.003441 -2.75874 Yes -0.01624 -0.00275

Ag Fish and Hunt 0.016167 0.010859 1.488846 No -0.00512 0.037451

58
Education and Health 0.022629 0.010336 2.189449 Yes 0.002372 0.042887

Manufacturing 0.014177 0.009387 1.510326 No -0.00422 0.032576

Recreation and
Entertainment -0.02529 0.012311 -2.05397 Yes -0.04941 -0.00116

Services -0.03015 0.007623 -3.95476 Yes -0.04509 -0.01521

Retail and Wholesale -0.01243 0.004203 -2.95694 Yes -0.02067 -0.00419

Utilities and Construction -0.02844 0.009379 -3.0321 Yes -0.04682 -0.01005

Establishments

Overall -0.01894 0.003132 -6.04681 Yes -0.02508 -0.0128

Ag Fish and Hunt 0.012334 0.007684 1.605149 No -0.00273 0.027394

Education and Health 0.03639 0.007432 4.896409 Yes 0.021824 0.050957

Manufacturing -0.00791 0.005833 -1.3556 No -0.01934 0.003526

Recreation and
Entertainment -0.01675 0.007434 -2.25261 Yes -0.03132 -0.00218

Services -0.02573 0.004998 -5.14832 Yes -0.03553 -0.01593

Retail and Wholesale -0.0124 0.003283 -3.77624 Yes -0.01883 -0.00596

Utilities and Construction -0.0305 0.006345 -4.80731 Yes -0.04294 -0.01806

Wage Bill

Overall -0.00744 0.004252 -1.74929 No -0.01577 0.000896

Ag Fish and Hunt 0.006792 0.01157 0.58701 No -0.01589 0.02947

Education and Health 0.018771 0.011993 1.565099 No -0.00474 0.042278

Manufacturing 0.016897 0.010884 1.55244 No -0.00444 0.038229

Recreation and
Entertainment -0.03343 0.014468 -2.31102 Yes -0.06179 -0.00508

Services -0.03386 0.008853 -3.82462 Yes -0.05121 -0.01651

59
Retail and Wholesale -0.01391 0.005552 -2.50501 Yes -0.02479 -0.00303

Utilities and Construction -0.03281 0.010911 -3.00667 Yes -0.05419 -0.01142

Crime

Aggravated Assault 0.008544 0.015301 0.558396 No -0.02145 0.038533

Simple Assault -0.01818 0.014104 -1.28937 No -0.04583 0.009458

Property Crime -0.00663 0.010144 -0.65341 No -0.02651 0.013254

Violent Crime 0.00741 0.013685 0.541513 No -0.01941 0.034232

Health Outcomes

Opioids 0.035233 0.019724 1.786297 No -0.00343 0.073892

Alcohol -0.04484 0.06087 -0.73667 No -0.16415 0.074464

Heart Disease -0.0195 0.007619 -2.55869 Yes -0.03443 -0.00456

Insurance

Liability Amount -0.04021 0.018384 -2.18709 Yes -0.07624 -0.00417

Indemnity Amount 0.24725 0.037257 6.636312 Yes 0.174226 0.320274

# Indemnity Policies 0.172356 0.023863 7.222675 Yes 0.125584 0.219127

Taxes and Revenue

Total Revenue 0.010326 0.011174 0.924095 No -0.01158 0.032227

Total Taxes 0.015817 0.00961 1.645963 No -0.00302 0.034653

60
Table 18: Proportional Impact of 2012 Drought on Employment and Establishments by
Agricultural Sector in Otero County, CO

Standard Lower 95% Upper 95%

Impact error t Significant CI CI

Employment

Livestock -0.02846 0.029502 -0.96454 No -0.08628 0.029368

Crop Production 0.001777 0.032634 0.054466 No -0.06219 0.06574

Agricultural Suppliers -0.07637 0.037727 -2.02438 Yes -0.15032 -0.00243

Establishments

Livestock 0.032409 0.018519 1.750085 No -0.00389 0.068706

Crop Production 0.065942 0.019796 3.330992 Yes 0.027141 0.104743

Agricultural Suppliers -0.05275 0.022261 -2.36967 Yes -0.09638 -0.00912

Wage Bill

Livestock -0.00886 0.030822 -0.28746 No -0.06927 0.05155

Crop Production 0.004865 0.033586 0.144861 No -0.06096 0.070694

Agricultural Suppliers -0.07931 0.040764 -1.94571 No -0.15921 0.000583

61
Table 19: Proportional Impact of 2018 Drought on Employment and Establishments by
Agricultural Sector in Otero County, CO

Standard Lower 95% Upper 95%

Impact error t Significant CI CI

Employment

Livestock 0.02267 0.013455 1.684844 No -0.0037 0.049042

Crop Production -0.00525 0.015723 -0.33416 No -0.03607 0.025562

Agricultural Suppliers 0.004429 0.016261 0.272355 No -0.02744 0.0363

Establishments

Livestock 0.017662 0.009574 1.844785 No -0.0011 0.036426

Crop Production 0.022956 0.010708 2.143942 Yes 0.00197 0.043943

Agricultural Suppliers 0.007662 0.010786 0.710342 No -0.01348 0.028802

Wage Bill

Livestock 0.011889 0.013935 0.853145 No -0.01542 0.039202

Crop Production -0.0217 0.016128 -1.34519 No -0.05331 0.009916

Agricultural Suppliers -0.00694 0.017203 -0.40347 No -0.04066 0.026777

62
Conclusions
This analysis evaluates the impact of drought on many sectors over an approximately forty-year
period across the Western United States. Importantly, our approach distinguishes between the effect
of prolonged drought conditions (as measured by PDSI) and the impact of contemporaneous variation
in temperature and precipitation. Thus, simulated impacts reflect the combined effect associated
with both prolonged drought conditions and short-term weather conditions within a given year.

Our primary results indicate that prolonged drought has a negative and statistically significant impact
on corn, hay, and wheat production. Each additional month of extreme drought conditions (PDSI<-3)
is associated with a decrease in total corn and wheat production of 3.2 and 3.7 percent, respectively.
We also find that higher temperatures and precipitation have a positive and statistically significant
impact on total agricultural production; however, these effects are diminishing with higher
temperatures and become negative for conditions well above the mean.

The average analysis masks important heterogeneity. We find that drought conditions that occur
during the growing season have a much larger negative impact on total production than those that
occur during the non-growing season. For corn and wheat, the reduction in total production is largely
driven by a decrease in harvested acres, whereas for hay the reduction in total production is primarily
due to a decrease in yields per acre harvested. We also find evidence that the impact of drought on the
agricultural economy is lower in counties with higher median incomes

Beyond agricultural production, we find that each additional month of extreme drought is associated
with a reduction in total wages paid to employees in agriculture of approximately 0.5 percent, with
total wages decreasing by roughly the same amount for those working in the agricultural supply
sector. Extreme drought is also negatively associated with wages paid to employees in the recreation
and entertainment sectors, but positively associated with total wages paid to those in the utility and
construction sectors. Results are similar for total employment and number of establishments in each
sector with the exception that livestock employment is negatively associated with extreme drought.

We also evaluate the impact of drought on public health and criminal activity. We find the average
effect of an additional month of extreme drought is not statistically significant for opioid and alcohol
related mortality, but we do find a negative association between extreme drought and heart disease
mortality. We find suggestive evidence that moderate drought is associated with a decrease in simple
assaults and property crime.

Finally, we explored the relationships between drought and insurance markets and total government
revenue and tax. Unsurprisingly, indemnity amounts and insurance policy counts are positively
correlated with extreme and moderate drought. We also find suggestive evidence that drought
conditions are positively associated with total taxes and revenue.

63
The novelty of our study is that we use our overall results, combined with data on drought conditions
in particular counties in particular years to simulate location and sector specific drought impacts. As
an example, we simulate the impact of 2012 and 2018 drought conditions across a variety of economic
sectors, public health, and criminal activity for Otero County, Colorado. This analysis can be
performed for any county in our sample.

References
Abatzoglou, J.T., D.J. McEvoy, and K.T. Redmond, in press, “The West Wide Drought Tracker: Drought
Monitoring at Fine Spatial Scales,” Bulletin of the American Meteorological Society

Branco, Danyelle, and José Féres. "Weather Shocks and Labor Allocation: Evidence from Rural
Brazil." American Journal of Agricultural Economics.

Dai, Aiguo & National Center for Atmospheric Research Staff (Eds). Last modified 12 Dec 2019. "The
Climate Data Guide: Palmer Drought Severity Index (PDSI)." Retrieved from
https://climatedataguide.ucar.edu/climate-data/palmer-drought-severity-index-pdsi.

64
Appendix 1: Regression Tables for Agricultural Impacts

Table A1: Drought Impacts on Crop Production

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

Corn Hay Wheat Sorghum

PDSI <- 3 -0.0323*** -0.0222*** -0.0364*** -0.0273***

(0.00402) (0.00194) (0.00370) (0.00589)

PDSI between -3 and -1 -0.00847*** -0.00793*** -0.0191*** -0.0168***

(0.00252) (0.00115) (0.00203) (0.00314)

PDSI between 1 and 3 -0.0106*** -0.000539 -0.00302 0.000438

(0.00238) (0.00107) (0.00224) (0.00288)

PDSI > 3 -0.0179*** -0.00480*** -0.00985*** 0.0151***

(0.00334) (0.00137) (0.00246) (0.00366)

Annual average temperature 0.268*** 0.0710*** 0.234*** 0.274***

(0.0448) (0.0210) (0.0303) (0.0606)

Annual average temperature -0.00894*** -0.00273*** -0.00641*** -0.00536***

squared (0.00124) (0.000596) (0.000857) (0.00145)

Precipitation 0.000679*** 0.000551*** 0.000175* 0.00123***

(0.000127) (4.62e-05) (0.000103) (0.000183)

Precipitation squared -3.62e-07*** -1.31e-07*** -2.29e-07*** -5.93e-07***

(6.73e-08) (1.61e-08) (4.76e-08) (9.54e-08)

Constant 11.10*** 9.941*** 11.41*** 9.015***

(0.432) (0.195) (0.297) (0.711)

A-1
County fixed effects yes yes yes yes

Year fixed effects yes yes yes yes

Observations 14,488 16,671 18,697 12,908

R-squared 0.333 0.124 0.176 0.223

Number of fips 618 773 808 544

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between
-1 and 1. All dependent variables are logged. PDSI variables are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

A-2
Table A2: Drought Impacts on Yield Per Acre

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

Corn Hay Wheat Sorghum

PDSI <- 3 -0.0157*** -0.0187*** -0.0164*** -0.0250***

(0.00154) (0.00111) (0.00166) (0.00202)

PDSI between -3 -0.000835 -0.00512*** -0.00832*** -0.00723***

and -1 (0.000906) (0.000653) (0.000975) (0.000991)

PDSI between 1 -0.00430*** -5.10e-05 0.00115 -0.00369***

and 3 (0.000821) (0.000603) (0.000786) (0.000890)

PDSI > 3 -0.00978*** -0.00375*** -0.00598*** -0.00897***

(0.00101) (0.000700) (0.000911) (0.00107)

Annual average 0.140*** 0.112*** 0.0551*** 0.152***

temperature (0.0184) (0.0123) (0.0132) (0.0225)

Temperature -0.00559*** -0.00355*** -0.00170*** -0.00536***

squared (0.000511) (0.000367) (0.000362) (0.000574)

Precipitation 0.000173*** 0.000371*** -5.43e-05 0.000778***

(5.67e-05) (2.47e-05) (5.86e-05) (8.22e-05)

Precipitation -4.71e-08* -8.39e-08*** -1.00e-07*** -3.69e-07***

squared (2.84e-08) (6.54e-09) (3.08e-08) (4.44e-08)

Constant 3.779*** -0.172 3.198*** 2.677***

(0.177) (0.107) (0.129) (0.232)

County fixed
effects yes yes yes yes

A-3
Year fixed effects yes yes yes yes

Observations 14,488 16,671 18,697 12,908

R-squared 0.319 0.191 0.161 0.313

Number of fips 618 773 808 544

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and
1. All dependent variables are logged. PDSI variables are number of months with PDSI in indicated bin.

*** p<0.01, **
p<0.05, * p<0.1

A-4
Table A3: Drought Impacts on Harvested Acres

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

Corn Hay Wheat Sorghum

PDSI <- 3 -0.0166*** -0.00347** -0.0201*** -0.00230

(0.00367) (0.00165) (0.00294) (0.00517)

PDSI between -0.00764*** -0.00281*** -0.0108*** -0.00956***

-3 and -1 (0.00244) (0.000917) (0.00171) (0.00303)

PDSI between -0.00633*** -0.000488 -0.00418** 0.00412

1 and 3 (0.00232) (0.000881) (0.00198) (0.00267)

PDSI > 3 -0.00812** -0.00106 -0.00387* 0.0241***

(0.00317) (0.00109) (0.00213) (0.00339)

Annual average 0.128*** -0.0412*** 0.179*** 0.122**

temperature (0.0408) (0.0151) (0.0243) (0.0580)

Temperature -0.00335*** 0.000818* -0.00471*** -5.92e-06

squared (0.00108) (0.000421) (0.000701) (0.00141)

Precipitation 0.000507*** 0.000180*** 0.000229*** 0.000448***

(0.000120) (3.63e-05) (8.17e-05) (0.000155)

Precipitation -3.15e-07*** -4.70e-08*** -1.29e-07*** -2.24e-07***

squared (6.12e-08) (1.27e-08) (3.61e-08) (7.83e-08)

Constant 7.317*** 10.11*** 8.216*** 6.338***

(0.424) (0.144) (0.242) (0.672)

-0.0166*** -0.00347** -0.0201*** -0.00230

County fixed effects yes yes yes yes

Year fixed effects yes yes yes yes

A-5
Observations 14,488 16,671 18,697 12,908

R-squared 0.219 0.045 0.215 0.228

Number of fips 618 773 808 544

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1
and 1. All dependent variables are logged. PDSI variables are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, *

p<0.1

A-6
Table A4: Drought Impacts on Crop Production-Groundwater Versus Non-Groundwater Dependent Counties

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

Corn Wheat Hay Sorghum

No GW GW No GW GW No GW GW No GW GW

PDSI <- 3 -0.0360*** -0.0310*** -0.0298*** -0.0126*** -0.0392*** -0.0315*** -0.0346*** -0.0207***

(0.00626) (0.00525) (0.00243) (0.00312) (0.00494) (0.00590) (0.0119) (0.00697)

PDSI between -3 and -1 -0.0116*** -0.00639** -0.00933*** -0.00556*** -0.0163*** -0.0175*** -0.0152** -0.0165***

(0.00425) (0.00313) (0.00156) (0.00170) (0.00324) (0.00259) (0.00601) (0.00370)

PDSI between 1 and 3 -0.0157*** -0.00602** -0.00241* 0.00257 -0.0135*** 0.00542* -0.00788 0.00233

(0.00400) (0.00302) (0.00141) (0.00170) (0.00354) (0.00276) (0.00563) (0.00344)

PDSI > 3 -0.0199*** -0.0179*** -0.00603*** -0.00275 -0.0242*** 0.000656 0.00458 0.0178***

(0.00525) (0.00425) (0.00169) (0.00243) (0.00364) (0.00313) (0.00603) (0.00471)

Annual average temperature 0.117 0.413*** 0.00455 0.207*** 0.153*** 0.391*** -0.0304 0.318***

(0.0745) (0.0534) (0.0234) (0.0448) (0.0386) (0.0581) (0.164) (0.0675)

Annual average temperature -0.00568*** -0.0119*** -0.00128* -0.00597*** -0.00407*** -0.0104*** -0.00150 -0.00562***
squared

(0.00205) (0.00148) (0.000680) (0.00123) (0.00121) (0.00151) (0.00359) (0.00166)

A-7
Precipitation 0.00109*** 0.000553*** 0.000507*** 0.000655*** 0.000834*** -0.000339*** 0.000569 0.00171***

(0.000287) (0.000140) (5.71e-05) (6.89e-05) (0.000180) (0.000122) (0.000387) (0.000217)

Precipitation squared -5.20e-07*** -3.12e-07*** -1.20e-07*** -1.62e-07*** -4.20e-07*** -7.64e-08 -2.68e-07 -9.00e-07***

(1.40e-07) (7.82e-08) (1.82e-08) (2.31e-08) (8.19e-08) (5.40e-08) (1.75e-07) (1.18e-07)

Constant 11.40*** 10.10*** 10.68*** 8.458*** 11.54*** 10.33*** 13.79*** 8.140***

(0.718) (0.527) (0.218) (0.420) (0.356) (0.589) (2.034) (0.773)

County fixed effects yes yes yes yes yes yes yes yes

Year fixed effects yes yes yes yes yes yes yes yes

Observations 5,206 9,282 8,505 8,166 8,523 10,174 3,243 9,665

R-squared 0.327 0.351 0.148 0.129 0.167 0.225 0.304 0.218

Number of fips 263 355 360 413 386 422 168 376

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and 1. All dependent variables are
logged. PDSI variables are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

A-8
Table A5: Drought Impacts on Yield Per Acre-Groundwater Versus Non-Groundwater Dependent Counties

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

Corn Wheat Hay Sorghum

No GW GW No GW GW No GW GW No GW GW

PDSI <- 3 -0.0108*** -0.0194*** -0.0161*** -0.0216*** -0.0168*** -0.0144*** -0.0206*** -0.0263***

(0.00232) (0.00198) (0.00152) (0.00157) (0.00225) (0.00237) (0.00401) (0.00227)

PDSI between -3 and -1 -0.000729 -0.000923 -0.00452*** -0.00517*** -0.00732*** -0.00687*** -0.00295 -
0.00820***

(0.00166) (0.00108) (0.000895) (0.000991) (0.00151) (0.00126) (0.00217) (0.00113)

PDSI between 1 and 3 -0.00404*** -0.00389*** -8.93e-05 0.000243 -0.000988 0.00281*** -0.00556*** -
0.00287***

(0.00144) (0.00101) (0.000821) (0.000864) (0.00122) (0.00102) (0.00201) (0.000993)

PDSI > 3 -0.0110*** -0.00892*** -0.00313*** -0.00435*** -0.00809*** -0.00424*** -0.0101*** -

0.00764***

(0.00154) (0.00132) (0.000939) (0.00106) (0.00134) (0.00120) (0.00215) (0.00130)

Annual average 0.104*** 0.155*** 0.103*** 0.172*** -0.0305* 0.198*** 0.00170 0.187***
temperature

(0.0252) (0.0247) (0.0142) (0.0268) (0.0162) (0.0260) (0.0574) (0.0239)

A-9
Annual average -0.00487*** -0.00595*** -0.00357*** -0.00488*** 0.000888* -0.00554*** -0.00276** -
temperature squared 0.00584***

(0.000694) (0.000692) (0.000450) (0.000743) (0.000480) (0.000662) (0.00134) (0.000633)

Precipitation 0.000505*** 1.52e-05 0.000356*** 0.000384*** 0.000296*** -0.000360*** 0.000549*** 0.00104***

(0.000109) (5.98e-05) (3.10e-05) (3.94e-05) (0.000104) (5.90e-05) (0.000179) (0.000100)

Precipitation squared -1.84e-07*** 1.52e-08 -8.11e-08*** -8.36e-08*** -2.20e-07*** 1.65e-08 -2.27e-07*** -5.64e-
07***

(5.03e-08) (3.00e-08) (7.83e-09) (1.08e-08) (5.06e-08) (3.22e-08) (8.15e-08) (5.77e-08)

Constant 3.926*** 3.760*** -0.103 -0.723*** 3.755*** 2.011*** 4.698*** 2.104***

(0.244) (0.240) (0.116) (0.245) (0.155) (0.266) (0.660) (0.236)

County fixed effects yes yes yes yes yes yes yes yes

Year fixed effects yes yes yes yes yes yes yes yes

Observations 5,206 9,282 8,505 8,166 8,523 10,174 3,243 9,665

R-squared 0.354 0.315 0.194 0.209 0.139 0.237 0.356 0.321

Number of Counties 263 355 360 413 386 422 168 376

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and 1. All dependent variables are
logged. PDSI variables are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, *

p<0.1

A-10
Table A6: Drought Impacts on Harvested Acres-Groundwater Versus Non-Groundwater Dependent Counties

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

Corn Wheat Hay Sorghum

No GW GW No GW GW No GW GW No GW GW

PDSI <- 3 -0.0252*** -0.0116** -0.0137*** 0.00902*** -0.0224*** -0.0171*** -0.0140 0.00554

(0.00553) (0.00489) (0.00192) (0.00269) (0.00397) (0.00472) (0.0101) (0.00628)

PDSI between -3 and -1 -0.0109*** -0.00547* -0.00481*** -0.000386 -0.00902*** -0.0106*** - -0.00825**
0.0122**

(0.00394) (0.00315) (0.00121) (0.00139) (0.00278) (0.00218) (0.00547) (0.00360)

PDSI between 1 and 3 -0.0117*** -0.00213 -0.00232** 0.00233 -0.0125*** 0.00262 -0.00232 0.00520

(0.00374) (0.00297) (0.00112) (0.00145) (0.00318) (0.00250) (0.00498) (0.00326)

PDSI > 3 -0.00896* -0.00894** -0.00289** 0.00159 -0.0162*** 0.00490* 0.0146** 0.0254***
*

(0.00486) (0.00418) (0.00129) (0.00201) (0.00314) (0.00283) (0.00557) (0.00439)

Annual average temperature 0.0125 0.258*** -0.0982*** 0.0351 0.183*** 0.193*** -0.0321 0.131**

(0.0675) (0.0495) (0.0174) (0.0306) (0.0315) (0.0455) (0.145) (0.0665)

Annual average temperature -0.000816 -0.00591*** 0.00229*** -0.00109 -0.00496*** - 0.00126 0.000222
squared 0.00486***

A-11
 (0.00180) (0.00132) (0.000494) (0.000840) (0.000984) (0.00123) (0.00327) (0.00166)

Precipitation 0.000588** 0.000538*** 0.000151*** 0.000272** 0.000538*** 2.05e-05 1.98e-05 0.000665**

* *

(0.000266) (0.000131) (4.31e-05) (5.90e-05) (0.000137) (0.000105) (0.00030 (0.000197)

0)

Precipitation squared -3.37e- -3.27e- -3.85e-08*** -7.87e- -1.99e-07*** -9.29e- -4.07e-08 -3.36e-
07*** 07*** 08*** 08** 07***

(1.27e-07) (7.21e-08) (1.37e-08) (2.09e-08) (5.98e-08) (4.48e-08) (1.39e- (1.08e-07)

07)

Constant 7.472*** 6.342*** 10.79*** 9.182*** 7.787*** 8.316*** 9.090*** 6.035***

(0.685) (0.526) (0.163) (0.293) (0.302) (0.453) (1.724) (0.756)

County fixed effects yes yes yes yes yes yes yes yes

Year fixed effects yes yes yes yes yes yes yes yes

Observations 5,206 9,282 8,505 8,166 8,523 10,174 3,243 9,665

R-squared 0.222 0.234 0.065 0.062 0.227 0.228 0.351 0.209

Number of Counties 263 355 360 413 386 422 168 376

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1 and 1. All dependent
variables are logged. PDSI variables are number of months with PDSI in indicated bin.

*** p<0.01, ** p<0.05, * p<0.1

A-12
Table A7: Drought Impacts on Crop Production by Growing Season and Non-Growing Season

(5) (6) (7) (8)

Corn Hay Wheat Sorghum

Growing season PDSI <- 3 -0.0468*** -0.0435*** -0.0620*** -0.0734***

(0.00745) (0.00379) (0.00689) (0.0102)

Growing season PDSI between -3 and -1 -0.00599 -0.00972*** -0.0247*** -0.0291***

(0.00434) (0.00184) (0.00334) (0.00548)

Growing season PDSI between 1 and 3 -0.0309*** 0.000145 0.0125*** -0.00623

(0.00425) (0.00178) (0.00334) (0.00463)

Growing season PDSI > 3 -0.0563*** -0.00182 0.0275*** 0.0135*

(0.00707) (0.00301) (0.00557) (0.00818)

Non-growing season PDSI <- 3 -0.0100 0.00705 -0.00192 0.0315*

(0.0110) (0.00510) (0.00950) (0.0170)

Non-growing season PDSI between -3 and -1 -0.0110** -0.00514** -0.0122*** -0.00310

(0.00502) (0.00224) (0.00424) (0.00673)

Non-growing season PDSI between 1 and 3 0.0145** -0.00159 -0.0236*** 0.00771

(0.00584) (0.00233) (0.00466) (0.00705)

Non-growing season PDSI > 3 0.0255** -0.00855** -0.0548*** 0.0162

(0.0100) (0.00390) (0.00749) (0.0106)

Annual average temperature 0.303*** 0.0685*** 0.205*** 0.264***

(0.0454) (0.0212) (0.0297) (0.0616)

Annual average temperature squared -0.00974*** -0.00270*** -0.00580*** -0.00503***

(0.00124) (0.000597) (0.000835) (0.00147)

Precipitation 0.000697*** 0.000552*** 0.000140 0.00125***

A-13
 (0.000127) (4.60e-05) (0.000102) (0.000182)

Precipitation squared -3.70e-07*** -1.29e-07*** -2.10e-07*** -5.99e-07***

(6.72e-08) (1.59e-08) (4.71e-08) (9.51e-08)

Constant 10.73*** 9.962*** 11.73*** 9.049***

(0.442) (0.198) (0.296) (0.725)

County fixed effects yes yes yes yes

Year fixed effects yes yes yes yes

Observations 14,488 16,671 18,697 12,908

R-squared 0.336 0.128 0.181 0.225

Number of Counties 618 773 808 544

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI between -1
and 1. All dependent variables are logged. PDSI variables are number of months with PDSI in indicated bin. Growing
season defined as April to September.

*** p<0.01, ** p<0.05, * p<0.1

A-14
Table A8: Drought Impacts on Yield Per Acre by Growing Season and Non-Growing Season

(5) (6) (7) (8)

Corn Hay Wheat Sorghum

Growing season PDSI <- 3 -0.0282*** -0.0234*** -0.0346*** -0.0300***

(0.00362) (0.00239) (0.00292) (0.00377)

Growing season PDSI between -3 and -1 -0.00938*** -0.00521*** -0.0141*** -0.00871***

(0.00165) (0.00111) (0.00144) (0.00193)

Growing season PDSI between 1 and 3 -0.00379** -0.000438 0.00783*** -0.00670***

(0.00159) (0.00104) (0.00136) (0.00171)

Growing season PDSI > 3 -0.00680** -0.00507*** 0.0130*** -0.0137***

(0.00282) (0.00175) (0.00232) (0.00271)

Non-growing season PDSI <- 3 -0.00133 -0.0121*** 0.00779** -0.0187***

(0.00416) (0.00289) (0.00371) (0.00514)

Non-growing season PDSI between -3 and -1 0.00908*** -0.00495*** -0.00134 -0.00555**

(0.00192) (0.00128) (0.00179) (0.00237)

Non-growing season PDSI between 1 and 3 -0.00547*** 0.000405 -0.00801*** -1.44e-05

(0.00198) (0.00128) (0.00158) (0.00195)

Non-growing season PDSI > 3 -0.0129*** -0.00227 -0.0290*** -0.00344

(0.00320) (0.00213) (0.00266) (0.00319)

Annual average temperature 0.135*** 0.113*** 0.0397*** 0.156***

(0.0180) (0.0123) (0.0127) (0.0229)

Annual average temperature squared -0.00545*** -0.00356*** -0.00137*** -0.00542***

(0.000502) (0.000367) (0.000348) (0.000583)

A-15
 0.000162**
Precipitation * 0.000372*** -7.54e-05 0.000778***

(5.65e-05) (2.47e-05) (5.82e-05) (8.23e-05)

-8.94e-
Precipitation squared -4.35e-08 -8.38e-08*** 08*** -3.70e-07***

(2.83e-08) (6.55e-09) (3.04e-08) (4.44e-08)

Constant 3.832*** -0.181* 3.367*** 2.627***

(0.175) (0.108) (0.126) (0.237)

County fixed effects yes yes yes yes

Year fixed effects yes yes yes yes

Observations 14,488 16,671 18,697 12,908

R-squared 0.322 0.192 0.170 0.314

Number of Counties 618 773 808 544

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI
between -1 and 1. All dependent variables are logged. PDSI variables are number of months with PDSI in
indicated bin. Growing season defined as April to September.

*** p<0.01, ** p<0.05, * p<0.1

A-16
Table A9: Drought Impacts on Harvested Acres by Growing Season and Non-Growing Season

(5) (6) (7) (8)

Corn Hay Wheat Sorghum

Growing season PDSI <- 3 -0.0185*** -0.0200*** -0.0274*** -0.0434***

(0.00679) (0.00287) (0.00572) (0.00944)

Growing season PDSI between -3 and -1 0.00339 -0.00451*** -0.0106*** -0.0204***

(0.00429) (0.00150) (0.00299) (0.00502)

Growing season PDSI between 1 and 3 -0.0271*** 0.000583 0.00464 0.000470

(0.00392) (0.00158) (0.00296) (0.00433)

Growing season PDSI > 3 -0.0495*** 0.00325 0.0145*** 0.0272***

(0.00649) (0.00257) (0.00471) (0.00779)

Non-growing season PDSI <- 3 -0.00870 0.0192*** -0.00971 0.0502***

(0.0100) (0.00414) (0.00827) (0.0151)

Non-growing season PDSI between -3 and -1 -0.0201*** -0.000189 -0.0109*** 0.00244

(0.00495) (0.00195) (0.00376) (0.00620)

Non-growing season PDSI between 1 and 3 0.0199*** -0.00199 -0.0156*** 0.00772

(0.00560) (0.00204) (0.00429) (0.00678)

Non-growing season PDSI > 3 0.0383*** -0.00628** -0.0259*** 0.0196**

(0.00932) (0.00314) (0.00663) (0.00991)

Annual average temperature 0.168*** -0.0442*** 0.165*** 0.108*

(0.0421) (0.0154) (0.0243) (0.0585)

Annual average temperature squared -0.00429*** 0.000861** -0.00443*** 0.000396

(0.00110) (0.000423) (0.000693) (0.00142)

A-17
Precipitation 0.000535*** 0.000180*** 0.000215*** 0.000468***

(0.000120) (3.62e-05) (8.16e-05) (0.000154)

Precipitation squared -3.27e-07*** -4.56e-08*** -1.20e-07*** -2.29e-07***

(6.15e-08) (1.26e-08) (3.60e-08) (7.82e-08)

Constant 6.895*** 10.14*** 8.363*** 6.423***

(0.439) (0.148) (0.244) (0.677)

County fixed effects yes yes yes yes

Year fixed effects yes yes yes yes

Observations 14,488 16,671 18,697 12,908

R-squared 0.225 0.048 0.216 0.229

Number of Counties 618 773 808 544

Standard errors in parentheses clustered at the county level. Impacts are relative to a month with PDSI
between -1 and 1. All dependent variables are logged. PDSI variables are number of months with PDSI in
indicated bin. Growing season defined as April to September.

*** p<0.01, ** p<0.05, * p<0.1

A-18
Table A10: Agricultural Results with HUC Level PDSI Counts

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

VARIABLES Corn Hay Wheat Sorghum

PDSI<-3 -0.0321*** -0.0211*** -0.0382*** -0.0296***

(0.00450) (0.00225) (0.00438) (0.00690)

PDSI -3 to-1 -0.00628** -0.00581*** -0.0174*** -0.0155***

(0.00286) (0.00133) (0.00228) (0.00350)

PDSI 1 to 3 -0.00880*** 6.66e-06 -0.00252 0.00215

(0.00282) (0.00120) (0.00253) (0.00327)

PDSI >3 -0.0164*** -0.00272* -0.00905*** 0.0155***

(0.00384) (0.00151) (0.00286) (0.00404)

HUC Level PDSI<--3 -5.95e-05 -0.000394 0.00129* 0.000373

(0.000959) (0.000495) (0.000736) (0.00113)

HUC Level PDSI -3 to-1 -0.00144* -0.000938*** -0.00125** -0.00161**

(0.000808) (0.000321) (0.000539) (0.000720)

HUC Level PDSI 1 to 3 -0.00122 4.98e-05 -0.000847 -0.00159**

(0.000914) (0.000333) (0.000666) (0.000703)

HUC Level PDSI >3 -0.000754 -0.000779** -0.000693 -0.000291

(0.000855) (0.000346) (0.000576) (0.000658)

Annual average temperature 0.283*** 0.0925*** 0.248*** 0.268***

(0.0459) (0.0209) (0.0315) (0.0630)

Annual average temperature

squared -0.00938*** -0.00337*** -0.00683*** -0.00513***

A-19
 (0.00128) (0.000578) (0.000888) (0.00151)

Precipitation 0.000709*** 0.000537*** 0.000194* 0.00117***

(0.000141) (4.89e-05) (0.000107) (0.000188)

Precipitation squared -3.84e-07*** -1.29e-07*** -2.44e-07*** -5.64e-07***

(7.59e-08) (1.74e-08) (5.03e-08) (9.78e-08)

Constant 10.99*** 9.773*** 11.33*** 9.071***

(0.440) (0.204) (0.313) (0.737)

Observations 13,537 15,027 17,055 12,087

R-squared 0.341 0.132 0.179 0.225

Number of fips 563 700 732 507

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

A-20
Table A11: Agricultural Results with HUC Level PDSI Counts without County Counts

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

VARIABLES Corn Hay Wheat Sorghum

HUC Level PDSI<--3 -0.00253*** -0.00235*** -0.00214** -0.000708

(0.000876) (0.000635) (0.000862) (0.000999)

HUC Level PDSI -3 to-1 -0.00202*** -0.00150*** -0.00283*** -0.00228***

(0.000749) (0.000334) (0.000570) (0.000691)

HUC Level PDSI 1 to 3 -0.00224*** -0.000253 -0.00147** -0.00166**

(0.000837) (0.000334) (0.000649) (0.000656)

HUC Level PDSI >3 -0.00254*** -0.00150*** -0.00236*** 0.000688

(0.000783) (0.000346) (0.000559) (0.000671)

Annual average temperature 0.295*** 0.0803*** 0.229*** 0.259***

(0.0474) (0.0210) (0.0315) (0.0631)

Annual average temperature

squared -0.00996*** -0.00345*** -0.00718*** -0.00586***

(0.00132) (0.000588) (0.000885) (0.00152)

Precipitation 0.000822*** 0.000669*** 0.000478*** 0.00153***

(0.000139) (5.29e-05) (0.000109) (0.000192)

Precipitation squared -4.37e-07*** -1.52e-07*** -3.21e-07*** -6.28e-07***

(7.47e-08) (1.94e-08) (5.33e-08) (9.84e-08)

Constant 10.92*** 9.944*** 11.67*** 9.394***

(0.448) (0.203) (0.312) (0.735)

Observations 13,537 15,027 17,055 12,087

A-21
R-squared 0.336 0.122 0.170 0.220

Number of fips 563 700 732 507

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

A-22