During the late 1970s, subsurface drip irrigation was the latest technology to improve water application and reduce water waste. In her empirical analysis of subsurface drip adoption amongst perennial crop growers in the San Joaquin Valley, Caswell finds that water cost and farm type are significant in predicting the likelihood of adopting drip relative to traditional irrigation technology . Caswell and Zilberman extend this analysis to include sprinkler irrigation, and confirm the previous results that water cost and farm type are significant in increasing the likelihood of adopting more efficient relative to traditional technology. Caswell and Zilberman also find that water source, specifically groundwater relative to surface water, increases the likelihood of adopting either sprinkler or drip irrigation. At the time of their analysis, many water districts did not have the appropriate infrastructure to deliver pressurized water needed for the newer technologies. Green et al. compare three technologies in their survey of San Joaquin Valley: two mature technologies and a new one . They find that the adoption of high-pressure sprinkler has similar characteristics to that of furrow, suggesting that the former may also be nearing the end of its product life cycle. For example, it is statistically unlikely that higher water costs would lead to adopting high-pressure sprinkler. As expected, crop choice positively and significantly influences the decision to adopt drip, with the adoption likelihood higher in citrus, deciduous fruit, and vineyards relative to truck crops. Both an increase in slope and soil quality positively and significantly impact adoption of drip. Water price is an important metric and, indeed, policy instrument with respect to irrigation technology adoption, but it tends to capture more than drought-related scarcity . When measuring the extent to which such adoption represents adaptation to climate change, it is thus important to control for climate, as done in subsequent studies. Dinar, Campbell, and Zilberman also find that water cost and farm type increase the likelihood of adopting modern technology at both the farm and field levels. In addition, they find that acreage has a positive and significant impact at the field level,grow shelving indicating that economies of scale exist at the field level.
They also include climate variables in their analysis. Schuck et al. also study irrigation technology adoption as an adaptation to climate change. While they do not include climate variables in their analysis, they evaluate irrigation technology adoption amongst Colorado growers immediately following a historic drought. They find that growers who switch irrigation systems following the drought were more likely to switch from gravity to gated pipe rather than gravity to sprinkler. This is an example of how growers tend to minimize the cost of transitioning to a new technology rather than maximize benefits from saving the most water possible by adopting the most efficient technology. There is also an income effect to this finding, as those who adopted gated pipe instead of sprinkler tended to have lower income. Schuck et al. also found that level of education had a positive and significant impact on adoption of sprinkler during the drought. Leasing land had a negative and significant impact on adoption of sprinkler. Mendelsohn and Dinar run individual logistic regressions on percent cropland irrigated by a given technology across a large sample of US counties. They find that the likelihood of adopting gravity and drip irrigation is significant with high temperatures, while there is a significant negative relationship with sprinkler systems and high temperatures. Higher precipitation increases the likelihood of adopting sprinkler, but at a declining rate. Increasing allocation of surface water also increases the likelihood of adopting sprinkler irrigation, while higher soil salinity levels increase the likelihood of adopting drip systems. At the nation-state scale, Su and Moaniba find that innovations in climate-technology may be driven by anthropogenic climate change. Specifically, they find that the number of climate-patents is positively associated with GHG emissions from certain fuel sources. Studies also find a significant relationship with adoption of efficient irrigation technologies and sources of information, particularly information from agricultural extension. Escalera, Dinar, and Crowley study the adoption of a broad range of soil monitoring technologies among California avocado growers.
They find that the likelihood of adopting tensiometers to monitor soil moisture is positively and significantly related to receiving farming information from University of California Cooperative Extension. Genius et al. study adoption of drip or sprinkler amongst olive growers in Crete. They find that extension services and social networks significantly increase the rate at which either of these technologies are adopted. In addition, measures of human capital also have a significant impact on increasing the adoption rate. As with the soil and salinity monitoring technologies, we created categories for primary threats to water scarcity and primary source of information from extension experts, previous surveys, and our pilot survey. We found that population growth and drought are the top concerns amongst growers in our sample. Both population growth and drought are irreversible trends that affect the sustainability of agriculture in the region. These may incentivize growers to improve productivity through intensive margin improvements such as soil moisture and salinity monitoring. Intensive margin adaptations are those that improve the efficiency per unit area compared to extensive margin adaptations that improve efficiency of the total planted area. We also found that industry and social networks are the primary sources of information.Based on Chi-Square tests on the Wald statistics, the fit for both the soil moisture and salinity models is significantly different from the respective null models. The McFadden R2 for both models is somewhat below the standard acceptable range of [0.20, 0.40] . We use a natural log transformation on the acreage variable due to the large range in values [0.25, 10625], and find that the transformed variable is statistically significant. A 2.7-fold increase in acreage increases the likelihood of adopting soil moisture and salinity monitoring by 40% and 45%, respectively. This is consistent with our hypothesis that farms with more acres will, on average, be more likely to adopt monitoring practices. The “acreage effect” may be more relevant for small- and medium-sized farms since it takes an almost 3-fold increase to increase the likelihood of adoption.
Indeed very large farms also tend to derive 75-100% of their income from farming activities, and we do not find a significant relationship with monitoring activities in these farms. Farms with 50-74% income from agriculture are 4.2 times more likely to adopt soil moisture monitoring than those with less than 25% agricultural income. And, farms with 25-49% income from agriculture are 2.6 times more likely to adopt salinity monitoring relative to those with less than 25% agricultural income.38 Farm type is a significant predictor of adopting at least one soil monitoring practice. Relative to orchards , vegetable and field crop farms are 84% and 76% less likely to adopt soil moisture monitoring. Additionally, growers who primarily receive their information from state and federal institutions are roughly 3 times more likely to adopt either monitoring practice than those who receive from other sources . Contrary to our hypothesis, poor soil moisture quality is not a significant predictor of soil moisture monitoring. This may suggest that the underlying soil quality has little to do with actual moisture retention due to the large quantity of amendments growers add to soils. Access to groundwater, number of water sources,grow benches or type of water source do not significantly predict adopting either monitoring practice. We did not find a significant relationship between the level of total dissolved solids in irrigation water and adoption of salinity monitoring. This is contrary to our hypothesis that growers with higher salinity levels would be more likely to adopt salinity monitoring. However, when we included a variable on salt tolerance of the plant, we discovered why we may have observed this contradiction. Growers are concerned about the specific tolerance of their crops rather than broadly concerned about salinity levels. We found that growers with sensitive crops were almost 3 times more likely to adopt salinity monitoring than growers with moderate crops.Perhaps the most counter-intuitive result from our analysis is that water price is not a significant predictor of adopting at least one of these practices, which is contrary to the empirical evidence on irrigation technology adoption in the literature. However, when we included a variable on the frequency with which the water price has increased over the past decade, we found that a one-unit increase in frequency of rate increase , a grower is 9% and 16% more likely to, respectively, adopt soil moisture and salinity monitoring. We do not observe a significant relationship with any of the climate normals, but find a significant relationship with short-run total annual precipitation. The results in Tables 5.1 and 5.2 suggest that a positive increase in the precipitation mean results in positive likelihood of adopting either monitoring practice.40 Additionally, the coefficient of variation across 5 years of precipitation also positively influences adoption of either monitoring practice.
Perhaps education is a more important variable in developing countries where it is equated with access to information. Experience may be reflected across different time horizons based on farm type, i.e., it may take longer to acquire experience for more complex farm systems with multiple crops. Additionally, perceptions of water scarcity do not significantly impact the likelihood of adopting either of these practices. This may suggest that growers are monitoring soil moisture and salinity to improve crop health rather than optimize water use. Due to the issue of endogenous regressors, we could not test our theory of “bundling” both monitoring practices with binary logistic regression. The multinomial logistic regression is more informative in identifying factors that influence adoption of both types of monitoring practices . We find that percent of income derived from agriculture, farm type, and using government institutions as the primary source of information may increase the likelihood of jointly adopting both monitoring practices. We use “both practices” as the baseline in order to compare this to the implementation of “soil only” or “salt only”. Due to the overlap in confidence intervals of these variables across the different categories, we cannot distinguish the effects of a single category relative to the baseline. A 2.7-fold increase in the 5-year precipitation mean results in roughly a 64% decline in likelihood that neither practice is being implemented, or that only soil moisture monitoring is implemented survey Murray Darling Basin farmer attitudes to having their children take over farming operations following the historic drought. They find that farmers who plan to have their children inherit their farm are more likely to have made irrigation efficiency improvements and less likely to have sold any land in the prior 5 years. Zollinger and Krannich survey Utah growers to determine the factors influencing their expectation to sell land for non-agricultural uses. They find that increased profitability over the past five years has a significant negative influence on the expectation to sell land, while the perception of increased urbanization exerts a significant positive influence. Deschenes and Kolstad study how weather and expectations on weather influence farmland productivity in California across a 20-year period. They assume that such expectations are derived from observing past weather, and thus include a 5-year moving average in their time-series model. Although none of their weather variables are significant, their study provides general intuition on the magnitude of these variables. The magnitude of the expected degree-day variable is larger than the annual average, suggesting that changes in expectation are more costly than annual weather changes. This section has highlighted micro-level analyses of likelihood of selling farmland. These studies differ from the subsequent analysis on likelihood of land sales because these do not include climatic impacts. Additionally, these focus on agricultural land sales for non-agricultural uses whereas our study focuses on agriculture-to agriculture land sales. This section also highlights land value studies that focus on short run fluctuations in weather, which we build upon for the subsequent Ricardian analysis. More detailed review of the Ricardian literature is presented in Chapter 4.We test the impact of 5 and 10 year lags of weather on both the likelihood of sale and value of farmland. We focus on the precipitation mean, as this has the greater contribution to drought relative to temperature .