While the negative impacts of drought are often disproportionately spread to low-income individuals, we have seen that windfall gains to agriculture create short-run spillovers to other industries . This spillover is likely to impact closely related industries in terms of input-output and exchange the most . In the case of agriculture, we understand this to be industries that directly rely on agricultural output such as food manufacturing and wholesale. When we consider the spillover impacts of a crisis event such as drought, there is evidence that the volatility of agricultural input exerts significant spillover effects on the volatility of agricultural output and retail food prices . I treat drought as a crisis event that creates volatility in the agricultural input price of water. Hornbeck and Keskin found that windfall gains to the agricultural industry can create short run spillover to other local industries. While they found no evidence of long-run sustained spillover, as my data does not include ex-post results, this does not pose a threat to the scope of my study. In general,vertical plant tower spillover is likely to be strongest in closely related industries and exert significant impact in times of crisis or in instances of volatile input prices.
I observe the crisis period of the 2012 to 2016 California drought and the volatility it created in agricultural input and outputs to evaluate how drought impacts local incomes and employment. To the best of my knowledge, this is the first study utilizing individual data to analyze possible spillover impacts of the 2012 to 2016 California drought.3 Previous studies focus on statewide impacts of the recent California drought or analyze different aspects of labor market impacts for either this or other historical droughts. Literature suggests that I would identify a significant negative impact on the agricultural industry and closely related industries during this time period.I estimate a difference in difference regression comparing outcomes in San Joaquin and Tulare, the two counties that experienced 90% of farmland fallowing, with outcomes in similar Central Valley counties. I find a significant 9% reduction in employment and an 11% reduction in individual income for those working in agriculture.4 Although I expected to discover contractions in closely related industries, I observe almost no impact on these industries’ employment and incomes. There were also no significant differences in the impact of the drought between males and females when my regression was run with a gender interaction. However, an additional interaction shows a significant and highly negative impact on Hispanic individual employment in agriculture by 12% and further reduction in wages by 13%.
This signals that although the economy was resilient, the drought disproportionately impacted Hispanic agricultural workers. Additionally, the small spillovers that occurred into related industries had impact only on Hispanic workers. This result represents a departure from traditional intuition that observes spillover between closely related industries, particularly during a crisis. Although these results are unusual, further robustness checks and a statistically optimized control group would be necessary to confirm the lack of spillover effects. This instance of limited spillover could reflect the recent popularization of water permit trading amongst farmers and the introduction of new drought-related welfare programs . Data on water trading rates and prices are not currently aggregated or publicly available but would be an area for potential further study. Prior research finds that water management policy coordinated with farmers has the potential to increase environmental and economic gains to all parties . A detailed input-output study would also further improve the validity of my results. These models are commonly used to analyze changes in farmer behavior in reaction to price changes among other purposes and could be fit to the scenario of a drought .Lund et al. synthesize their past research on drought with contributions from other prominent researchers in the field to create a full picture of the impact in “Lessons from California’s 2012-2016 Drought”. I draw from components focusing on employment and revenue losses.
In their preliminary findings, agriculture was the industry primarily impacted through increased pumping costs of $600 million per year and half a million acres of fallowed crop area. When water supplies reached a low in 2012 to 2015, certain negotiated contracts with water projects received zero deliveries. Lund et al. touch on the uncertainty for future strength during drought caused by overdraft of groundwater, first reported by MacEwan et al. This will most likely hit rural areas the hardest as they have the least access to water and lower aquifer elevations available for groundwater pumping. The paper finds that overall resilience was due to strong prices for key specialty crops, ability to rely on groundwater, effective water management, and the beginnings of a robust water trading market. Despite this, they acknowledge that these costs were likely concentrated in areas with a lack of easily accessible groundwater. There is no detailed analysis of county level impacts on these rural and dry counties in the San Joaquin Valley and Tulare River Basin due to the 2012 to 2016 drought. Cooley et al. similarly find that overall impacts were mitigated, but discuss the need for local variability estimates for areas that experienced intense fallowing. Related literature has indeed shown that rural and low-income individuals have less tolerance for natural disasters. A drought of a similarly intense magnitude occurred in Australia from 2001 to 2004. Carroll et al. used life satisfaction survey data to estimate that the occurrence of the drought was equivalent to an annual reduction in income of $18,000 . Using fixed effects to control for unobserved area characteristics, this impact appeared only for individuals living in rural areas. While the Australian economy suffered more heavily due to a lack of drought infrastructure, the divide between rural and urban individuals in this case is clear. I use a similar regression with fixed effects and demographic controls to look at labor market outcomes for the California Drought from 2012 to 2016. As with the Australian drought, this recent California drought has been proven to be hydrologically severe and sustained marked losses within the agricultural sector . Following the focus on rural and low-income individuals I estimate differences between the hydrologically dry rural counties with counties that were able to mitigate most drought losses with groundwater and water project contracts. Based on further studies I determine that San Joaquin and Tulare counties were the most heavily impacted during this time period and faced the heaviest groundwater pumping costs. My study differs in its approach,10 liter drainage collection pot data and focus. I choose to use survey data and look at individual characteristics within the more closely focused county groups. Additionally, I test for differences in outcomes for Hispanic individuals and females. The 2012 to 2016 California drought was found to create emotional distress regarding food insecurity, particularly in Hispanic households . My results and analysis provide further evidence of the harsher penalties imposed on rural and Hispanic agricultural households due to drought conditions.I additionally confirm the question theorized by earlier research in this field that there indeed was variability in county level impact due to the drought.“Does Agriculture Generate Local Economic Spillovers? Short-Run and Long-Run Evidence from the Ogallala Aquifer” by Hornbeck and Keskin is the most closely related and influential paper in the design and understanding of my topic.
This paper analyzes the impact of new technology that allowed farmers to utilize a new groundwater source, the Ogallala Aquifer. This windfall gain to the agricultural sector allows Hornbeck and Keskin to estimate the differences between counties with a high proportion of areas with increased water access and those that largely missed the benefits of this new water source. They estimate a difference in difference regression controlling for various agricultural effects and time effects to estimate the spillover impact of increased water access. They find that areas with high exposure to the Ogallala had increased agricultural gains through land value and revenue. This also caused an exogenous increase in rural farm employment. Similar to my paper, they set manufacturing, wholesale, retail, and services as comparison industries for their economic closeness. While this did not extend to the long-run, Hornbeck and Keskin did find short-run statistically significant expansions in these industries. While this result is different from the lack of spillover seen in my results, I attribute this limit of negative spillover to efficient water management and programs to limit contractions to the agricultural industry itself. Notably, Moretti demonstrated that spillovers occur between closely related industries with greater frequency and intensity than in industries that are distant. Instead of focusing on measures of agricultural workers or rural areas, Moretti looks to the proportion of college-educated workers within a data set cataloging production plant productivity. He finds an increase in plant productivity as a result of the faster growth of the proportion of college-educated workers in an area. This effect is larger for economically close industries, reflecting the spillover of knowledge and physical capital accumulation. Additionally, Kang et al. find that there is a strong impact of spillover during and after the crisis period by estimating commodity futures returns. This reflects a premium on uncertainty and increased supply chain costs for closely related industries that rely on crude commodities. We would expect to see the greatest impact on industries purchasing and relying on outputs of the agricultural sector . My findings that closely related sectors were not impacted is a departure from this intuition and is reflective of the effective water management and drought mitigation techniques that contained heavy losses to parts of the agricultural industry while keeping agricultural produce prices stable. Nazlioglu et al. find that after the occurrence of a crisis in oil markets there is significant market volatility on key agricultural commodities. Using a GARCH model they show that there is a growing linkage between agriculture and energy markets due to their similarities and investor profile. Further work done by Apergis and Rezitis delves further into the links between agricultural input prices and output commodities. They used agricultural commodity prices in Greece from 1985 to 1990 to test for links in equilibrium price patterns. The study finds that there are significant linkages in price variation between agricultural input and output prices, and between agricultural output prices and retail food output prices. They also find evidence of imperfect price transmission among the three categories so that exogenous shocks would create disparate welfare changes among market participants. Since output prices were observed to be more flexible than input and retail prices, this indicates that general price decreases in a crisis would create short term losses for farmers as their prices decrease faster than input prices. This aligns with my findings that agricultural earnings had large short-run decreases due to drought-related shocks. The main purpose of this study is to quantify how economic spillovers between industries impacted individuals living in areas severely affected by the 2012 to 2016 drought. I chose to use U.S. government survey data to have access to one of the largest data sources on my target counties while retaining other significant data measures on the socio-economic profile of the individuals. The American Community Survey collects cross-sectional data on individuals with attached characteristics and publishes annually to the Integrated Public Use Microdata Series . The ACS uses a series of monthly samples on 250,000 addresses to produce an annual estimate of data for the same small areas on 3,000,000 addresses. My data extract is limited to individuals in the California Central Valley in the years 2006 to 2017 for sample size consistency. I use the California Research Bureau classification of the 18 Central Valley counties.To ensure the accuracy of my results I used the IPUMS provided CPI adjustment factor to convert income to 2005 dollars, so estimates are standardized to the beginning of the observed time period. Additionally, only individuals in the age range of 20 to 65 that did not reside in group quarters were kept, to ensure individuals not typically in the labor market did not distort income estimates. Before performing analysis, observations with missing values for labor industry classification or income were removed. After these modifications, the data includes 435,996 individual observations on individuals living in counties categorized as the Central Valley. I used sex, educational attainment, and race control variables to add accuracy to the estimate without over fitting my model.6