I will investigate the effects of pesticide usage on children’s health and determine if there is any noticeable between children’s health, pesticide usage, and income. There is a long history of pesticide exposure being linked to a variety of human health issues. Many types of cancer such as Non-Hodgkin’s lymphoma and leukemia are associated with pesticide exposure (Alavanja et al. 2004). Experts have discovered possible links between pesticides and developmental disorders such as cerebral palsy and the autism spectrum (Goldman et al. 2000). Children in particular are susceptible to negative health impacts of pesticides during their developmental stages and are found to have higher exposure rates than older residents (Eskenazi et al. 1999).

Question and Problem

I will be addressing the question: “What is the correlation between pesticide usage, children’s chronic health problems and income?”

Location and Grain

My analyses are focused on the state of California at the County level.


The data was gathered from the California Department of Pesticide Regulation, the State of California Franchise Tax Board, and KidsData (a Program of the Lucile Packard Foundation for Children’s Health). I analyzed my data in both qualitative and quantitative methods. I first created bar graphs that visualize the differences of median income level, pesticide usage, children’s cancer diagnoses and children’s asthma diagnoses in the California counties. I then used statistical modeling to identify correlation in a quantifiable manner.


When I analyzed the following graphs and tracked specific counties, I was unable to visually identify any strong correlations. I did note that it appeared counties with high incomes appeared to have less pounds of pesticides applied. Looking at the data in this manner was helpful to me because when I ran the correlation tests I already had a sense of what the results would be, and it would therefore be easier for me to know if I had made any mistakes with the code for the correlation tests.

income_counties$County <- factor(income_counties$County) %>%

colorCount = length(unique(income_counties$County))
getPalette = colorRampPalette(brewer.pal(11, "Spectral"))

ggplot(income_counties, aes(x=County, y=Median.Income)) + 
  geom_col(fill= getPalette(colorCount)) + coord_flip() + 
  ggtitle("Median Income  of California Counties") +
  labs(x= "California Counties", y= "Median Income in 2013") 

cancer_counties$County <- factor(cancer_counties$County) %>%

cancer_counties1 <- subset(cancer_counties, !

colorCount1 = length(unique(cancer_counties1$County))
getPalette1 = colorRampPalette(brewer.pal(5, "Set1"))

ggplot(cancer_counties1, aes(x=County, y=X2009.2013)) +
  geom_col(fill= getPalette1(colorCount1)) + coord_flip() +
    ggtitle("Rate of Child Cancer Diagnoses in California Counties") +
  labs(x= "California Counties", y= "Number of New Cancer Diagnoses per 100,000 Children for 2009-2013")

pest_counties$County <- factor(pest_counties$County) %>%

getPalette3 = colorRampPalette(brewer.pal(4, "Paired"))

ggplot(pest_counties, aes(x=County, y=X2015.Pounds.Applied)) + 
  geom_col(fill =getPalette3(colorCount)) + coord_flip() +
  ggtitle("Pesticide Usage of California Counties") +
  labs(x= "California Counties", y= "Pounds of Pesticides Applied in 2015") +
  scale_y_continuous(labels = comma)

asthma_counties$County <- factor(asthma_counties$County) %>%

getPalette4 = colorRampPalette(brewer.pal(5, "Pastel1"))

ggplot(asthma_counties, aes(x=County, y=X2013.2014)) + 
  geom_col(fill= getPalette4(colorCount)) + coord_flip() +
  ggtitle("Percent of Children with New Asthma Diagnoses in California Counties") +
  labs(x= "California Counties", y= "Percent of Children Diagnosed with Asthma in 2013-2014")

Next, I ran correlations tests to determine whether or not there were any positive or negative relationships between the data. With the Spearman method of correlation, a -1 indicates a strong negative relationship, a +1 indicates a strong positive relationship, and a 0 indicates that there is no correlation.

income_asthma <- cbind(income_counties, asthma_counties) 

ia_results <- with(income_asthma, 
  cor(Rank.asthma, Rank.income,
    use = "pairwise.complete.obs",
    method = "spearman")

## [1] -0.178769

The result is near 0, so we can conclude there is no significant correlation.

income_cancer <- cbind(income_counties, cancer_counties) 

ic_results <- with(income_cancer, 
  cor(Rank.cancer, Rank.income,
    use = "pairwise.complete.obs",
    method = "spearman")

## [1] 0.3628602

The result is greater than 0 but not high enough to indicate a strong positive correlation.

income_pest <- cbind(income_counties, pest_counties)

ip_results <- with(income_pest, 
  cor(Rank.pest, Rank.income,
    use = "pairwise.complete.obs",
    method = "spearman")

## [1] -0.5175958

The result is closer to -1 than it is to 0 so we can conclude there is some correlation between income and pesticide usage. This is understandable because I would expect counties with higher median incomes to be more urban than rural and therefore have less land on which pesticides would be needed.

pest_asthma <- cbind(pest_counties, asthma_counties)

pa_results <- with(pest_asthma, 
  cor(Rank.pest, Rank.asthma,
    use = "pairwise.complete.obs",
    method = "spearman")

## [1] 0.08572549

The result is almost exactly 0, so we can conclude there is no identifiable relationship.

pest_cancer <- cbind(pest_counties, cancer_counties)

pc_results <- with(pest_cancer, 
  cor(Rank.pest, Rank.cancer,
    use = "pairwise.complete.obs",
    method = "spearman")

## [1] -0.3539666

The result is less than 0 but not close enough to -1 to indicate a strong negative correlation.

asthma_cancer <- cbind(asthma_counties, cancer_counties)

ac_results <- with(asthma_cancer, 
  cor(Rank.asthma, Rank.cancer,
    use = "pairwise.complete.obs",
    method = "spearman")

## [1] -0.123814

The result is fairly close to 0, so we can conclude there is no identifiable correlation.

Main Conclusions

The main conclusion from my work is that there is a lack of strong evidence to support linkage between pesticide usage and children’s cancer and asthma diagnoses at the county level in California. My project did, however, identify a negative correlation between median income and pesticide usage. This means that areas with higher incomes had lower amounts of pesticide usage and vice versa. This conclusion could be helpful for further research into my main question because now that I know low income areas have higher amounts of pesticide usage, I would hone in my focus specifically on the health problems in these lower income areas. If I were to continue this research, I would also consider looking at other health problems and choosing to widen the grain of my analyses to the entirety of the United States. Overall, I believe the results of my work show that research on the health effects of pesticides is still of high importance.


Alavanja, Michael C.R., et al. “Health Effects of Chronic Pesticide Exposure: Cancer and Neurotoxicity.” Annual Reviews, Annual Review of Public Health, 21 Apr. 2004,

California Department of Pesticide Regulation. (2016). Total pounds of pesticide active ingredients reported in each county and rank during 2014 and 2015 [PDF file]. Retrieved from

California Franchise Tax Board (2018). California median income by county [Map with data exportable as a CSV file]. Retrieved from

Eskenazi, B, et al. “Exposures of Children to Organophosphate Pesticides and Their Potential Adverse Health Effects.” Environmental Health Perspectives, U.S. National Library of Medicine, June 1999,

Goldman, L R, and S Koduru. “Chemicals in the Environment and Developmental Toxicity to Children: a Public Health and Policy Perspective.” Environmental Health Perspectives, U.S. National Library of Medicine, June 2000,

Lucile Packard Foundation for