Specific Objectives of the Project:

1. Describe how nitrogen management practice adoption varies across farm size, crop types, and location. Analyze how grower behavior varies across fields and crop types.
2. Analyze whether there are diminishing marginal returns to nitrogen efficiency if multiple nutrient management practices are adopted.
3. Evaluate the effectiveness of practices designed to alter grower decision making on nitrogen application rates.
4. Determine a set of practices that growers find beneficial and have the strongest impact on nitrogen efficiency, to provide tangible recommendations to policymakers and technical advisers.

Summary of Research to Date:
Current results rely on data collected from the Irrigation and Nitrogen Management Plans that are required of agricultural producers in the Central Valley. The focus area is the East San Joaquin Valley (ESJV) in 2019, as this was the first region where data were required to be made publicly available. In the sample, there are 1713 growers applying nitrogen to 4810 fields.

We find in the ESJV, nitrogen application rates are highest for nut crops, with vegetable row crops not far behind. On average, vegetable row crops have the lowest nitrogen efficiency (calculated as N applied/N removed through biomass). The low nitrogen efficiency in vegetables is driven primarily by sweet potatoes. In total, split fertilizer applications, testing soil for residual nitrogen, and tissue testing are the most popular nutrient management practices adopted by growers. These results are similar to those found in surveys of growers in the Central Valley (Rudnick et al. 2021). The use of neutron probes, pressure bombs, and cover crops are the least likely to be adopted.

We regress nitrogen management practices adopted on nitrogen application rates, and nitrogen use efficiency, while accounting for irrigation system and crop and farmer fixed effects. We find that the adoption of cover crops reduces nitrogen application rates, but that several practices, including foliar nitrogen application, split fertilizer applications, tissue testing, soil testing, and pressure bombs, actually increase the nitrogen application rate. Tissue testing was significantly likely to improve nitrogen use efficiency, but use of soil testing and pressure bombs were associated with reduced efficiency.

Altogether, these results were a bit puzzling, and require a deeper exploration. We think there are two main challenges at play. The use of individual fixed effects, while they help combat omitted variable bias (farm and demographic characteristics), limit the variation in our current sample, especially since the majority of growers in the sample have a single field. We are in the process of securing more recent data to both expand the area of study to the full Central Valley, as well as the years of our sample, adding additional variation to the sample. Secondly, the number of practices in our data is likely leading to multicollinearity issues. We see significant evidence that farmers regularly adopt
“bundles” of practices in tandem, such as irrigation N tests and fertigation. To combat this issue, we use random forest techniques to isolate the variables that have the largest predictive power to explain nitrogen use, and reduce the number of practices in our model. The four main practices that have explanatory power include testing irrigation water for nitrogen, tissue testing, split nitrogen application, and foliar nitrogen application. We are also speaking with agronomists on alternate methods to group practices that are likely to be adopted together. We expect further insights and a concrete working paper to emerge with the application of these techniques to the expanded dataset.

Specific Objectives of the Project:
1. To conduct a study and provide data analysis on effects of severity of the pandemic measured in terms of county level health outcomes (deaths, deaths, hospitalizations) by day on the mobility and social distancing behavior across counties in California, and how mobility and social distancing changes when county level mitigation strategies and stay at home mandates are implemented using an econometric model and unique policy county level data merged with data on daily county level mobility measures (via a data sharing agreement between U C Berkeley and Unacast).
2. To develop an econometric model to measure the effects of the severity of the pandemic, the mobility changes, and the mandate effects on business revenues and rate of closures by county over time, investigating such patterns for different types of business.
3. Distinguish such patterns of effects by the county rural classification, and also as a function of availability of hospitals using a unique data set of hospital locations by county and characteristics of hospitals in terms of number of beds, trauma centers, and number of recent closures of hospitals, especially in rural counties.

Summary of Research to Date:
Results are preliminary. They rely on mobility data collected from Unacast by county by day, and the county level and city level extensive mandate and policy implementation data we collected. We also used womply business data by week, for less representative samples than county level data (we only realized this after we had access to the data). In addition, we define counties according to population, rural definition, and merge hospital census data with hospital characteristics and capacity into our analysis.

Objective 1: On average, mandates have a significant effect on mobility and encounters with other individuals. The measures we use are average distance traveled, distance to non-essential businesses, and the number of encounters (based on the number of cell phones in a radius to a certain cell phone by day) in the data. All three distance measures drop significantly even before the first policies are implemented, in all counties (first implementers and also for the later implementing counties) The incremental effect of mandates is a proportion of the initial drop.

Objective 2: There is a significant exit of businesses in the data during covid, resulting in tremendous sample selection that does not allow us to look at the effect on business revenues, as the businesses that survive in the sample during covid are the ones that are different from the ones that exit, so we do not have comparable samples throughout as the policies are implemented. Second, we find that almost all business sectors suffer the exit of small businesses. The effect is long-lasting and we do not have data long enough to investigate recovery. We plan to do so in a future project.
Objective 3: we did not have enough coverage of small business data in rural areas to investigate the differential effect of the pandemic and policies to investigate exit intensity. We do not find differential effects in mobility and human encounter responses to the mandates between rural and nonrural counties. We did not have enough power
(sample size) to distinguish the effect on mobility due to hospital capacity in rural versus non-rural counties.

Specific Objectives of the Project:
Gather data on consumer stated valuations for seafood and aquaculture attributes

Project Report/Summary of Results:
We have gathered a comprehensive data set on the geographic and time patterns of production of aquaculture in the United States and have started developing a survey instrument to producers and also an additional survey instrument to consumers in the production areas, as well as a representative sample of US consumers . We do not have a draft available yet of your data analysis and plan to have one in a year.

Specific Objectives of the Project
This project empirically investigates consumer demand for brands, private labels, and generics. We use a reduced form model to investigate consumer demand changes when information about generics versus brand name is provided at the point of purchase. We also reveal the share of consumers in a store purchasing generics and assess whether a larger peer group presence of generic purchases sways consumers to also try out generics.

Project Report/Summary of Results
Using a field experiment we empirically investigate factors that may explain why consumers do not buy cheaper private label alternatives to brands. Our approach is reduced form and we use a difference in difference strategy. We estimate the effect of each treatment on total purchasing and on generic share, using a set of comparable untreated stores as control.
There were three treatments implemented: (1) Enhance the point that generics are “the same” (and here effects are not statistically significant) ; (2) Making price differences more visible (salient) causes significant increases in generic share and the entire effect is driven by customers with previous generics purchase history, (3) Informing people about generic purchasing propensity of others causes increases in generic share that are large in magnitude. The category used is a first step towards implementing similar treatments in a wider and more heterogenous set of categories such as staple foods.

Objectives of the Project
1. Determine whether egg consumption decreased due to the news of the Salmonella outbreak.
2. Determine whether consumers are substituting away from conventional eggs towards other types of eggs (organic, cage free, free range) or egg substitutes.
3. Determine how long this substitution effect (if any) lasts.
4. Examine whether there are any heterogeneous effects based on income and demographics.

Summary of Results
We examine how consumers in California reacted to three consecutive egg recalls during the 2010 Salmonella outbreak. Eggs infected with Salmonella were recalled through codes clearly labeled in egg boxes, leaving no infected eggs in stores. Using a large product-level scanner data set from a national grocery chain, we test whether consumers reduced egg purchases. Using a difference-in-difference approach, we find a 9 percent reduction in egg sales in California following the three egg recalls. Given an overall price elasticity for eggs in U.S. households of -0.1, this sales reduction is comparable to an almost 100% increase in price. We find no evidence of substitution toward other “greener" type of eggs, such as organic or cage free eggs. We also find no correlation with demographics such as income, but we do find that areas that had a larger than average household size decreased egg purchases more. We also find differentiated effects among Northern and Southern Californian stores. Although the national grocery chain had infected eggs only in Northern California, we find that Southern Californian stores had lower egg sales as well. The sales reduction in Southern California was half as large as the reduction in Northern California, and is consistent with media and reputation effects being significant determinants of demand, even in the absence of an actual food infection occurring in a region.

Objective of the Project
Objective 1:
• We will match weekly data from 2006-2010 in California stores with foreclosure data (at
the zip code level). We will exploit the geographic and temporal variation in the degree of
severity of the foreclosure crisis in order to measure its impacts on food purchases. We will
focus on key product categories in order to measure the effect of the foreclosure crisis on:
1) Consumption shifting between items in product categories
2) Willingness to pay for product attributes (i.e. health)
3) Private Label Consumption
4) Consumption of promotional items

Objective 2:
• Use results from above to quantify and highlight implications of little known
consequences of the foreclosure crisis, such as:
1) Modified dietary composition
2) Nutritional intake
We use a unique, product- and store-level scanner data set containing five years of weekly
information as well as panel data on each consumer shopping behavior. The data feature quantity
and price of each product sold at all retail stores in California. We shall combine these scanner
data with socio-demographic statistics provided by the United States Census Bureau (by zip
code) and also with frequent loyalty card based demographic data collected by the retail chain,
that we hope to obtain also. We also combine the purchase and census data with monthly
foreclosure rates indicators at the zip code level. We also include other housing market indicators
such as home prices, new constructions, and census level information on renting vs. ownership.

Summary of Results
We collected and analyzed a detailed foreclosure data-set from the RAND corporation, that includes the number of foreclosed properties by house type for each month from 2002-2012 across all zip codes in California, in order to investigate the spatial and time series variation in foreclosures in California. We use both foreclosures of single-family and multi-family residences including condominiums, townhouses, and owned apartments. Importantly, a foreclosure is recorded in the data only when the proceedings are completed and the home is actually recaptured by the lender. Thus, the data do not include homes where foreclosure filings were initiated but the process was not completed because the borrower caught up with mortgage payments. We rescale the number of foreclosures by the number of housing units in the zip code, as measured by the 2010 census of population and housing. In all of our analysis we therefore measures foreclosures as the total number of foreclosures per 1,000 housing units in the zip code.

Figure 1 shows the evolution of foreclosures over time and relate those waves to the location of a large representative retail chain in California that will be used in future studies that combine the foreclosure to the retail scanner data. The foreclosure crisis picked up speed in 2007, but there is clear variation across locations in terms of the intensity of the crisis. Monthly foreclosures per 1,000 households peaked at around 4 in early 2008 in zip codes that were most severely impacted, or the orange line in the graph. In contrast, areas that were less affected by the crisis had similar levels of foreclosures up until 2007 but did not experience such large shocks during the height of the crisis. Most importantly, there is a notable amount of variation across space in the intensity of the foreclosure crisis.

We demonstrate the spatial and temporal variation in foreclosures across zip codes in a map in Figure 2.

The map shows the change in the number of foreclosures per 1,000 households in each zip code from 2005-2010. As is well known, the California foreclosure crisis clearly began in some areas in 2007, while in many areas the largest year-on-year increases in foreclosures took place in 2008. In many zip codes the change in foreclosures from 2007-2008 represents more than 1.5% of households (the darkest blue areas in the map). While there is substantial variation across zip codes, the California counties of Kern, Riverside, San Bernardino, and San Joaquin were heavily impacted. These constructed data exhibits promising variation to be used in future studies that combine foreclosure data to consumer level changes in expenditures for grocery items. In particular, future work uses this variation by comparing changes over time in areas that were heavily affected by the foreclosure crisis with changes in areas that were less affected. We will also combine these foreclosure data with tax records at the county level for a variety of taxable product categories, ranging from automobile, new and used, to clothes items to investigate the impact of the income shocks due to foreclosures on the allocation of consumer expenditures across categories.

Conservation organizations seeking to reduce over-fishing and promote better fishing practices have increasingly turned to market-based mechanisms such as environmental sustainability labels (eco-labels) in order to shift patterns of household consumption. The real-world evidence-increasing per capita seafood consumption and continued decline in the size of some of the very fisheries that have been certified–suggests that these mechanisms may be falling catastrophically short of their objectives. This Giannini-funded project explores this apparent paradox with an empirical analysis of consumer response to an advisory for sustainable sea food adopted by a supermarket in the United States. The advisory consisted of a label in which one of the three "traffic light" colors was placed on the pin tag of every fresh seafood product to inform consumers about the relative environmental sustainabil ity of that product. Analogous to the food labeling system currently being proposed in the European Union, green meant "best" choice, yellow meant "proceed with caution", and red meant "worst choice." Using a unique product-level panel scanner data set of weekly sales data, we apply a difference-in-differences identification strategy to estimate the effect of the advisory on seafood sales. We find some evidence that the advisory led to a slight decline in overall seafood sales (significant -15.5%). We find strong evidence that the sale of yellow-labeled seafood significantly decreased (-31.3 to -34.9%, significantly). We fail to reject the null hypothesis of no change for green and red sales. Furthermore, yellow products on a mercury safe list had the largest drop in sales (-41.3%). These results suggest that we need a much better understanding of whether and when eco-labels achieve their goals before continuing to make large investments in household consumers as a primary conservation tool. Co-author: Dr. Eric Hallstein. The paper is under revision at the Journal of Environmental Economics and Management, second round.

This Giannini funded project investigates whether information costs prevent consumers from making healthier food choices under currently regulated nutritional labels in a market-level expe riment. Implemented nutritional shelf labels reduce information costs by either repeating information available on the Nutritional Facts Panel, or providing information in a new format. We analyze microwave popcorn purchases using weekly store-level scanner data from both treatment and control stores in a difference-in-differences and synthetic control method approach. Our results suggest that information costs affect consumer purchase decisions. In particular, no trans fat labels significantly increase sales, even though this information is already available on the package. Low calorie labels significantly increase sales, while correlated low fat labels significantly decrease sales, suggesting that labeling response may also be influenced by consumers' tast e perceptions. Finally, combining multiple claims in a single label reduces the effectiveness of the implemented labels. Our results provide direct implications for changes to the format and content of nutritional labeling currently considered by the Food and Drug Administration. Co author is Professor Kristin Kiesel. The paper has been published in the International Journal of Industrial Organization.

There exists a large literature that analyzes the impact of expert opinion on consumer choice and demand for experience goods. Due to the non-experimental nature of most prior studies, endogeneity is a concern. Specifically, there is likely a spurious correlation between good reviews and high product demand that exists because of an underlying correlation with unobservable quality signals. This paper avoids such obstacles through an experiment approach.

Specifically, this study examines the impact of expert opinion on retail wine purchases. Anecdotal evidence suggests that consumers are largely uninformed regarding the quality of wine products: thus, wine is an experience good. There exist many sources of expert opinion for wine, including Wine Spectator and the Wine Advocate, which have reviews for more than 300,000 wines. However, little is known regarding the impact of such expert reviews.

To examine the impact of expert opinion, we utilize an experimental approach at a large national retail grocery chain. In particular, wines in two retail stores in Northern California were randomly chosen to display wine scores from the Wine Spectator and Wine Advocate. Wine opinion labels were displayed for one month during Spring 2006 for 200 wines. Wine sales in the treated stores were then compared to sales of these wines in stores that displayed similar trends in wine sales both before and after the
treatment period.

We find that while demand decreases for wines that receive low scores, demand for average- and higher-than-average-scored wines increases. The results indicate that expert opinion labels transmit quality information that affects demand as opposed to solely increasing the wine's shelf visibility to the consumer.

These results have important implications for understanding how to effectively market wine and other consumer products reach and influence consumer behavior. Specifically, our results demonstrate that expert opinion labels can substantially increase wine sales. More broadly, for products where consumers' product knowledge is incomplete, the display of quality information may induce consumers to enter the market or to purchase such products on a more regular basis.