Public Acceptance of and Willingness to Pay for Mosquito Control, Texas, USA

Mosquito control is essential to reduce vectorborne disease risk. We surveyed residents in Harris, Tarrant, and Hidalgo Counties, Texas, USA, to estimate willingness-to-pay for mosquito control and acceptance of control methods. Results show an unmet demand for expanded mosquito control that could be funded through local taxes or fees.

effective mosquito control methods, a better understanding of what motivates public opinion is needed. In the Key West example, survey results showed that people opposed to the GM mosquito trials tended to cite concerns about disturbing ecosystems and using unproven technologies (8,9) whereas some who supported the trial said they thought this method was more natural and less harmful than chemical-based controls (8). Perceptions of the risks that mosquitoes pose, and their general nuisance level, can also be important motivators of support for control efforts. A survey of Madison, Wisconsin residents showed that dislike for nuisance mosquitoes was a stronger driver of willingness-to-pay for mosquito control than concerns about disease (12). Concerns about the environmental impacts of mosquito control were found to decrease willingness to pay for mosquito control in Key West, Florida (13). Recent studies have also found that residents would be willingness to pay more to expand mosquito control programs in New Jersey (14) and North Carolina (15).

This study protocol was reviewed and approved by the Colorado Multiple Institutional
Review Board (COMIRB) (Protocol #: 18-0348, approved March 2, 2018) and the Texas A&M University Institutional Review Board determined the proposed activity was not research involving human subjects on July 2, 2018 (Protocol #: 2018-0774). Participants provided written consent to take the survey.

Survey Design and Administration
To inform the design of our survey, we interviewed key stakeholders involved in mosquito control at state level and in Harris, Tarrant, and Hidalgo counties in Texas (Appendix 1 Figure 1).
Using findings from these interviews, we developed a survey to address the following: 1) household characteristics, 2) perceived impacts of mosquitoes on quality of life, 3) vectorborne disease knowledge and concerns, 4) mosquito avoidance and control behaviors, 5) willingnessto-pay for expanded mosquito control, and 6) opinions on specific control methods (e.g., adulticides, traps). The full survey is available below. After a pilot study to recruit by mail that resulted in a low response rate (~10%), we contracted Qualtrics, a software and analytics company, to recruit survey participants. Qualtrics uses multiple actively managed panels of survey participants recruited for academic and market research. While relying on Qualtrics panels limited our ability to recruit at random from specific geographic areas within each county, it did allow us to set quotas for gender, race/ethnicity, and median household income to obtain samples that largely reflect the demographic makeup of each county. The survey was administered to a total of 1,831 persons during October 24-November 15, 2019. Participants could choose to take the survey in English or Spanish.

Measuring Willingness to Pay
Appendix 1 Figure 2 depicts the design of our triple-bounded WTP question format. A key benefit of this approach over alternative dichotomous choice contingent valuation question formats is that it gathers more information from each respondent, allowing us to estimate WTP more precisely for a given sample size (16). Our approach is largely consistent with recommended best practices in WTP measurement (e.g., the NOAA Blue Ribbon panel report (17)). For example, we use a referendum format (yes or no vote on a proposed ballot measure), allow a "don't know" option, emphasize that all answers are reasonable and valid, and remind the respondent that money used toward mosquito control would not be available for other uses. A sample question is below: Suppose that there were a proposal on the next election ballot to expand mosquito control across the county. If the proposal passes, the number of mosquitoes in this area would be cut in half. To fund this expansion, your household and others in the county at your income level would be charged a fee of $XX once per year.
Would you support this proposal?
The fee amount for this initial question was randomized across respondents ($5, $10, $25, $50, or $100). Participants who said yes to the initial fee were then asked if they would support a fee of twice that amount. If they said yes to that, the fee doubled again. Participants who said no to the initial fee were asked if they would support a fee that was half the initial amount. If they said no again, the fee was halved again (Appendix 1 Figure 2).
Resulting data include upper and lower bounds on WTP for each participant. The exceptions are for participants that answered "Yes" to all questions (in this case, we do not know the upper bound on their WTP) or "No" to all questions (in this case, we do not know the lower bound because WTP can be negative, meaning participants would need to be compensated to accept increased levels of mosquito control).

Respondent Characteristics
Participant characteristics and corresponding census data for a subset of variables are presented by county in Appendix 1 Table 1. Overall, the average participant was 41 years of age and had lived in their county for ≈22.5 years. A little over half of participants were women.
Approximately 37% of respondents identified as Hispanic or Latino/a/x, and 56% identified as a race other than White, with substantial racial and ethnic variation across counties. About a quarter of participants had a high school education or less, while 37% had a bachelor's degree or higher. Regarding income, ≈40% of respondents earned <$40,000 per year, whereas <10% earned >$150,000. A little less than half of participants had children.
To analyze relationships between WTP and respondent characteristics within each county, we use an interval censored regression model in which the dependent variable is the respondent's WTP range and independent variables were gender, race/ethnicity, age, education, income, political ideology, whether or not the respondent knew someone who had had WNV, dengue, or Zika, whether the respondent noticed many mosquitoes outdoors at the time of the survey, and county.
To analyze variation in support for each control method by sociodemographic characteristics, we used ordered logistic regression models for each of the 6 control methods. In these analyses, the dependent variable was the 5-point Likert scale response to the level of support for the method in question and independent variables were gender, race/ethnicity, age, education, income, political ideology, whether or not the respondent knew someone who had had WNV, dengue, or Zika, whether the respondent noticed many mosquitoes outdoors at the time of the survey, and county (Appendix 1 Table 3; Appendix 1 Figure 3).

Discussion
As public health risks from vector-borne disease continue to evolve in the U.S. and around the world, understanding public attitudes and support for control programs can help guide effective policies and interventions. Results from our survey offer key insights into the public's willingness to pay for expanded mosquito control in Texas, as well as their support for a range of different control measures. Across all three counties surveyed, we found that residents were willing to pay much more for mosquito control than is currently allotted per capita in each of their respective county budgets. On average, participants were willing to pay an additional In this vein, a 2017 bill introduced in Texas (S.B. 1695) sought to establish mosquito control districts for counties on the Mexico border that either 1) had experienced >1 locallytransmitted case of Zika, or 2) were located adjacent to a county that had experienced such a case (18). These districts would conduct vector surveillance, support county public health work on communicable diseases, and educate the community about vectorborne disease prevention. The bill died in committee, but our results suggest that public opinion is in favor of similar efforts.
In assessing options, Texas officials may find it helpful to examine programs in other states. There is substantial heterogeneity in the scale, scope, and funding of organized vector control activities at the national level (19). Legislation varies across states and counties, and policies can either facilitate or prevent revenue collection through property taxes, services charges, or other contracts. California and Florida allow for abatement districts, mostly at the county level, with some county vector control programs having annual budgets of >$10 million per year (20). As an example, the Orange County Mosquito and Vector Control District in California has an annual budget of $15,651,880 and a population of 3,190,000, equating to $4.90 per person per year. This is significantly higher than the funding levels in the Texas counties we surveyed, but still lower than our estimated WTP. Property taxes derived from millage rates are not the only revenue source for funding local mosquito and vector control programs. For example, Illinois and Florida both collect fees for tire disposal that fund vector surveillance, control, and research activities (21)(22)(23).
In terms of control methods, we found that participants were most supportive of lethal traps as a form of mosquito control, which involve neither chemicals nor the release of additional mosquitoes. Ongoing research is shedding light on the effectiveness and feasibility of scaled up mosquito control using such traps (24). Participants were also in support of traditional methods of mosquito control (adulticides and larvicides). Consistent with prior work in Key West, Florida (9), participants were somewhat more skeptical of methods involving the release of genetically modified mosquitoes or mosquitoes infected with Wolbachia bacteria. However, more respondents were in favor of these methods than those in opposition. With intentional community engagement, education, and outreach, these methods may serve as an effective tool in control programs going forward.
Vector control programs have historically been either top-down, bottom-up, or a combination (25)(26)(27). Top-down approaches led to effective control of Ae. aegypti over much of the Americas in the 1950s, but political instability and a lack of sustained efforts caused these programs to fail over time (1,26,28). Currently, many area-wide operational vector control programs are organized at the city or county level. Some counties house multiple mosquito control districts (e.g., Cook Co., Illinois) while others have multi-county programs (e.g., the Metropolitan Mosquito Control District in St. Paul, Minnesota). In some instances, states take on vector control activities: in addition to performing vector-borne disease surveillance, Arizona's state health department runs annual vector control workshops (29). In light of the strain that the COVID-19 pandemic is putting on health departments, vector control programs around the country have faced cutbacks, and recently, there has been a call for a national vector surveillance program (30)(31)(32). A hybrid approach including top-down and bottom-up elements will likely be advantageous. Critically, the funding for this program needs to have bottom-up support through tax-based revenue aligned with the willingness and preferences of local citizens. Support from local communities is key to ensure feasibility and cooperation.
This study is limited in scope to three counties in the large state of Texas, and we cannot generalize our results beyond these counties. While our sampling strategy created a sample that is largely representative of the populations of the selected counties in terms of observable characteristics, the fact that respondents were selected from survey panels recruited for marketing purposes likely results in some unobserved differences between this sample and the general population (e.g., more internet-savvy persons). In addition, stated preference methods such as the WTP protocol we employed here are subject to hypothetical bias and may result in overestimates of true WTP.