Running on debt: Financing land protection with loans

Abstract

New land protection is expensive, and many conservation NGOs rely on loans to help fund land acquisition in the short term. Conservation loans are offered by a range of philanthropic organizations that often allow much more flexible terms than traditional loans. Thus, conservation loans may behave differently from other types of loan. There are costs and benefits to relying on loan financing to fund land protection that organizations need to consider, but few data are available to inform such evaluations. Here, we focus on estimating the financial cost of these loans, by analyzing loans used to support land protection projects that were provided through an internal revolving fund at a large U.S. conservation NGO. We estimate loan financing cost through accrued interest and test deal‐level characteristics for their ability to explain or predict loan interest. We find that loan performance can be highly uncertain and costs can be substantial in relation to the total purchase price. An improved ability to estimate the overall cost of conservation loans upfront may determine just which conservation projects are prioritized for investment and avoid costly misallocations of conservation resources.

INTRODUCTION

Land protection is a critical tool for conserving biodiversity. Unfortunately, conservation is chronically underfunded (McCarthy et al., 2012; Shaffer et al., 2002) and new land purchases are increasingly expensive (Davies et al., 2010). This funding deficiency is accentuated in situations where a large investment is needed quickly, as is often the case when a property comes on the land market for a short window. If this timing does not coincide with grants or fundraising seasons, then a conservation organization may be caught shorthanded. Preparing for such moments is critical for efficient decisions and action (McDonald-Madden et al., 2008). In some countries, NGOs active in land protection turn to loan financing options to supply the quick cash infusion that many conservation opportunities need. Taking a loan enables paying for a property over time with future income or recouping funds by selling the property into public ownership or to another conservation buyer.

Loans for conservation land acquisition are unique among real estate borrowing practices. Conservation loans are sometimes supplied by traditional lending organizations, such as banks, with set repayment schedules and contractually obligated interest. Using such traditional loans can make estimating borrowing costs straightforward. However, small non-profit land trusts can struggle to qualify for competitive interest rates or provide adequate collateral to secure such loans. Typically, the property itself would serve as collateral in real estate loans, but easements on conservation properties will lower the property value and not qualify as full collateral. Therefore, conservation groups often rely on loan programs specifically created for and targeted toward conservation and other charitable causes for which loan obligations may not be as restrictive. Various national and regional organizations manage revolving funds explicitly for conservation (Clark, 2007). Some, such as The Conservation Fund, the Open Space Institute, and the Maine Coast Heritage Trust, provide external loans (Clark, 2007). Other organizations manage internal revolving funds to provide a temporary funding source for their own projects, such as The Nature Conservancy's Land Preservation Fund (Birchard, 2005). Because these organizations have philanthropic motivations, they often allow much more flexibility in the loan terms (e.g., accepting unusual collateral types or early repayment options) (Clark, 2007). The more accommodating structure of such conservation-specific loans may result in differing performance from traditional real estate loans and warrants further investigation.

Loans are commonly used to support land protection efforts in a number of countries. In the United States (U.S.), land trusts play a growing role in establishing new conservation lands, often acquiring lands directly, creating easements, or facilitating public agencies in taking ownership of new protected areas by serving as temporary land holders (McQueen & McMahon, 2003; Merenlender et al., 2004; Land Trust Alliance, 2016). Over 50% of land trusts report in a survey conducted by the Conservation Fund as having used loan financing to assist in purchases (pers comm. Amundsen). Loans are also used by conservation organizations in Australia; Hardy et al. (2018a, 2018b) review the operation of Australian revolving loan funds active in land protection, focusing in particular on factors that affect what properties are chosen for protection. In addition, recent interest in conservation has arisen in the impact investment sector, which could provide a source of capital to expand the use of loans in conservation. In impact investment applications, individuals and private groups provide loans for habitat conservation for below-market-rate returns, with the expectation of additional social benefits (Hamrick, 2016).

Loans enable quick action when land protection opportunities arise allowing conservation organizations to benefit from particular moments in which a timely allocation of funds allows for efficient conservation gains (Larson et al., 2014; Radeloff et al., 2013). Loan financing also enables conservation organizations to aggregate funding in time allowing larger projects to be undertaken. Larger protected areas that minimize fragmentation and reduce edge effects offer important ecological benefits (Peres, 2005; Woodroffe, 1998), and can achieve cost efficiencies through economies of scale in acquisition and management costs (Armsworth et al., 2011; Kim et al., 2014). Loans can also allow conservation organizations to unlock new or additional sources of funding. The opportunity to provide loans directly, fund repayment costs or contribute to a wider financial management structure of which loans are a component, may appeal more to some potential conservation donors than funding traditional gifts. Also, Pinnschmidt et al. (2021) find evidence suggesting making investments in land protection, potentially funded by loans, may lead to greater philanthropic donations to conservation in the future.

These advantages of loans must be weighed against the added financial cost a loan imposes on a conservation organization in terms of interest payments over time. Debt repayments impose opportunity costs on conservation organizations because other opportunities may need to be passed over during the repayment period. In extreme cases, defaulting on a loan risks a conservation organization losing financial or other collateral used to secure the loan. There are many possible circumstances that can cause an organization to deviate from its loan repayment plan. Perhaps, a planned resell or transfer of the property falls through, a fundraising campaign does not bring in as much capital as expected, or external circumstances, such as a market crash lowers income to the organization and the value of the property.

Conservation organizations need to evaluate these pros and cons of relying on loans to fund land protection carefully. However, data and theory on use of loans in conservation to inform such evaluations remain scarce. Here, we focus on some of the costs of relying on loans to fund land protection. We investigate how conservation-specific loans impact the financial cost incurred by a borrowing organization. In particular, we ask whether known characteristics of a proposed land acquisition deal or a borrower explain variation in interest accrued over the lifetime of a highly flexible loan. We also offer an initial exploration of whether the magnitude of additional costs involved in relying on loans could be large enough to change project prioritizations. Our cost-side focus of this study complements other work examining the benefits of loans allowing quick movement of funds when opportunities arise (Lennox et al., 2017) or scaling up of land protection projects (Armsworth et al., 2018). Ultimately, better understanding of the factors that lead to excessive interest payments will help land trusts be strategic in the use of borrowing and improve their return on investment valuations.

DATA AND METHODS Case study

We use a case study of land acquisition deals to examine interest accumulation on loans for land purchases. That is, we examine the additional financial cost that results from relying on loan funding over and above the basic acquisition cost of land protection, something explored in other studies (Kim et al., 2014; Nolte, 2020). The Nature Conservancy (TNC) provides internal financing opportunities for state chapters through the Land Preservation Fund (LPF). The LPF was first established in 1979 as a $20 million internal revolving fund, and has since grown significantly (Birchard, 2005; Clark, 2007). Inside the United States, TNC is structured with state chapters that perform much of the on-the-ground conservation work. The LPF, managed at the TNC headquarters, provides loans to chapters pursuing large land acquisition projects.

TNC's LPF loan program is of particular interest because of the high degree of flexibility allowed to individual loans. As discussed earlier, increased flexibility is a common theme among conservation loans programs. These LPF loans can be paid off as quickly or slowly as needed. Therefore, sometimes no interest may be accumulated, or alternatively, a high amount of interest may accumulate if fundraising plans fall through. Eliminating some of the contractual obligations, however, means that simple assumptions about overall interest costs and times to repayment may no longer apply. Additionally, the use of a single-organization case study allows us to track projects from proposal stages through loan repayment within the same record-keeping and organizational structure.

We extracted characteristics of the deal and financial plans from a set of land deals acquired between the years 2000–2011.This decade witnessed a variety of economic conditions, including two recessions followed by periods of economic expansion. In fact, the LPF interest rates tracked these external market conditions, with peak annual interest rates coinciding with high points in the U.S. prime rate (years 2001, 2007, and 2008). Therefore, this time period should encapsulate a variety of behaviors for pursuing loans, while also allowing ample time for loans to be repaid. The variable interest rates for LPF loans are set annually and are equal for all outstanding accounts. The interest rate ranged from 4% to 7.5% across the years examined.

Sample

We focused on 181 projects spanning 24 TNC state chapters (median of 6 loans per state, Figure 1) that all had purchase prices of at least $1,000,000 and both requested and received a loan account from the LPF. We used only projects where a single financial transaction could unambiguously be associated with an individual land deal (e.g., excluding instances in which a single loan was used to fund a growing set of land acquisitions through time). The minimum purchase price we consider $1,000,000 served as an organizational threshold for TNC during the time period over which these loans were made. The organization required greater documentation of proposed land deals from state chapters for purchases surpassing this threshold, which included narrative descriptions of the ecological attributes of the property and how protecting it would align with TNC's goals as well as greater detail regarding the state of project financing at the time the deal was developed and the regarding the proposed repayment plan. We extracted information on some of our predictor variables from this documentation. TNC land deals below this threshold would therefore not be able to provide the same predictive information but also likely were in less need of loan financing. Thus, this selection process excluded deals that only sought outside financing options or did not apply for a loan at all. By focusing on expensive deals that received internal financing, the selection pool may have a higher propensity for poor loan performance.

Response variables

We measured the total interest paid on each of these 181 loans. Start date was the purchase date and repayment date was defined as the first date in which the remaining loan balance fell below 0.1% of the purchase price (because these loan accounts could be used for project expenses after the initial purchase, they were not always zeroed out or closed). Interest accrued for the loan was defined as the sum of any interest charges listed in the account between these dates. To control for inflation and allow time consistent estimation, all monetary values are reported in 2010 USD (translated using Consumer Price Index from the Bureau of Labor Statistics). We used a regression approach to explain variation in total interest accrued across loans. The distribution of total interest across loans is characterized by the presence of a number of zeros (loans that were paid back immediately) and by a skewed distribution of interest among loans where some interest accrued (Figure 2). In light of this, we first ran models that considered all loans but treated interest as a binary (i.e., was some interest paid or not). Next, we restricted our analyses only to transactions where some positive amount of interest accrued. We used generalized linear models assuming binomial errors to explain whether any interest was paid and generalized linear mixed models (GLMMs) with log-link and Gaussian errors to explain how much interest was paid for those loans where some interest accrued.

Explaining variation in the response variables

We sought to explain how much interest was paid for loans using information that a decision maker would know ahead of time. Given the sample sizes (N = 181 for presence of a loan and N = 150 for loan amount), we limit the number of parameters in our base model (Model 1) to 9 fixed effects chosen a priori as those most likely to influence loan performance in order to prevent overfitting (Harrell, 2015). We identified likely predictors through conversations with practitioners experienced in managing conservation loans. Model predictors focus on project characteristics such as future land use plans or funds dedicated to the purchase. We hypothesize that projects with established take-out plans, meaning a strategy in place to resell or transfer the property to a partner organization, are less risky and would accrue less interest. Take-out plans varied in how they were configured—some included only a portion of the property, some involved TNC placing easements on properties before transferring management responsibilities, and so forth. Furthermore, we hypothesize that some characteristics of the borrowing chapter, like average expenditure and income, will also correlate to better loan performance. Additional explanations of a priori hypotheses are given in Table 1. Collinearity of predictors was checked using variance inflation factors and fell within tolerated ranges (all VIFs <4; James et al., 2021). For models predicting how much interest accrued, we also included a state-level random effect.

TABLE 1 Hypotheses motivating inclusion of particular predictor variables, including sign of association that would result

[Table omitted. See PDF]

We use these models to assess how different project characteristics affect the additional cost of relying on loans due to interest payments. We based our inferences on the significance, sign, and magnitude of different coefficients in the resulting model fits and, for the model of loan amount, on a comparison of the marginal and conditional r-squared values for the random effect. At the same time, we recognize other choices could have been made about just what variables to include. Therefore, we examine whether the inferences one would draw with this base model about loans would change if instead we had relied on three alternative model specifications that also consider the role of habitat types featured in the narrative descriptions detailing how protecting a particular property would align with TNC's goals (Model 2 in Table 3), type of take-out partner (Model 3) and the 2008 recession (Model 4). As an additional test of how robust our inferences are to the design choices we made, we fitted the same models (and same additional specifications) using an alternate response variable to measure loan performance, the total length of loan measured in days (see supplemental Table S1). Loan length may serve as an additional useful indicator of loan performance because even in scenarios where an interest-free loan could be obtained, holding debt for long periods of time still may incur an opportunity cost to conservation by making other acquisitions less likely in the interim. This approach to model selection of fitting a full model motivated from a priori hypotheses and then testing whether the inferences we would draw from it would have changed had we specified the model differently or measured loan performance differently is common in economics and policy sciences (Armsworth et al., 2009). We recognize, however, that other approaches to model selection, including approaches that emphasize parsimony more, would also have been appropriate.

Potential effect on project prioritization

We also undertook an initial exploration of whether the additional costs incurred as interest payments on loans may be large enough to change what conservation projects are prioritized when compared to prioritization approaches that only consider land purchase price. To do so, we compared rankings of the full set of deals, the first prioritization using only purchase price and the second using purchase price and the cost of interest payments on the loans. To retain a focus on the cost component of our analysis, we focused on a simple area-based benefit function and ranked all projects based on dollar per hectare protected.

RESULTS

Land acquisitions in the data set ranged from USD $1 to $81 million, although three quarters of purchase prices fell under USD $5.7 million (Table 2). Among states, Texas, Georgia, North Carolina, and Montana all stood out for having large areas acquired under loan transactions at a high cost (Figure 1). Most projects protected forested habitat. The median project area was 330 hectares. The average acquisition had 7% of the purchase price either in hand or pledged to the project at the time of acquisition. Some pledged funds covered the full cost of the project, which, if realized, allowed for a quick repayment. About 57% of projects had a take-out plan to resell or transfer the acquired parcel to another agency or organization. Over half of the take-out plans presented in the land deal proposals were with state government partner organizations.

TABLE 2 Summary characteristics of data for the response variable and predictor variables

	Loan response variable	Unit	Data summary
	Interest accrued	0 / 1	150 loans accrued some interest, 31 did not
	Total Amount of Interest	2010 U.S. Dollars	Quartiles [Q1, Q2, Q3] = [13,700, 114,000, 322,000]
	Predictor	Unit	Data Summary
Deal Characteristics	Purchase price	2010 U.S Dollars	[Q1, Q2, Q3] = [1.8, 2.6, 5.7 million]
	Funds in hand	% of purchase price	Deals where there are no funds in-hand: 72% Of deals with some fund in hand: [Q1, Q2, Q3] = [2, 9, 24]
	Funds pledged	% of purchase price	Deals with no funds pledged: 52% Of deals with funds pledged: [Q1, Q2, Q3] = [22, 49, 91%]
	Human land use, Recreational	1/0	46% of deals had some recreational use
	Human land use, Extractive	1/0	29% of deals had some extractive use
	Number of partner organizations	Count	[Q1, Q2, Q3] = [0, 1, 4]
	Take-out plan	1/0	57% of deals had a take-out partner
State Characteristics	State experience	Total chapter expenditure 2000–2009 (dollars)	[Q1, Q2, Q3] = [73, 107, 212 million]
	State fundraising	Total chapter fundraising 2000–2009 (dollars)	[Q1, Q2, Q3] = [27, 48, 63 million]
	Random effect, State	Categorical	24 states; Median 6 loans per state

Across all projects, the median interest accrued per loan was $114,000 (Figure 2, Table 2). The median time to repayment was a year and a half, with some loans taking up to 8 years. With longer loans, the interest rate experienced fluctuated prior to repayment (range 4%–7.5%). On average, interest reflected about 4% of the purchase price, but the range was large, and the data set included situations where loans amassed interest equivalent in value of up to 36% of the purchase price. Perhaps surprisingly, 17% of projects accrued no interest. These projects either had pledged funding, grants that came in quickly, or a take-out plan completed within the first month of the loan, so no interest payments occurred.

To determine correlates of loan performance, we tested the significance of predictors in our regression framework (Table 3). First, we considered what might explain whether any interest accrued or whether a loan was paid off immediately. Loans for larger transactions were more likely to accrue interest, while those with a greater percentage of the acquisition cost in hand at the time of acquisition were less likely to do so. While significance was marginal, the models also suggest loans that cover acquisitions of land where TNC plans to allow recreational uses (e.g., hiking, hunting, and fishing) or that are made by state chapters having stronger fund-raising histories may be less likely to accrue any interest. Of loans that accrued some interest, only overall purchase price predicted how much interest was accrued. Other variables we included (e.g., the number of partnering organizations or presence of a take-out plan for transferring the property to another organization) showed little association with the amount of interest accrued on this subset of loans. Overall, our model was able to explain 36% of the variation in how much interest accrued, in situations where at least some interest was paid, with this being almost entirely accounted for by the model's fixed effects.

TABLE 3 Regression results for four model specifications—Our base model (model 1), and three alternative models (models 2–4) that contain additional predictor variables. For each specification, results are given for a model including all transactions (N = 181) while treating the presence of interest as a binary response variable and for a model that only includes transactions where some interest was paid (N = 150) while treating interest amount as the response variable and including a state-level random effect. Significance levels: “^” .1; “*” .05; “***” .001

	Model 1				Model 2				Model 3				Model 4
	Binary interest		Without zeroes		Binary interest		Without zeroes		Binary interest		Without zeroes		Binary interest		Without zeroes
	Estimate	SE	Estimate	SE	Estimate	SE	Estimate	SE	Estimate	SE	Estimate	SE	Estimate	SE	Estimate	SE
log Purchase Price	0.63*	0.294	1.186***	0.159	0.783*	0.326	1.151***	0.164	0.635*	0.302	1.145***	0.156	0.634*	0.295	1.177***	0.161
% of Price In Hand	−2.666*	1.14	−0.287	1.219	−2.775*	1.183	−0.396	1.256	−2.692*	1.203	−0.269	1.209	−2.64*	1.143	−0.309	1.223
% of Price Pledged	−0.643	0.506	−0.573	0.385	−0.756	0.529	−0.524	0.392	−0.554	0.532	−0.596	0.392	−0.652	0.506	−0.568	0.387
Take-out Plan	−0.224	0.445	−0.283	0.304	−0.321	0.467	−0.243	0.306	−1.045	0.712	−0.14	0.475	−0.224	0.445	−0.289	0.305
Total Number of Partners	−0.012	0.1	0.006	0.072	−0.002	0.107	0.024	0.074	0.002	0.106	−0.022	0.073	−0.017	0.101	0.01	0.073
Land Use, Recreational	−1.109 ^	0.593	−0.442	0.395	−1.103^	0.646	−0.342	0.413	−1.03 ^	0.608	−0.357	0.404	−1.134	0.599	−0.419	0.4
Land Use, Extraction	−0.334	0.532	−0.206	0.402	−0.56	0.561	−0.26	0.418	−0.462	0.547	−0.207	0.408	−0.332	0.533	−0.202	0.404
State Expenditure	−0.001	0.002	0.001	0.002	0	0.002	0.001	0.002	−0.002	0.002	0.001	0.002	−0.001	0.002	0.001	0.002
State Fundraising	−0.015 ^	0.009	−0.005	0.008	−0.015	0.01	−0.003	0.008	−0.008	0.01	−0.006	0.007	−0.015	0.009	−0.005	0.008
Forest					−0.248	0.73	−0.116	0.376
Grassland					−0.153	0.772	0.484	0.436
Shrubland					−0.46	0.882	−0.457	0.514
Desert					16.353	1063.51	0.865	0.659
Wetland					0.775	0.714	0.16	0.361
Freshwater					−0.537	0.637	0.195	0.356
Marine					−0.726	0.907	−0.202	0.502
State Govt.									1.115	0.713	−0.382	0.456
Private/Non-profit									16.025	1002.387	0.784	0.537
Federal Govt.									1.796	1.241	−0.667	0.544
Local Govt.									−0.356	0.84	0.352	0.605
Post-recession (purchase after 2008)													0.174	0.572	−0.169	0.414
Random Effect, State	—		Var: 0.108	SD: 0.328	—		Var: 2.858	DS: 1.691	—		Var: 2.773	SD: 1.665			Var: 2.752	SD: 1.659
Marginal R-sq.	0.139		0.336		0.165		0.346		0.194		0.354		0.138		0.335
Conditional R-sq.				0.361			0.346			0.356			0.36

Our robustness tests checked how these findings changed when augmenting the original set of predictors variables with additional predictors describing habitat type (Model 2), type of take-out partner (Model 3) and a dummy variable controlling for any change in behavior before and after the 2008 recession (Model 4). None of these inclusions improved significantly upon the performance of the basic model and the only changes to inferences one would draw about other predictors was that the marginal significance of a state chapter's fundraising history on whether any interest accrued on a loan disappeared (Table 1). Alternate model specifications explaining variation in the total length of loan (days until loan repaid) again emphasized the role of purchase price. Loans for projects with larger purchase prices were less likely to be paid off immediately and also tended to take longer to be repaid. Loans where TNC planned to allow recreational uses of land were more likely to be paid off right away, something less common after the 2008 recession. However, the role of percent acquisition cost in hand was less clear in influencing repayment period (Table S2) and the amount of variation explained by the fixed effects in the models was lower.

When ranking projects by purchase price alone versus ranking them based on purchase price and the cost of interest on the loan, 75% of projects changed rankings with some deals shifting up or down in rank by two positions.

DISCUSSION

We analyzed land acquisition deals financed through a major land trust's internal revolving fund to estimate costs involved in relying on loans to fund land protection. We show that costs associated with loan financing can be substantial, contributing up to 36% of the purchase price to the total investment on the project in extreme cases. At the same time, we found that it is difficult to predict the overall cost of loans. Only some loans incurred any interest. Of those that incurred some interest, only the purchase price explained variation in how much interest accrued. Variation in loan performance metrics was not explained well by other characteristics of the deal itself nor of the borrowing state chapter. This poses a challenge for conservation organizations, because in order to decide whether a loan makes sense it is important to predict what the overall cost of the loan will be upfront.

Loan financing may further impact conservation planning decisions by changing prioritization of proposed deals. If loan costs are large enough and show different patterns to other upfront costs more traditionally used for ROI analyses (i.e., purchase price), then they may change recommendations about which projects should be a priority to pursue (Boyd et al., 2015). To illustrate, we included a simple ranking exercise for the deals we considered where we focused on overall area protected as a simple representation of the conservation benefit from each transaction. Although we showed in our model that purchase price and interest are highly correlated, loan financing costs proved large and varied enough to result in some changes to the relative priority of conservation projects. Therefore, had TNC not had funding to support all of these projects, consideration of interest costs likely would have impacted the “marginal deal” chosen for protection. Projects that were a top priority likely would have remained top priorities regardless, but those that were on the margin of acceptance could easily have changed position, when accounting for the costs of loans more fully. More refined ROI analyses would include richer representations of conservation benefits than just area, but the basic point revealed by this illustrative example would continue to apply.

At the time of requesting the loans we examined, TNC state chapters had to outline their proposed repayment plan. These repayment plans embed subjective predictions by TNC staff of what interest costs will result for a particular transaction. Promisingly, predicted interest in these plans is positively correlated with the interest that actually accrued for the transactions we considered (Pearson's r = .58, p < .001, n = 181; after log transformation). Moreover, only 15% of loans incurred more interest than the proposal writers anticipated. However, those that did incur more interest than expected tended to be expensive (median of $174,000 per deal in additional interest above expected values). Yet, when exploring other metrics of loan performance, such as whether the expected repayment time was met or the difference between anticipated and realized interest, we found little association with the covariates that we included in our models.

Our results here offer initial insights into the cost of loans for land protection and there is a lot of scope to extend and complement what we have done. We focused on hypotheses suggested to us by the literature and practitioners experienced in managing conservation loans and our models explain 36% of the variation in interest costs. However, the remaining variation remains unexplained. One potential avenue that might shed more light on this variation would be to explore differences between staff involved in loan applications in terms of their risk tolerances, experience managing loans or other factors. Our finding that transactions may be less likely to incur any interest when there is potential to support recreational activities on protected parcels was also intriguing and may highlight more reliable revenue streams are available when recreational uses are promoted. Future explorations of loan performance also consider how to treat situations in which a loan is requested and received but paid off immediately. This occurred for 17% of the loans in our data set. We chose to run two separate sets of regression analyses, one including all transactions that sought to explain whether any interest was accrued and another set predicting the amount of interest accrued when restricting attention to loans involving some interest. Of course, we could have chosen not to consider the zero-interest loans at all. We believe ignoring these zero-interest transactions when studying loan performance may be a mistake, because the financial security offered by having a loan available, may be important for getting a project off-the-ground and bringing relevant partners together, even if the financial support is not ultimately needed.

Working with one organization had the advantage of providing us access to a large sample of loans with consistently reported financial information on repayment rates and project characteristics. However, it would be clearly valuable to contrast our findings with comparable studies focused on other loan programs. For example, repeating analyses of this type for smaller land deals could help inform many smaller conservation organizations considering the benefits and costs of relying on loan financing. It would also be particularly valuable to compare our results here for an internal revolving loan fund with results from loans made by external organizations. Loans for conservation made by external organizations appear to exhibit very low default rates (Clark, 2007). This may indicate that conservation organizations have a higher aversion to missing scheduled repayments on external loans. Exploring this comparison further would be particularly useful.

The greater flexibility involved in internal loans of the type that we have examined may be useful for revealing other information. For instance, with less danger of penalties for poor loan performance, the financial costs of taking on debt may better reflect how conservation organizations perceive opportunity costs, something critical to evaluating how best to prioritize among conservation opportunities through time (McDonald-Madden et al., 2008). A final obvious extension would be to combine our work here on costs of loans with studies on the ecological benefits that loans for land protection provide (e.g., Armsworth et al., 2018; Lennox et al. 2016) and on the evaluation of which properties are prioritized for protection (Hardy et al., 2018a,  2018b).

CONCLUSION

Utilizing loans as a tool for financing conservation land acquisition is a widespread practice and critical to expanding capacity of non-profit land trust organizations. It is important to understand the inherent costs and risks associated with this mechanism as well as particular aspects of projects that predict these costs for the lending and borrowing organizations. We demonstrate the potential magnitude of loan-associated costs, which sometimes may be sufficient to change which particular conservation projects go forward. Only basic financial characteristics of the loan itself significantly helped explain the large variation in interest payments that we observed, which was independent of more nuanced characteristics of the land deals in question. While loans continue to be an important and beneficial tool for conservation, additional research is required to help nonprofits manage the risk associated with repayment costs of borrowing.

AUTHOR CONTRIBUTIONS

RF, PA, and JF contributed to the study conception and design. Material preparation, data collection, and analysis were performed by RF and SR with input from PA and JF. The manuscript was written by RF with all authors contributing to editing. All authors read and approved the final manuscript.

ACKNOWLEDGMENTS

This work is supported by the National Science Foundation Graduate Research Fellowship Grant No. DGE-1452154; additional support from The University of Tennessee, Knoxville. The authors would like to thank TNC staff, specifically Monica Garrison, for providing critical data sets and feedback. Thank you to A. Segovia for assistance in data extraction. Additional thanks to C. Dumoulin, D. LeBouille and C. Stachowiak for comments on this manuscript.

FUNDING INFORMATION

This study was supported by the National Science Foundation Graduate Research Fellowship Grant No. DGE-1452154 and the University of Tennessee, Knoxville. Author J. Fargione is an employee of The Nature Conservancy, the primary data provider and organization examined in this study.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Raw data used in this study are the proprietary product of the Nature Conservancy and subsequently cannot be publicly shared. All descriptive data of variables used (including distributions, medians, quartiles, etc.) can be found in the supporting information Table S1.