Â«Toward More Effective Endangered Species Regulation By Jacob P. Byl Dissertation Submitted to the Faculty of the Graduate School of Vanderbilt ...Â»
A second benefit of the recent sample is that the North Carolina Department of Environment and Natural Resources feels that the RCW data are more accurate for later periods, decreasing the amount of measurement error.
woodpeckers will move to the property ( Îł in the conceptual model). Prediction 1 is that Îž is positive. For controls, X is a vector of plot characteristics including the siteâs tree
Although the above model provides a test for whether RCWs impact the timber harvest decision, the model relies on an assumption that forest plots near RCWs are otherwise similar to plots that are not near RCWs after controlling for available variables within the model. With nonrandom distribution of RCWs, one may be concerned that there is heterogeneity across geographic areas that is not observed in the FIA data.
A difference-in-difference model, specified in Equation 3, can help control for this type of unobserved heterogeneity by comparing the difference in harvesting of pure pine sites that are near RCWs with the difference in harvesting of mixed sites near RCWs. Since RCWs will not nest or forage in pine forest that is mixed with hardwoods, landowners should not be concerned with proximity to RCWs. By comparing the differences in harvest rates between these two groups, one that is impacted by RCWs and one that is not, I can control for this unobserved heterogeneity. In this equation, the variable of interest is ÎŽ, which picks up the impact of RCW on pure pine sites as compared with mixed sites. The coefficient Îž estimates the impact of RCW on sites that are not suitable habitat. The coefficient Ï is the measure of how pure pine sites differ from mixed sites in ways other than RCW. The heteroskedasticity-robust error term Î”
in ways that are favorable for wildlife habitat. This is true because managing land in ways that encourage wildlife like deer and game birds tends to improve conditions for RCW as well (U.S. FWS 2010). Equation 4 lays out a difference-in-difference model to test whether this prediction is true. The variable Improve takes a value of one for sites that have been managed in a way to improve habitat for wildlife in the past five years. Other variables in this equation take similar forms as those in Equation 3.
geographic area, restricting to a dozen counties in and around the Sandhills area, but from a longer timeframe starting in 1982. This narrower and longer dataset allows me to look at habitat destruction both before and after implementation of the safe-harbor program using treatment and control areas that are close to each other and have similar terrain and features.
counties that are targeted for the safe-harbor program run by FWS. Post takes a value of one in years after 1995 when the safe-harbor program was started. The variable of Standard errors can also be clustered at the county level to allow for arbitrary correlation within counties and the same variables remain statistically significant. Running this and subsequent equations with a probit model also yields similar results.
interest is a three-way interaction between Eligible, Post, and the variable RCW. The coefficient ÎŽ measures the impact of the safe-harbor agreement on eligible landowners near RCW compared with those not eligible or not near RCW. Interactions are a set of two-way interactions and dummies for Eligible, Post, and RCW to control for the areas, times, and site characteristics of interest. Controls for other site characteristics and year dummies are also included. A heteroskedasticity-robust error term, Î”, is the final variable
The safe-harbor program may also encourage landowners to manage land in ways that are favorable to wildlife, as in Prediction 4. Equation 6 takes a similar form as Equation 5, but tests for the impact of the safe-harbor program on the probability of improving wildlife habitat.
An alternative methodology to measure the impact of the safe-harbor program is to match sites that have entered safe-harbor agreements with similar sites that are not in safe-harbor agreements, then look at differences in outcomes across the groups. Selection issues preclude the use of simple models to test the effectiveness of the program because there are reasons to believe that landowners who choose to participate in safe-harbor agreements may differ from landowners who do not participate. Nearest-neighbor matching allows me to match sites based on a combination of factors including location, proximity to woodpeckers, the starting value of timber and value of growth, and tree type. Each site in a safe-harbor program is matched with the three most similar sites, all surveyed after 1995. If outcomes differ across the treatment group of sites in the safeharbor program and the control group of matched sites, there is evidence of the effectiveness of the program. Matching estimators can measure the average treatment effect on the treated, which focuses on the sites within the treatment group, or the average treatment effect across the entire population. In this case, average treatment on the treated is a measure of whether the safe-harbor program is effective for landowners who have chosen to enter it, where average treatment effect is a measure of whether the safe-harbor program would make a statistically significant difference across the area of interest.
Matching methods are also used to look at the effectiveness of the safe-harbor program on increasing management for wildlife habitat.
VIII. Results Results from the probit model of harvest probability are presented in Table 4, alongside results from the 1990s sample analyzed in Lueck and Michael. While each additional woodpecker colony increased the probability of harvest by about 0.1% in the 1990s sample, there is not a statistically significant impact of RCWs in the 2000s sample.
Coefficients on other variables maintain similar sign and significance across the two samples, so this model indicates that perhaps habitat destruction to avoid having RCWs move onto property has slowed down. This could be taken as a sign that the safe-harbor program has been successful. However, as mentioned in previous sections, the results from this model may be largely driven by unobserved differences between areas with RCW populations and areas without RCWs.
Table 5 presents results from the difference-in-difference model that controls for this heterogeneity and points to a continued impact of RCWs on timber harvests. The variable of interest of pure-pine sites interacted with RCWs has a positive and significant coefficient. With this preferred model of harvest behavior, pure pine sites that are near RCWs are 25% more likely to be harvested than pure pine sites that are not near RCWs.
Although the magnitude of the effect is about half the size of the effect found by Lueck and Michael with older data, there is evidence that Prediction 1 holds and landowners continue to destroy habitat to prevent RCWs from moving onto their property. The smaller magnitude of the coefficient may indicate some success of the safe-harbor program, as discussed below. The smaller effect of RCWs is also consistent with changes in the market for timber products that have occurred over the past twenty years, as described in Section IX.
Table 6 presents results of the difference-in-difference model of landowner behavior with respect to improving land as wildlife habitat. The variable of interest is again pure-pine sites interacted with RCWs, which has a negative and significant coefficient indicating that landowners near RCWs are about two percentage points, or 40%, less likely to manage their pure pine sites in a way that improves habitat for wildlife. Prediction 2 also appears to hold when describing recent landowner behavior.
Predictions 3 and 4 relate to the effectiveness of the safe-harbor program in the Sandhills area of North Carolina. Table 7 presents results of the triple-difference model that tests whether landowners with the option of a safe-harbor agreement are less likely to destroy RCW habitat. The variable of interest is the three-way interaction of RCW with sites that are in the area targeted for the safe-harbor program (Eligible), and surveyed in the years after the program was started (Post). This variable has a negative but insignificant coefficient, indicating a lack of strong evidence of the effectiveness of the program, although the negative coefficient is in the expected direction of Prediction 3.
The interaction term between Eligible and Post is significant and negative, providing some evidence that landowners in the area may have reduced the amount of habitat destruction by approximately 6 percentage points, or a 33% decrease from the mean, following introduction of the program.
Table 8 presents results from the triple-difference model predicting whether landowners improve wildlife habitat on their land. The triple-difference variable of interest has a positive and significant coefficient indicating that landowners in the target area of the safe-harbor program after the program started are half a percentage point, or 10%, more likely to improve wildlife habitat compared with landowners who are not in the programâs target area.
As presented in Table 9, matching estimators show mixed effectiveness of the safe-harbor program as well. Three control sites that are not in safe-harbor agreements are matched with each site in a safe-harbor agreement using nearest-neighbor matching based on location, number of RCW within ten miles, tree type, starting value of timber and value of an added year of growth. For harvest probabilities, the average treatment for landowners in the treatment area (ATT) is a statistically significant negative thirteen percentage points, which is a 72% reduction from the average harvest probability. This indicates that landowners who are in safe-harbor agreements have greatly curtailed their harvesting behavior. When looking at the average treatment effect for sites across the relevant area (ATE), there is no significant change in harvest probabilities across treatment and control sites. The average treatment on the control group (ATC) is positive but not statistically significant. This would mean that if landowners in the control groups were forced to enter safe-harbor agreements, those landowners may actually increase harvest behavior, perhaps as acts of defiance or distrust in the government. However, since these treatment effects are based on data from a voluntary program, it is difficult to say with confidence how well the results would apply to compulsory programs.
When looking at changes in improving wildlife habitat, the nearest-neighbor matching estimators do not show significant effects. The ATT estimate for landowners in the safe-harbor program is a positive seven percentage point increase in the probability of improving wildlife habitat. This estimate is in the expected direction, but is not statistically significant. The ATE estimate for landowners across the area is an eleven percentage point increase in the probability of improving wildlife habitat that is not statistically significant. The ATC estimate for landowners in the control group is similar.
IX. Explanations and Policy Implications The results from this study suggest that private landowners continue to engage in preemptive habitat destruction. Although the estimate from a model similar to that used by Lueck and Michael does not show evidence of habitat destruction, the estimate from a model that controls for unobserved heterogeneity of forest plots points to a 25% increase in the probability of harvest for plots near RCWs. Although the FWS has decreased the acreage that is set aside for each RCW colony and offered voluntary agreements, as described below, to lessen the threat of regulation, landowners still appear to be avoiding ESA regulations by destroying habitat.
Changes in market conditions may be one factor working against efforts of the FWS to have a more cooperative relationship with landowners. Technological advances in wood products with things like oriented-strand board (OSB) allow landowners to harvest young timber with less of a penalty for not allowing trees to reach a diameter sufficient for saw timber. OSB and other alternative lumber products caused the price gap between chip and saw timber to narrow, so landowners can put trees on a thirty-year rotation and avoid the threat of RCWs moving in without suffering a large financial penalty for not allowing the trees to reach a diameter suitable for saw timber.
One way that the FWS has tried to encourage cooperation with landowners and mitigate the perverse incentive to destroy habitat is through safe-harbor agreements. One of the first safe-harbor programs in the country was for RCWs in the Sandhills area of North Carolina. This program therefore offers one of the richest data sources for empirical tests of the effectiveness of the program. As the above results describe, the variable of interest in the model of harvest behavior does not show the program having a significant impact on the landowners we would most expect to be affected, namely those who have RCWs nearby. Instead, the coefficient on a variable that interacts the area and time of the program shows a 33% decrease in harvest behavior in the treatment area after the program was started. This estimate uses the surrounding counties that are not in the safe-harbor target area as controls.
The lack of significance on the triple-difference variable in Table 7 means that this result is not driven by landowners with RCWs nearby, so it should not be considered strong evidence of the effectiveness of the program at decreasing habitat destruction through tree harvests. Instead, it may point to spillovers of conservation behavior to landowners who are not immediately threatened with RCWs, but may be in the future.
FWS data on the safe-harbor agreements indicates that a third of the agreements are in place for properties that are not currently within ten miles of any existing RCW colonies, so these landowners are locking in an agreement with a zero baseline as an insurance policy. Although this behavior will tend not to have an immediate benefit to RCWs because these sites are outside the traditional range of fledgling RCWs searching for new colonies, it may have long-term benefits by promoting more mature pine in the area.