Phragmites Removal Increases Property Values in Michigan’s Lower Grand River Watershed

The presence of Phragmites australis, an invasive wetland plant, negatively affects coastal property values and home prices rise with distance from Phragmites. Home prices increased as distance to Phragmites increased at a rate of $3.90/meter. Removing Phragmites from a property so that the next closest Phragmites was 400 m away results in a property value increase of over $1,500. Removing all Phragmites within 400 m of any property results in a total property value impact of $837,000. This generates about $13,457-$15,121 in additional property taxes each year once the prices and taxes adjust to the plant’s removal. We estimated the cost of Phragmites removal at $687/ha. Removing the approximately 36 ha of Phragmites in the area would cost about $25,041. Future treatments would likely be less than that of the first year. The estimated cost of the first year of Phragmites removal is less than the estimated two years of annual property tax revenue increases. This application notes is available in Journal of Ocean and Coastal Economics: https://cbe.miis.edu/joce/vol4/iss1/5


INTRODUCTION
The introduced genotype of Phragmites australis (Cav.) Trin. ex Steudel (or common reed, hereafter referred to as simply Phragmites) is an invasive wetland plant. The genotype native to North America was relatively uncommon and in recent years, the introduced genotype has become a fast-spreading nuisance (Saltonstall 2002). Phragmites, a warm-season perennial grass, can grow up to four meters tall and has flower clusters that are open and feathery at maturity. The State of Michigan (2017) lists Phragmites as an invasive species. The introduced, invasive genotype has darker leaves than its native cousin, has lighter-colored rhizomes, and forms more monotypic stands (Great Lakes Commission 2017) ( Figure 1, Figure 2). Public agencies across the United States spend more than $4 million per year treating areas with Phragmites (Martin and Blossey 2013). The Michigan Department of Natural Resources (DNR) spent more than $126,000 between 2005 and 2012 treating Phragmites on public lands in the Bay City area near Lake Huron (Michigan Department of Natural Resources 2015). Though the economic costs of controlling Phragmites are clear, the economic benefits and financing for control programs are less clear.  A thorough review by Hazelton et al. (2014) documented how Phragmites establishment in coastal wetland ecosystems is associated with decreased biodiversity, reduced habitat quality for fish and wildlife, and disrupted biogeochemical cycles. Phragmites also negatively effects human use of coastal and wetland areas. The tall, monotypic stands may impair water access and viewsheds. Additionally, a large amount of slow-decaying biomass in the form of dense standing thatch remains following every growing season, resulting in a major fire hazard accumulating year after year. Coastal residents are also becoming more educated regarding the harmful effects of invasive species and are therefore more likely than ever to pressure their neighbors into removing invasive populations before they cross property boundaries. Removing Phragmites and restoring native ecosystems at a watershed scale should enhance ecosystem function and service provision. The economic value of ecological benefits from Phragmites removal are not well understood. This paper, therefore, seeks to fill that gap by estimating the property value effects of Phragmites removal.
Many studies have estimated the economic value of wetlands and the results are generally positive. For example, Brander et al. (2006) conducted a review and metaanalysis of nearly 200 wetland valuation studies, but only five of those used hedonic (property sales) models. The median wetland value from the five hedonic studies was $5/ha (in 1995 dollars) but mean value was about $8,000/ha. Other valuation methods exhibited similarly wide ranges. Wetland values were positively associated with both per capita GDP and population density.
Studies from Ohio, Minnesota, Oregon, and Australia found that home sale prices are inversely associated with distance to most, but not all, types of wetlands. That is, property prices increase as distance to wetlands decreases (Babb 2012;Doss and Taff 1996;Mahan et al. 2000;Tapsuwan et al. 2009). Associations ranged from $0.13/m (Mahan et al. 2000) to about $33.60/m (AUS$42.40/m in the original paper) (Tapsuwan et al. 2009). Metrics of wetland quality, such as size, buffer, and habitat diversity, also had a positive influence on housing values (Babb 2012). An analysis of coastal marsh along Michigan's Saginaw Bay using travel cost and contingent valuation methods found that people valued the wetlands at $756/ha ($1,870/ac) in 2005 dollars (Whitehead et al. 2009). This equates to about $928/ha ($2,292/ac) in inflation-adjusted 2015 dollars. There are few valuation studies that focus directly on Phragmites. One study of a Greek wetland found that people had a positive willingness to pay for reducing the area of the wetland covered by Phragmites (Birol, Karousakis, and Koundouri 2006). These studies suggest that people see high-quality, biologically-rich wetlands as an asset and are willing to pay to obtain the ecosystem services that flow from them. As far as we know, our study would be the first economic valuation of Phragmites removal using a hedonic model.
Environmental economists have developed various techniques for measuring the economic values of ecosystem goods and services. Economists often use hedonic models to estimate the willingness to pay for various attributes of consumer goods including environmental quality. In a hedonic model, the consumer good is viewed as a bundle of characteristics and sales price is regressed against these attributes. With a reasonably large sample size, the analyst can estimate the marginal effect each attribute has on the sales price. The repeat-sales model is a variation on the standard hedonic model that is often used in residential home sales. Economists can regress the change in sales price over time against the relevant time-variant characteristics, while leaving out characteristics that do not change over time. This simplifies the process considerably, although it limits the available data to those goods that have sold multiple times (Freeman 2003). The repeat-sales model was developed by Palmquist (1982) and was first used to analyze the effect of highway noise on residential properties.
Since 2007, The Nature Conservancy and other conservation entities in the Michigan Dune Alliance collaborative have implemented extensive terrestrial invasive plant control efforts throughout the dunes, wetlands, and nearshore forests of Eastern Lake Michigan, ultimately aimed at restoring and maintaining the ecosystem health, processes, and services of this globally-unique coastal system. Traditionally, financial resources for this type of ecological restoration come from public and private grant sources, a system primarily based on short-term (1-3 year) funding cycles and defined project start and end dates. While these traditional funding sources have been vitally important to both define the extent of invasive species impacts across this 500-mile stretch of shoreline and reduce those populations to a manageable level, they do not currently offer secure, long-term funding to maintain those outcomes in the future. To best maintain the ecological integrity of this system for natural habitat and human well-being, Michigan Dune Alliance members began to investigate alternative funding models that could derive resources for sustainable ecosystem management from the human-use benefits these coastal areas provide.
There are anecdotal reports that realtors in coastal areas such as Traverse City and Grand Haven, Michigan, USA were actively steering potential home buyers away from properties on which stands of Phragmites were established. Simultaneously, The Nature Conservancy developed a whitepaper on coastal conservation financing specifically focused on the aforementioned goal of identifying sustainable funding options for invasive species control (The Nature Conservancy 2010). Included in that document was an evaluation of multiple financing options, including tax increment financing (TIF).
Tax increment financing is based on the idea that improvements in, for example infrastructure, within a designated district can stimulate incremental growth in property assessments and tax revenues. Those additional revenues over time are earmarked to pay for the original improvements. The improvements, therefore, should be self-financing. Since its origins in California in 1952, TIF has become the most popular tool in the United States for financing economic development. It also has expanded from a tool to invigorate depressed city centers to an approach for financing more general public investments in infrastructure (Briffault 2010). In Michigan, TIF can be used to promote economic development in downtown districts, manufacturing and technology parks, commercial districts outside of city centers, and brownfield industrial sites (Bassett 2009).
This expansion in scope also includes the use of TIF for conservation purposes. In 2008, the Michigan legislature passed the Water Resource Improvement Tax Increment Finance Act (PA 94). The act enables local government units to create TIF districts to promote water resource improvements or access to inland lakes (State of Michigan 2008). Programs that remove invasive aquatic plants, such as Phragmites, could potentially be funded through TIF.
While TIF appeared to be well-aligned with the desired funding stream in terms of potential governance and timeframe, it was undetermined whether the revenue generated would be sufficient to support ongoing, long-term control of a species such as Phragmites in coastal areas (The Nature Conservancy 2010).
The Grand River flows more than 250 miles from its headwaters near Jackson to its mouth at Grand Haven where it empties into Lake Michigan. Phragmites has invaded many of the wetlands in the lower Grand River area, including the river's bayous and tributaries as well as Spring Lake ( Figure 3, Figure 4). We used a repeat-sales model to estimate the property value effects of Phragmites removal in Ottawa County, Michigan. We hypothesize that coastal property values are negatively affected by proximity to Phragmites and that removing Phragmites will increase property values. As property values increase, so should property tax revenues. If the additional tax revenues are greater than the cost of Phragmites removal, then the publicly-funded management regime should be self-sustaining. This is similar to the TIF concept used to improve blighted urban and industrial neighborhoods (Briffault 2010). The results of this study will help local units of government understand the economic benefits of removing Phragmites and will inform TIF-style policy options.

Study Area and Data Sources
The study area was the lower Grand River watershed including the communities of Grand Haven and Spring Lake in Ottawa County, Michigan, USA ( The Phragmites survey protocols, however, were not identical. This is a source of uncertainty. The absence of Phragmites points in Spring Lake in the 2015 survey was confirmed by Spring Lake Township Supervisor John Nash (personal communication). Spring Lake Township treated the lake for Phragmites in 2013, and remaining Phragmites stands were treated again in 2014. No treatment was necessary in 2015 because it was eradicated. Phragmites stands on Harbor Island near the mouth of the Grand River were also treated and eradicated. The untreated areas along the Grand River show relatively consistent patches of Phragmites in both 2010 and 2015. This gives us confidence that, although the data were not collected identically and systematically between the organizations, they both found the major patches of Phragmites, and the differences can be attributed to treatments on Harbor Island and in Spring Lake.
Ottawa County established a database and protocol that they will follow starting in 2015 so that future Phragmites conditions in the lower Grand River watershed can be tracked. Beginning with the 2015 survey, Ottawa County Parks is collecting not only Phragmites locations but also area and density. The data were categorical (Table 1). We assumed that each 2015 point location is at the top of its category and used 0.81 ha for the >0.40 ha category (the median value). The high-end values were chosen to simulate a "worst case" scenario. If the actual area of Phragmites is less, then the cost of treatment will likewise be lower. The high-end estimate therefore is about 36.42 ha of Phragmites in the lower Grand River area in 2015. Using the midpoint of the first three categories and 0.40 ha for the highest yields an estimate of 21.21 ha of Phragmites. The data also included a category for individual stalks, but this was ignored in the area calculation. Sales data from 2004 to August 2015 for Ottawa County were obtained through the Ottawa County (Michigan) Assessor's Office. Sales less than $10,000 were considered invalid (not arms-length transactions) and were removed from the dataset. The tabular sales data were joined to the parcel polygons through the PIN field. The extreme northern end of Spring Lake extends into Muskegon County. Sales from this area were not included in the analysis.

Economic model
A repeat-sales model was used to measure the effect of Phragmites on property values. Repeat-sales models compare the change in a property's sales price with the change in the variable(s) of interest. The model assumes that all other attributes of the property, including both housing and neighborhood characteristics, are unchanged between the sales. It is important to note that homeowners do invest in renovations and some houses may violate this assumption. However, at least one study has shown that repeat-sales models provide comparable results to hedonic property value models that include a suite of home and neighborhood characteristics (Hansen 2009).
The regression model can take one of two forms: random effects or fixed effects. The Hausman test, a form of chi-square, can determine whether the unique errors are correlated with the regressors. The null hypothesis is that the errors and regressors are uncorrelated, in which case the random effects model is preferred (Greene 2008). We performed the Hausman test (χ 2 = 33.38, Prob>χ 2 = 0.00) and rejected the null hypothesis. Therefore, a fixed effects model was chosen.
The dependent variable was the natural log of the sales prices (ln_price). The independent variables included distance to the closest Phragmites location (in meters, dist_phrag) and a property value index for each year. The 2004 variable was withheld to prevent multicollinearity problems.
Hedonic models can be hampered by spatial autocorrelation within the data. That is, property prices for nearby houses tend to be similar and violate the statistical assumptions about independence of observations. Spatial autocorrelation can result in biased regression estimates. The sales price dependent variable was tested for spatial autocorrelation using Moran's I in ArcGIS 10.1. Most cases of significant spatial autocorrelation are dealt with using spatial lag or spatial error term models. With fixed effect repeat-sales models, however, this is not possible. A reasonable approach that mimics the spatial error model is to adjust the standard errors for clusters (repeated sales of the same property) ( where λt is the annual dummy variable for year t (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015); αi is the fixed effects for parcel i; zit is the vector of time-varying parcel-level characteristics (in this case, distance to Phragmites); β is the vector of regression coefficients; and εijt is the error term including error clustering for group j.
Distance to Phragmites was calculated based on the year of sale. For sales before 2013 (when Spring Lake Township began removing Phragmites), distance was calculated based on the Nature Conservancy's 2010 points. For 2013-2015 sales, the Ottawa County Phragmites points were used. This allows us to capture the change in Phragmites distribution over time. This method assumes that the Phragmites stands were relatively stable in size and location from 2004 to 2010. Although it would be better to use sales that are closer in time to the Phragmites assessments, limiting the sales years also limits the number of observations. Including repeated sales back to 2004 was necessary to create a sufficiently large dataset for the regression model.
The sales were further restricted to properties within 100 meters of a waterbody including the Grand River, Spring Lake, or the various bayous and tributaries. The dataset comprised a total of 967 sales of 384 properties. Repeat-sales models require a minimum of two sales. The average number of sales in the dataset was 2.5 with a maximum of eight.
The average sales price for homes in the dataset was about $185,000. This is slightly higher than the median home value for Ottawa County of $153,200 (US Census Bureau 2014). This is to be expected because of the higher proportion of waterfront homes in the dataset. The total value of Phragmites removal was calculated using the following formula (Equation 2): where ValueRemoval is the property value generated by totally removing Phragmites within 400 meters of a property; #Homes is the number of homes with Phragmites within 400 meters; MeanPhragDist is the mean distance to the nearest Phragmites location; and PricePhragDist is the change in sales price that results with each one-meter increase in the distance to Phragmites. We assume that this change in value happens immediately when the Phragmites is removed. In reality, the change in value, and in property tax revenue, is captured when the home is sold.
Changes in the property values due to Phragmites removal will affect property tax revenues. Property taxes in Michigan are calculated based on millages (1/1000 th of a dollar) and the taxable value is, by Michigan law, not more than 50 percent of the sales price. In the base case, we assumed that all homes were primary residences (homesteads). Area homestead millage rates for 2014 ranged from 26.51 in Grand Haven Township to 39.69 in the Village of Spring Lake (Michigan Department of Treasury 2015). For example, a house in Grand Haven Township that sells for $200,000 would have a taxable value of $100,000. At the millage rate of 26.51, the homeowner would pay $2,651 per year in property taxes.
The average homestead millage for the five municipalities was 32.14, which was applied to one-half of the sales price. Millages are higher for second homes, rentals, and businesses. Non-homestead millages ranged from 44.69 (Grand Haven Township) to 57.87 (Village of Spring Lake) with an area average of 50.29 (Michigan Department of Treasury 2015). The owner-occupied housing rate for Ottawa County Michigan (2009-2013) was 78.1 percent (US Census Bureau 2014). The remainder can be assumed as non-homesteads (rentals or second homes). The alternative model used a weighted average of homestead and non-homestead millages applied to one-half the sales price to estimate the property tax revenue change.

RESULTS
The Moran's I test for spatial autocorrelation showed that the log of sales price was significantly autocorrelated (Moran's I = -0.57, p<0.01). The negative Moran's I statistic indicates that high value properties are more dispersed than would be expected under a random distribution. We used error clustering to generate robust coefficients that account for the spatial effects (Heintzelman and Tuttle 2012).
Distance to Phragmites (dist_phrag) was statistically significant after controlling for the effects of the sales year (Table 2. Fixed Effects Regression Results.). The rho metric of intraclass correlation indicates that about 55% of the variance is due to difference across groups. The regression results indicate that a one-meter increase in distance to Phragmites is associated with a 0.0002 change in the natural log of the price, that is, a $3.90 change in the sales price (Table 3). Removing Phragmites from a property so that the next closest Phragmites patch is a quarter mile (400 m) away would lead to a sales price increase of more than $1,500. All Phragmites patches in the study area were less than 400 meters from the closest property.
The value of the change in distance to Phragmites was calculated for each house with Phragmites within 400 m (Equation 2). The total value of removing all Phragmites from the study area, found by summing all the per house values, was estimated at $837,391. This is assumed to be an immediate effect of removing the Phragmites. Assuming all the affected homes are primary residences and using the area average homestead millage (32.14), increasing property values by removing Phragmites would increase property tax revenues by $13,457 per year. Including the higher millage rate for non-homestead homes using a weighted average results in an annual property value increase of $15,121.  (Table 1), the cost of treating all of it would be $25,041 in the first year. Treatments in subsequent years would likely be less since there would be fewer patches of Phragmites to treat. This was the case in Spring Lake in which initial treatment in 2013 was followed by a modest treatment in 2014. In 2015, no treatment at all was needed.   (Doss and Taff 1996). Removing Phragmites therefore has a positive economic benefit not only to the property owner, but the entire community.
Higher property values lead to greater tax revenues which can fund additional Phragmites removal.
We found that removal of all Phragmites within the study area would increase annual tax revenues by $13,457-$15,121. Removing Phragmites, however, comes at a cost of about $687.47/ha. Removing all the Phragmites, about 36.42 ha, would cost $25,041. The cost of removal is just slightly less than two years of additional annual tax revenue. Since Phragmites, once controlled, does not need to be treated each year, the additional annual tax revenues can be put to other uses. These could range from lowering other taxes to improving infrastructure or maintaining key services. This suggests that TIF may be an appropriate tool for financing Phragmites removal.
This study has several limitations. The costs of Phragmites removal occur in the present, while the benefits from increased property values occur in the future after the properties have sold or are re-assessed. The full effect could take many years to be seen. The estimates of the property value impact come with some uncertainty. The years 2004-2015 included the inflation, bursting, and recovery of the housing bubble. While Michigan was spared the worst of the bubble's effects, there was substantial volatility in the Ottawa County housing market. We have taken steps to account for these swings, but the market instability could affect our estimates.
Repeat-sales models are limited to properties that have sold multiple times. It is possible that there is something unusual about a property that sells two or more times in a ten-year period. The median housing value in our dataset ($185,000) was slightly higher than that of Ottawa County as a whole ($153,000). This suggests that the properties in the dataset are representative of the county's housing stock. As noted previously, the repeat-sales method assumes that the condition of the home, such as renovation or deterioration, does not change between sales. Comparisons of repeat-sales and hedonic models that include a suite of housing and neighborhood characteristics show that repeat-sales models provide results that are consistent with hedonic models with housing attributes (Hansen 2009). We do not, however, have data that would confirm the stability in housing characteristics in our study area, and this is a source of uncertainty in the model. The temporal scope of sales, going back to 2004, assumes that the Phragmites stands were stable from 2004 to 2010. Inclusion of the early sales was necessary to create a sufficiently large dataset to run the model. Limiting the temporal scope would have resulted in an unworkably small dataset.

CONCLUSIONS
Phragmites can quickly dominate a wetland, which displaces native vegetation and disrupts ecosystem functions and services. Communities often struggle, however, with funding Phragmites removal programs, and the economics benefits of such removals are unclear. This paper shows that in Ottawa County, Michigan, home sales prices are negatively associated with proximity to Phragmites. Removing Phragmites and increasing the distance to the next closest patch raised property values at a rate of $3.90 per meter. This demonstrates that Phragmites depresses property values and homeowners have a positive willingness to pay for properties that are farther away from Phragmites. The total property value benefits of removing all Phragmites within 400 meters of all affected properties were estimated to be $837,391 once all the benefits are internalized into sales prices, which could take a decade or more. The increased annual property tax revenues ($13,457-$15,121/year) is about half of estimated Phragmites treatment cost ($25,041). That is, two years of additional property tax revenues would pay for the removal of Phragmites, which should last many years. The treatment would not need to be conducted annually, so treatment should have a positive net benefit to coastal communities.