Abstract
This paper attempts to estimate the price elasticity of demand for the amount of nonresident hunting licenses in five states. This has a very real policy implication in that it could help states determine the effect of changes in the priceof nonresident hunting licenses, and hence revenue, on the demand. Population and wages had very little effect on the outcome, while the price of fuel alternated between a positive and negative association with the number of nonresident licenses sold. This analysis found each state has differing elasticities for the demand of nonresident hunting licenses. Colorado has the lowest price elasticity, or inelastic demand, implying that their price for nonresident hunting licenses could be increased with little effect on demand. The opposite is true for Washington, as they have higher price elasticities, or more elastic demand.
Introduction
Hunting and fishing have been long held pastimes for generations. Fathers taught sons, sons taught grandsons and so on, until recently. Starting in the mid-1920s, states turned this pastime into a revenue producing vehicle through permits and license sales. Fish and wildlife departments were formed with the mission of preserving and expanding the art and tradition of wildlife recreation. However, amount of hunters in the US has declined from 13 million people in 2001 to 12.5 millionin 2006. West Virginia alone has reported a loss of nearly $1.5 million within the last decade from declining license sales.[1]
Outdoor recreation, which is typically defined as hunting, fishing, and other nonconsumptive activities, has a significant economic impact. In 2006, sportsmen spent $76.6 billion on equipment and trip related expenses;$42.2 billion in fishing and $22.9 billion in hunting. Of that,$11.6 billion was spent permits, licenses, and certifications.[2]
All states of the union have some version of a natural resources department, and a corresponding division of wildlife. The agencies administer conservation strategies, create community activities, ensure the quality of the parks, regulate licenses and permits, and disseminate knowledge of the outdoors through publications and, usually, free courses. However, these agencies are reliant on permits and license sales to help fund their mission. Comprehensive state, and small portion of federal, funding for various programs provide the greatest portion of revenue needed to operate. An equally important part of the budget comes from the sale of licenses and permits, by the division of wildlife. The state departments use the revenue derived from the sales of permits and licenses to fund their work and increase the quality of their services and environment.
The US Fish and Wildlife Service conducts a direct survey every five years which details the various aspects ofoutdoor recreation; hunters, fisherman, and bird watchers are asked to fill out a survey. The survey found that nearly 87.5 million participants spent $123 billion in 2006 on outdoor recreation, throughout the country. The value of the hunting and fishing to the economy of the entire country is evident. The goal of this paper is to determine the price elasticity of demand for hunting licenses, and its policy implications. This analysis has a very real policy implication of determining the effect of pricing changes has the demand for nonresident hunting license sales. The determination of the price elasticity would allow the states for project their revenue after a change in the price and readjust their budget accordingly.
Literature Review
The amount of work currently published relating directly to hunting and demand for such is rather limited. Hunting demand can be related to demand for recreation and some authors have used different recreation demand models interchangeably. There are various methodologies used by authors analyzing the demand for hunting: Travel-cost, contingent valuation, various empirical methods, and benefit-cost strategies have been used in the past to determine various aspects of hunting. Whitehead, et al (2003), uses contingent valuation analysis to explore the loss of temporal reliability in contingent valuation surveys. Contingent valuation surveys tend to vary with the researcher who administered the questionnaire. Whitehead, et al, set out to determine whether the contingent valuation analysis was reliable. Whitehead (2003) was the first to use comparative data to determine if the willingness to pay data obtained through contingent valuation surveys was sufficient to determine the change in willingness to pay. Using the U.S. Fish and Wildlife Service's (USFWS) and the National Survey of Fishing, Hunting, and Wildlife-Associated Recreation (FHWAR), they note the differences in the two questionnairesand attempt to create a standardized data set. The authors find that changes in sampling, questioning wording, and temporal unreliability could be at fault for compounding the results and the significance of their previous findings.
John A. Curtis (2002) delves into the controversial issue of salmon angling in the Ireland. Curtis is the first to develop a demand estimation function and determinants for salmon angling. Curtis asserts that fisheries could use the determinants of his demand for salmon angling to increase the desirability of their own facilities by adjusting their pricing. Policy implications can also be made at the federal level as to the management of recreational and commercial salmon harvesting. Curtis (2003) uses the travel cost approach to estimate the salmon angling demand. The demand is defined as a function of the number of days taken to fish, a travel cost proxy, and some angler characteristics. The data used in the paper was collected from surveys administered to fishermen at the popular sites for salmon fishing. The results, from their analysis, show that the fishermen value the salmon resource highly at $206/day. The analysis also shows a consumer surplus of nearly 67% of total willingness to pay. Producersmight be able to increase revenues by increasing the price of the fishing opportunities to be better closer to the willingness to pay.
Milon (1991) explores the demand for hunting for various species of large and small game. While many studies have looked at demand for hunting for one species or one specific group of game, this paper is the first to look at the demand for species variety. This paper is important in that it highlights that many hunters enjoy hunting for various types of game. Milon states that it is hard to determine the actual demand for a variety. Over time, hunter expenditures have increase 60 percent while participation has only increased by 14 percent. This could mean that consumption associated with the demand for wildlife resources has been more influenced by changing user preferences rather than new user pressures. Rather than attempting to measure quality of the hunting by using the number of animals allowed to be harvested of one species, Milon (2003) instead uses the number of species hunted as a measure of quality. The model uses the 1985 National Survey of Fishing, Hunting and Wildlife Associated Recreation, NSFHWAR dataset. The model incorporated the travel cost data from the survey. The model uses the household production framework. Nearly all of determinates of the results discovered proved to be significant. This model intended to expand the existing literature by using an alternative approach to quality measurement of wildlife recreation.
Brown, et al (1994), developed a model to predict big- and small-game hunting license sales in the state of New York. Traditionally, models used to predict usage and willingness to pay used cross sectional data sets. Survey work is very expensive and time consuming and in some cases the cost is prohibitive. As an alternative, the authors decided to use time series data. Their objective is to show that use and revenue from license sales can be estimated by using an inexpensive database; such as time series. The results indicated that big game license sales correlated positively with miles of interstate driven to get to the area hunted, deer harvest of the previous year, and license sales from the previous year. Sales correlated negatively with nonagricultural employment and the weighted license fee unadjusted for inflation. Brown, et al (1994) found that these model regressions are effective and significant to help wildlife managers understand the aspects that affect license sales, and hence revenue and usage demand change after fee increases.
Loomis, et al (1999) provides an alternative approach to evaluating the validity of contingent value estimates of willingness to pay by estimating demand for elk and deer hunting permits using the historic variation in non-resident license prices. The authors compare their results with those of a dichotomous survey measuring contingent values. The standard consumer demand model is used to determine the quantity of elk licenses sold. The contingent valuation model is based on a survey that aims to determine if hunters would be willing to pay for a hunting license if the price were increased of discrete increments. The comparison between the two analyses allows for assurance that the contingent valuation method is significant. The authors found that the contingent valuation method tends to underestimate the willingness to pay.
The following analysis attempts to determine the price elasticity, or the effect of price changes on demand, for nonresident hunting licenses for five states, of demand; one state from each major region of the United States. This knowledge will assist wildlife departments by illustrating the effect of pricing changes on demand for nonresident hunting license sales. This will help the departments make more informed policy decisions.
Empirical Model
The model is based upon previous research by Loomis, et al (1999), and is a simple variation of the standard consumer demand model. This model incorporates income, prices, substitutes, complements, and population. The concept model is as follows:
Licent=βi+β2Incomet+β3[Nat]Pricet+β4Pricet+β5Popt+β6Yeart+ β7FuelPricet+έit
where Licen represents the number of nonresident licenses sold for each state in year ‘t’, Income is the average yearly wage of an individual in each statein year ‘t’; NatPrice is the national average price of a nonresident hunting license in year ‘t’ and is used as a substitute for hunting in any given state; Price in the inflation adjusted price of a nonresident hunting license each statein year ‘t’; Pop is the estimated yearly population in each state in year ‘t’; Year is time trend variable that corresponds to the actual year of observation;FuelPrice is the average yearly price of a gallon of gasoline in each state in year ‘t’and is used as a complement to hunting. According to theory, Income, Pop,NatPrice, and Year, should have a positive effect on the dependent variable. Also, Price, and FuelPrice should have a negative effect.
Following Loomis, et al (1999), the dependent variable was logged to create the log-linear form of regression. However, by also logging Price the elasticity can be found; which has greater policy implication; transforming this regression a log-log function. The results will illustrate the percentage change in licenses sold for a given percentage change in the independent variable. The final model is as follows:
lnLicent=βi+β2Incomet+β3[Nat]Pricet+β4lnPricet+β5Popt+ β6Yeart+ β7FuelPricet+έit
Data Description
The US Fish and Wildlife Service collect data from states and compile their findings into an easily accessible format. This data includes the number of licenses sold to residentand nonresident hunters, the total amount spent for each category, and a national aggregation. They have compiled this data, nationally, back to 1929 and back to 1965 at the state level. Each statehas 38 observation years, 1965-2002. The price data used is the nonresident price of a license, in keeping with the Loomis framework. The price was derived by taking the total expenditure on nonresident hunting licenses and dividing by the number of nonresident hunting licenses sold. Due to systems changes and archived proprietary data in each state, the actual prices charged could not be found and analyzed in the limited amount of time available for this study. Nonresident data was used because a non-resident hunter is willing to travel to hunt and has the most choice in substitutes and compliments. For instance, a hunter which is prepared to travel Colorado could just as easily have picked Idaho or Texas or even Georgia. These hunters also have the most choice when it comes to compliments. For example, the cost of fuel is used as compliment in this analysis, so these non-resident hunters could choose between flying, riding a bus or train, or driving their personal vehicles. Population figures are derived from the Census Bureau’s yearly population estimates. The population is a yearly estimate of the number of people in each state. Income data were also taken from Census reports and is also a yearly estimation of average wages. The fuel pricing data isappropriately adjusted for inflation and is taken from the Bureau of Labor Statistics. Leaded fuel is used until 1981 when unleaded fuel came under popular usage.
This analysis uses one state from each major region of the United States: New York for the East, Ohio for the Midwest, Texas for the South, Colorado for the Central, and Washington for the West. Each region has its own set of huntable species and each is somewhat different. In an attempt to be as objective as possible, the states chosen from each region were chosen at random. However, Ohio was chosen because of the affinity for its landscape by the author. This paper focuses on hunting and each state has some form of big game, and at least one variation of deer. New York has whitetail deer, bear, and bobcat as its big game hunting opportunities. Ohio is the most limited state, in that it only offers whitetail deer for big game hunting; while black bear is making resurgence in Ohio, they are not hunted. Texas is the most diverse state with hunting available for alligator, pronghorn antelope, whitetail deer, mule deer, and javelina. Colorado is full of differing opportunities as it offers big game hunting for mountain lion, pronghorn, whitetail deer, elk, moose, mountain goat, RockyMountain big horn sheep, and black bear. Washington also has elk, black bear, cougar, whitetail deer, blacktail deer, mule deer, and lynx for hunting. The variation in the opportunities for hunting creates differing elasticities.
Results
Each time series, state level data set is subjected, separately, to the same regression process described above. Time series has many advantages and drawbacks. Relatively low cost of data collection and ease of manipulation are some of the advantages. However, serial correlation and multicollinearity can be serious issues when dealing with time series data. Serial correlation occurs when there is correlation between with the same variables over time. This can be tested for by using the Durbin-Watson test. The critical value for Durbin-Watson cannot be estimated exactly, so a range is used which is determined by the number of observations and the degrees of freedom associated with a model. This range has an upper and lower bound and is between 0 and 4; 2 being the mean. This analysis uses 38 observations and has 6 degrees of freedom. The Durbin-Watson statistics are 1.21 and 1.71, for the lower and upper bounds, respectively. If the Durbin-Watson statistic is below the lower bound then there is definitely serial correlation and the appropriate measures must be taken to correct. If the Durbin-Watson statistic falls between the upper and lower bounds then it is not known whether there is serial correlation and still must be corrected to ensure unadulterated results. To correct for serial correlation, a first order autoregressive process is used. This process will give the coefficient of rho, ρ, which is the error term of the previous time period for a given variable. This separation of the error terms allows the true error term to be realized. The explanation of the each state’s Durbin-Watson statistics and their implications follows in each state’s explanation.
Typically, a trend variable, Year in this case, can be highly multicollinear with a population variable, Pop. To correct for multicollinearity, one of the collinear variables should be dropped from the regression. To confirm multicollinearity, a correlation test can be done, and was done showing an average of 98% correlation between Pop and Year for the states. If not corrected these issues can cause skewed results and can create incorrect implications. So, for each of the state’s regressions, Year has been dropped to correct for multicollinearity.
Descriptive statistics and estimation results can be found in table form in the Appendix under Tables 1 and 2, respectively.
Colorado, the biggest hunting state in analysis, has only 2 significant variables, except the autoregressor. The Durbin-Watson statistic is 1.1733and falls below the lower bound of 1.2, according to the statistical tables. This requires that the regression be corrected for serial correlation using an autoregressive sequence. However, after correction, the Durbin-Watson is 1.5363, which is within the gray area, but it did improve greatly. The R-squared of the regression is 82.48% and the root mean squared error (RMSE) is 0.18821. After correction, the main variable of interest, lnPrice has a coefficient of -0.3526and is proved significant with a t-value of -2.12, at 95% significance level. Since this is a log-log function, a 10% increase in price will decrease the amount of hunting licenses sold by 3.526%. Pop and Income, which are not significant, both have negative signs, with coefficients of -0.0000001098 and -0.000009079, respectively. Both of these would be contrary to theory as it would be expected that these should be positive. The regression shows that as income increases the amount of licenses sold decreases inferring that hunting is an inferior good. Possibly, increasing wages may indicate promotions or increased responsibilities which would increase the opportunity costs to hunting and hence decrease the occurrence of the purchase of hunting licenses. Either way, the variables are not significant and even if they were, their effect would be extremely small. Natprice, the substitute, is the other significant variable in this case with a t-value of 2.24 and a parameter estimate of 0.0203. In this regression, RealFuel is positive, with a parameter estimate of 0.1603, and is contrary to theory. While being insignificant, it is odd the relationship would be positive and shows that an increase in the price of fuel would increase the amount of licenses sold. One answer might be that nonresidents who hunt in Colorado have aninelastic demand for gasoline. Meaning, the price of fuel is a not an important cost to them. The autoregressor, which corrects for serial correlation, has a coefficient of -2.33. With the correction for multicollinearity, by dropping Year, Natprice became significant where it wasn’t in the original situation (including year).