The next four columns vary that distance from 30-70 km in 10-km increments. As can be seen in Panel A, the impact of an additional fire is considerably larger when we focus on nearer fires, but this pattern of results no longer holds when we standardize our outcome measure based on the variability of test scores, as in Panel B. Unsurprisingly, the results become smaller as we include test takers further away from the fire. At a 70-km radius, as seen in column of Table 4, the results are no longer significant. Together, these results highlight the relatively localized impacts of agricultural fires. In columns – of Table 4, we explore the sensitivity of our results to alternative central angle measures to determine whether an individual is upwind or downwind of a fire. Recall that our baseline model specification uses the angle of 45 degrees to define upwind and downwind fires . As we alter the angle to 30, 60, and 90 degrees, the estimates remain significant, but become smaller as the angles become larger. This pattern of results is consistent with standard models of pollution dispersion,raspberry grow in pots as wider angles will expand the ‘treated’ upwind sample to include more individuals with peripheral levels of exposure. It also further validates that our upwind and downwind measures are doing a reasonable job of capturing the relevant transport of pollution from fires to test centers.
Table 5 experiments with alternative ways to define a fire. Column reproduces our core results from Table 2, while column takes a more aggressive approach to classifying fires as exogenous by limiting our attention to those fires within the 50-km radius of a county administrative center but that take place in a different county. While our use of wind direction is meant to capture the economic effects from agricultural fires, the enforcement of any policies designed to limit agricultural fires or protect air quality occurs primarily at the county level . Thus, our focus on non-local fires should help address any potential concerns about the endogeneity of local policies vis-à-vis testing outcomes. The results using this specification are largely unchanged. 15 In column , we inverse-distance weight fires to better reflect the distance of the fire from the county administrative center. In column , we account for the intensity of the fire by weighting by the fire radiative power in Watts of each event. The estimates remain statistically significant, but are slightly smaller in magnitude than those under our preferred specification. Finally, we use reliability measures from the fire dataset to adjust for the probability that a hotspot is genuinely a fire . The results after this adjustment are statistically significant and slightly larger in magnitude. In Table 6, we explore a final set of robustness checks. As before, the first column reproduces our core results for ease of comparability. We report the estimates using alternative ways of clustering standard errors either by prefecture in column , or by county and by year in column . The estimates are robust to these different clustering approaches, suggesting that spatial and temporal auto correlation is not a big concern in our setting. In column , we add controls for visibility.
These controls are important as impaired visibility may trigger avoidance behavior in the lead up to the exam.16 In addition, gray skies can impair one’s sense of psychological well-being, particularly if worried that diminished air quality might affect their test performance. In column , we expand our focus in Shandong to the third day, which only takes place in this province. In column , we add the data we have from Jiangsu Province, which only covers part of our study period. The coefficients barely budge across the first three checks. The results are slightly smaller and now only significant at the 10-percent level under the final one. In the end, our results appear quite robust to alternative methods of measuring fires, assigning exposure, clustering standard errors, and defining our sample population. That the magnitudes of results change in expected directions as we tighten or liberalize the approach we use to assign fires to testing facilities is particularly reassuring. In this section, we estimate the effect of agricultural fires on air pollution, to confirm that air pollution is the channel through which agricultural fires affect students’ exam scores and to place our results in a broader context. As described earlier, we do so by using data from the 2013–2016 period for which daily air pollution measurements, even in more rural areas, are available.
The ideal design for this analysis would focus exclusively on the two-day exam period, but this leaves us with limited statistical power. Instead, we construct a panel of two-day moving averages of pollutant concentrations in June and link them with proximate agricultural fires during the same period. The empirical model for this estimation is nearly identical to the one described in Equation , except that the dependent variable is now one of the six criteria air pollutants. Weather variables are now measured as two-day averages of the corresponding to each moving two-day period in June for which we have pollution measures. The results are shown in Table 7. The first two rows list the two-day averages and standard deviations of each pollutant in June during 2013–2016. The PM10 concentration is approximately 78 µg/m3 and the PM2.5 concentration is approximately 46 µg/m3 , both of which greatly exceed World Health Organization guidelines. The other pollutant levels are more modest, although still higher than those typically found in developed countries. Turning to our estimates, we find a significant and substantial effect of upwind agricultural fires on PM10 and PM2.5. A one-point increase in upwind agricultural fires increases PM10 and PM2.5 concentrations by 0.476 µg/m3 and 0.262 µg/m3 , respectively. We also detect a weak effect of downwind fires on PM10, and the coefficient of upwind-downwind difference becomes insignificant compared with that of PM2.5. This may be due to the fact that PM10 is heavier than PM2.5 and thus less responsive to wind direction. The impacts on PM2.5 are non-trivial: a one-standard-deviation change in the upwind-downwind difference is associated with a 5.6 percent standard-deviation change in PM2.5. In contrast, downwind fires have no impacts on air quality, providing further validation for our empirical strategy to uncover the pollution-driven impacts of agricultural fires on NCEE test performance. We find no effect of agricultural fires on other pollutants, including SO2, NO2, CO, and O3.
In general, these estimates are consistent with those found in the scientific literature and recent empirical analysis done by Rangel and Vogl in Brazil, both of which find that agricultural fires primarily emits PM. Given that the samples are different for our estimates of the impacts of fires on pollution and the impacts of fires on test performance, we are unable to provide an instrumental variable estimate of the effect of PM on student scores. We provide a rough estimate akin to Wald estimator as an alternative. Using the ratio of the reduced-form estimates over the first-stage estimates based on the differences in upwind and downwind fires, we find that a one-standard-deviation elevation in PM2 5 will lower average student scores by 13.6 percent of a standard deviation . While these magnitudes are quite modest, they are roughly three times as large as those found for the impact of PM on Israeli test takers . A simple transformation further shows that a 10 µg/m3 increase in PM2.5 reduces test scores by 4.6 percent of a standard deviation, which is larger than the 1.7 percent estimated from Ebenstein et al. . This likely reflects the higher levels of pollution in our setting,square plastic pots but may also be the result of our empirical strategy which relies on wind direction rather than an approach that assigns pollution equally to all of those within a certain distance of a pollution monitor. In addition, our estimates are also larger than those estimated for temperature . That said, our estimates here should be treated with some caution, as our ‘two-stage approach’ relies on data from adjacent but distinct time periods. California agriculture in 2004 is a very different industry than it was in 1950, 1850, or, for that matter, at its beginning in the late 18th Century. California, until well into the 18th century, was one of the few remaining “hunter-gatherer” societies left in the world . The origins of sedentary California agriculture began with the development of Spanish missions over the period 1769–1823. Over its brief history of 250 years, the character of California agriculture has been in a perpetual state of transition and adjustment: from early mission attempts to raise livestock, grow grains, and develop horticulture; to the era of ruminants ; to the development of large-scale, extensive wheat and barley production; to the beginnings of intensive fruit, nut, and vegetable agriculture based on ditch irrigation and groundwater; to pioneering large-scale beef feedlot and dairy production; to the intensified and expanded production of an increasingly diverse portfolio of crops resulting from massive public irrigation schemes; to today’s highly sophisticated, technologically advanced, management-intensive agricultural industry, which is embedded in a rich, urban state of 35 million people. It is a history of perpetual, profound, and often painful change. The turn of the millennium was marked by hard times in California agriculture: low prices seemingly across the board, water-supply woes, contracting growth in export markets, more stringent environmental regulations, and declining farm income. What does the future hold for California agriculture? Is it as bleak as it sounds? California agriculture has experienced recurrent challenges over its history and has survived. Can it do so again? This report is a modest attempt to throw some light on these questions. In this introduction we first review the situation in 2000–2002 to identify an assortment of “turn of the century” problems confronting California agriculture. In Chapter II we then place these symptoms/ indicators in a historical context in a stylized, epochal history of California agriculture circa 1769 to the present.
Chapter III is a more detailed examination of major structural shifts from 1950 to 2000, providing a look at internal performance indicators as well as comparisons to the performance of U.S. agriculture. In Chapter IV we develop a list of major factors that have driven California agriculture from the early mission agricultural period through the 20th Century and then make our qualitative assessment of the importance of these “drivers” in the early 21st Century. Chapter V contains our thoughts about the future of California agriculture. Overall, California agriculture has always battled economic adversity. While blessing California with good weather and fertile soils, nature did not provide adequate rainfall in the right place or at the right time. The downside is that investments are needed to bring water to the soil to grow crops. The upside is that irrigation potentially allows watering crops at the precise time of need and in the correct amounts, greatly increasing the range of production options. Thus, water management is a critical additional dimension of complexity for California agriculture. California is a long distance from everywhere; therefore, importing and exporting have always been expensive in terms of both money and time. Finally, California, because of its subtropical, Mediterranean climate, has different and more complex problems with pests and diseases than does the rest of mainland agriculture. Yet, since the 1850s, California agriculture has grown and adjusted many times. Each time, the composition and character of agriculture have changed, but the state’s overall industry, in terms of value and volume of output, has grown steadily and has always returned to profitability. Will California agriculture be unable to adjust and grow in the 21st Century? We cannot find evidence to support this proposition. In what follows we try to justify that conclusion.The 21st Century began with great uncertainty in the minds of many California farmers and ranchers.In the two sections that follow, and in no particular order of significance, we list some suggested indicators compiled from a variety of media.California has a highly diversified agriculture. Historically, when some prices have been low , fruit, nut, vegetable, or livestock prices were high, but this was not generally true in 2000–2001. Hence, the widespread concern following recognition that the last time everything was down was during the Great Depression of the 1930s.