Willingness to pay for a clear night sky: use of the contingent valuation method

Abstract

This article applies the Contingent Valuation Method (CVM) to the issue of night sky pollution. Light pollution decreases the ability to view a clear, unobstructed night sky. We administered a survey to the students of the Rochester Institute of Technology (RIT) to obtain estimates of Willingness To Pay (WTP) to improve night sky visibility and to prevent deterioration in visibility. This is the first CVM study that attempts to distinguish between these different WTPs. We find that students are willing to pay significantly more for a larger improvement in night sky conditions. We also find significant differences in WTP to improve versus prevent deterioration in night sky conditions.

Notes

¹We use the terms ‘sky glow’ and ‘light pollution’ interchangeably in this article. However, sky glow, which is the focus of this article, is just one type of light pollution that is a brightening of the sky caused by outdoor lighting and natural atmospheric and celestial factors. Other types of light pollution are glare, light trespass and spill light. For an explanation of these terms see McColgan (2003).

²In 1958 Flagstaff first addressed the issue at the request of astronomers at Lowell and US Naval Observatories by banning advertising searchlights used by retailers such as car dealerships (Hoversten, 1999). In addition, the state of Arizona has enacted light pollution mitigating legislation (Greeley, 1985; Arizona State Legislature, 2007). Alvarez del Castillo et al. (2003) write that, as of 2003, six states in the United States had state lighting ordinances.

³Light pollution is beginning to be discussed more in the popular media. The feature article of the National Geographic Magazine of November 2008 discussed this issue (Klinkenborg, 2008). The New York Times (Sharkey, 2008), The Wall Street Journal (Hotz, 2008) and The New Yorker (Owen, 2007) have all recently featured articles on this topic, reflecting a growing public awareness of the matter.

⁴The Rochester, NY metropolitan area has a population of 1 098 201 (2000). Rochester City has a population of 219 773 (2000) and a population density of 6 104.8 per square mile. RIT is located in the town of Henrietta, a suburb of Rochester City, with a population of 39 028 (2000) and a population density of 1 096 per square mile.

⁵See Hackl and Pruckner (1999) and Cameron and Huppert (1989) for further details.

⁶We carry out our statistical analysis using the Lifereg procedure of SAS (PROC LIFEREG, SAS version 9.1; SAS Institute, Inc., Cary, North Carolina). The parameters are estimated by maximum likelihood using a Newton–Raphson algorithm. The SEs are estimated from the inverse of the observed information matrix. See Maddala (1983), section 6.7, and Greene (2003), section 22.3 for a detailed explanation of the estimation method.

⁷160 questionnaires were completed. 21 of these had missing demographic data.

⁸The large percentage of males in the sample is due to the fact that almost 70% of all students enroled at RIT are male. The large percentage of freshmen and sophomores in the sample reflects the fact that three of the four classes that were surveyed were principles classes. All WTP and income figures are in US dollars.

Table 2. Explanatory variable descriptive statistics

Table 3. WTP percent frequency distribution

⁹Note that the WTP values that we use in this test are the midpoints of the WTP intervals. However, for the case where zero is circled on the payment card, a zero value is used.

¹⁰We ran a simple linear regression between mid-WTP2 and mid-WTP1 (with no intercept); the parameter estimate on the mid-WTP1 variable was 2.02 with SE equal to 0.1; thus, we cannot reject the hypothesis that the WTP for the large improvement is, on an average, twice the WTP for the large improvement.

¹¹The average WTPs are all significantly different from zero but the magnitudes seem relatively small. However, as a percentage of the sample average quarterly income ($1379.57), these are 1.04, 2.36 and 1.22% for WTP1, WTP2 and WTP3, respectively. Further, if administrators were able to impose the average WTP as an additional fee on each enroled student in the fall, winter and spring quarters, for an institution with 15 000 students, that would generate approximately $647 000, $1 466 000 and $756 000, annually for WTP1, WTP2 and WTP3, respectively.

¹²The survey also allowed us to gather information on whether the student had heard of light pollution or had experienced sky glow, how much time they spend outside on campus at night, and what their college major is. We experimented with including these in the analysis and found that their parameter estimates were statistically insignificant across all WTP equations and that their omission did not significantly affect our maximum likelihood estimates.

¹³We also experimented with logistic and normal distributions but found the log-normal distribution to fit the data relatively well.

¹⁴Since ln( y predicted) ≠ predicted ln( y), we have to adjust the exponential function of the predicted natural log WTP values to get the predicted WTP in dollars. The details of how we do this are given in footnote (c) of Table 6.

¹⁵While it is then counter-intuitive that the other home population dummy variables are statistically insignificant in Table 5, this may be because of a lack of variation in the sample data.

¹⁶A lack of variation in the data may be causing the other ‘year-in-college’ dummy variables to be statistically insignificant.

¹⁷That is, to measure the relationship between sky glow and electric lighting one must not only consider the lighting, but also the angular distribution of the light emitted from the light fixture, the light reflected from the ground and its angular distribution, as well as atmospheric effects of humidity, for example.

¹⁸Chalkias et al. (2006) describes a complex methodology that uses Geographic Information System (GIS) and Remote Sensing (RS) technology to model light pollution over the past decade in Athens.