State Implementation of Federal Environmental Policy: Explaining Leveraging in the Clean Water...

Introduction

Much of the recent discussion of the redefinition of federalism, and concomitant changes in environmental policy implementation, is predicated on the assumption that states possess the resources to assume responsibility for the administration of federal programs (Lester & Lombard, 1990). While some states clearly do possess the necessary resources, other states clearly do not (Heilman & Johnson, 1991; Walker, 2000). Environmental programs represent a particularly important area of policy, because the effects of pollution do not stop at state (or national) borders. As Kraft and Vig (1994) have pointed out, environmental problems are truly "public" problems. When the national government devolves responsibility for these programs to the states, the ability of state governments to administer these programs and meet national environmental standards is brought clearly into focus.

A case in point is the Clean Water State Revolving Loan Fund (CWSRF)1 program established in 1987 by the Water Quality Act (WQA) of 1987 (RL. 100-4). Designed to replace the Construction Grants program, the CWSRF program provides states with capitalization funds to create loan accounts from which communities may borrow funds for wastewater treatment projects. The CWSRF program requires states not only to concern themselves with the administration of an environmental program to meet their needs for clean water but also to grapple with a complex loan fund model to provide the financial resources for the program.

This article examines factors pertinent to administrative and financial decisions made by states in the implementation of their programs. After 12 years of program implementation, the CWSRF program has reached a level of maturity that allows analysts to discern clear programmatic trends among the states. Specifically, this article examines the factors that cause states to leverage their CWSRF funds, and once that decision has been made, how much to leverage. We explore these factors through three sets of explanatory "lenses"-EPA's official version of the factors leading to leveraging, Lester's (1994) model employing state commitment to environmental protection and state institutional capacity, and a model of our own design to combine several different elements to explain leveraging.

This research is important for several reasons. First, the CWSRF program is truly an experiment in New Federalism (Conlan, 1988), as defined by the legacy of the Reagan presidency, and thus allows us to discern the viability of the precepts of Reagan federalism. second, this research addresses directly the resource allocation decisions made by states to meet federal programmatic requirements. The CWSRF program is the only existing federal program to address water quality needs, and it has been one of the more "big ticket" environmental expenditures during the past decade. The factors that influence allocative decisions are thus brought sharply into focus. Finally, it provides an empirical test of both EPA's understanding of the program, as well as Lester's (1994) contention that state commitment to environmental protection and institutional capacity to achieve environmental goals can explain environmental outcomes in the states.

The CWSRF Program

The federal government has been an important source of funds for wastewater treatment facility construction since the 1950s. The Federal Water Pollution Control Act of 1956 provided direct grants to communities to fund water infrastructure projects (Freeman, 1990). Congress established the Construction Grants program through the Federal Water Pollution Control Act Amendments of 1972, which provided communities with funds directly from the federal government to pay for construction and associated costs of building wastewater treatment plants. The federal government led the way in water quality facility construction by supplying 75% of the construction costs directly to local governments. The federal appropriation for water quality projects rose because of this throughout the 1970s. The stated intent of Congress was that available federal funds would supplement state funding efforts for water quality needs.

By the 1980s, the level of funding and needs associated with this program were increasing, and the estimated costs of future construction were increasing sharply (General Accounting Office [GAO], 1992). In addition, the political changes signified by the election of Ronald Reagan had taken hold. There were two important parts to this movement. First, Reagan advocated a return, or devolution, of programmatic authority to the states. The argument was that states were willing and able to assume responsibility for the implementation of national policy. second, budgetary pressures created by the combination of increased federal spending and a broad tax cut had, by the mid-1980s, led to significant pressures to reduce the size of the federal budget by moving costs off-budget.

A working group created in 1984 by the U.S. EPA to study the agency's role in local water quality reported that state and local governments were not searching for other ways to fund water quality programs, other than the Construction Grants program. Another concern involved user fees; local governments were not raising their user fees to maintain water quality systems in the communities. In response, Congress reduced the level of federal funding through the Construction Grants program to 55% of the eligible construction costs (GAO, 1992). After considering several options, the working group recommended a new funding mechanism to replace the Construction Grants program. This mechanism, a revolving loan fund program, would be funded initially by Congress and administered by the states (Environmental Protection Agency [EPA], 1984).

In 1987, the Water Quality Act was passed with language creating the CWSRF. The CWSRF was created to provide a base of funding for states to create a program to address the costs of building and upgrading wastewater treatment systems to meet water quality needs in their state. By the late 1980s and early 1990s, communities would begin replacing the water treatment plants built in the 1970s, and Congress knew the demand for grant assistance would continue to grow. The program allotted each state a federal appropriation for 6 years (1988-1994) to build a sustainable CWSRF program. By law, each state was required to add a 20% match to their CWSRF loan pool each year. The key to the program, and its innovative idea, was that the state was allowed to create any type of loan program it wanted to address the building and upgrading of water systems in the state. The legislation was intended to stimulate a fund in perpetuity (GAO, 1992) in the state to meet local water quality construction needs. The appropriation was the only funding the federal government intended to give to the states for water system improvements. Through extended federal involvement, the federal appropriation for the CWSRF program is currently scheduled to continue until FY 2004.

Each state provides loans, with an interest rate determined by the state CWSRF program, to qualified communities ready to proceed with qualified projects.2 In turn, the loan repayment stream recapitalizes the CWSRF pool which is then lent again to other communities seeking funding. By combining interest payments along with loan principal, the value of the fund would continue to grow, thus ensuring a funding source to address future water quality needs in perpetuity.

The CWSRF program, while not the only wastewater funding source available, is the largest federal funding program. As such, it is often the largest program available to communities seeking public-sector funding assistance. Some states have separate state programs that are larger (better capitalized, provide more loans or grants, etc.) than the CWSRF program in that state, and several states have combined the CWSRF with state programs (although the funds must be accounted for separately). In most states, however, the CWSRF represents the largest public source of wastewater infrastructure funding. Many states provide "packages" of assistance to applicant communities consisting of funds from several sources (both federal and state funds), particularly for small or disadvantaged communities.3 In these instances, the funding drawn from the CWSRF often represents the bulk of the funds made available.

Leveraging

One of the important concepts in the CWSRF structure is the concept of leveraging. The EPA working group realized early on that the level of initial funding available from Congress would not be sufficient to meet current needs, much less future water quality needs. After the passage of the Water Quality Act, the EPA encouraged states to adopt two strategies in the CWSRF development: to contribute more than the minimum required 20% state match and to leverage their CWSRF monies as a way to generate additional funds for loans. Leveraging is a process through which a state places its annual federal appropriation (and perhaps its state match funds) in a reserve pool to guarantee the sale of revenue or general obligation bonds. The additional funds are then placed in the CWSRF fund pool and are available for loans. Because of a high level of state discretion in the implementation of the CWSRF program, states are free to choose not only whether to leverage but also what form of leveraging to employ.4 Some states engage in aggressive leveraging that although financially more risky, can generate upwards of five dollars for every dollar of initial investment. On the other hand, states may opt for a safer leveraging strategy, with the tradeoff of fewer dollars available (1.5 to 1 or 2 to I).

Although leveraging can be used to increase the pool of available funds, it is not without costs. First, leveraging mechanisms often involve extremely complex financial arrangements (Heilman & Johnson, 1991). Such complexity often requires highly trained financial expertise, which must be obtained from other agencies in state government or from the private sector (Morris, 1996, 1997). second, because leveraging involves the sale of state bonds, the bonds must be paid back to the bearers in the future, including the interest due on the bonds. The additional interest due on the bonds means that the CWSRF funds generated must be paid back by communities at an interest rate that not only stays ahead of the prevailing rate on inflation, but also covers the cost of the bond interest (Holcombe, 1992). This has the effect of increasing the costs of a CWSRF loan to communities (unless the state subsidizes the loan interest rate, which in turn has negative consequences for the long-term viability of the fund).

The implication of leveraging is that it may involve a tradeoff between meeting the demand for loans for wastewater needs (and environmental need) and the longterm viability of the CWSRF fund (a financial need). Because Congress explicitly directed that the CWSRF exist in perpetuity, states must be cognizant of the longterm viability of the fund. If the long-term viability of the fund becomes paramount, the cost of assistance to communities will be higher, because interest rates must exceed the prevailing rate of inflation. On the other hand, if a state chooses to leverage to meet significant environmental need, then it raises the costs of assistance to communities and results in fewer dollars available to meet future needs. States thus must balance several opposing forces-keep loans affordable but keep ahead of inflation, meet current needs but preserve the fund in perpetuity, and meet environmental needs (both current and future).

An important aspect of leveraging is that it significantly complicates the implementation structures required for the program, in that leveraging requires substantially more involvement by financial actors to design, implement, and administer the fund. Heilman and Johnson (1991) and Morris (1996, 1997) have described the tensions within the implementation structures that result from the pressures to meet environmental needs and maintain the long-term financial viability of the fund. When states leverage, the pressure to maintain financial viability grows; that pressure may make it more difficult for states to meet their environmental needs in the long-term. The analogy of a child's seesaw is instructive here: as one side rises, the other falls. Thus, as states become more involved in complex financial arrangements (such as leveraging), their ability to meet environmental needs may suffer. Part of the decision calculus for states is the degree to which they choose to focus on the financial or environmental aspects of the program, while meeting the host of tradeoffs discussed above.

Previous research (Heilman & Johnson, 1991; Morris, 1994) also indicates the decision to leverage is often a complex process on the part of states and may involve a calculation of risk, perceived long-term and short-term benefit, demand from communities for loan assistance, water quality needs in the state, and myriad of other factors. At the same time, the EPA (and, tangentially, Congress) has a less complicated view of the leveraging decision: states make near-term leveraging decisions based on demand and long-term decisions based on state water quality needs (see Heilman & Johnson 1991, chaps. 4-6). Taken together, the decision to leverage can thus be defined as a two-stage process. The first decision is whether to leverage at all. For states that decide to leverage, the second decision involves a question of scale-how much (or how aggressively) to leverage.

A cursory glance at both previous survey and case study data (Heilman & Johnson, 1991; Morris, 1994, 1997) fails to detect a clear pattern of leveraging or a consistent pattern of decision making among the states. Furthermore, there appears to be a discrepancy between the EPA's position on leveraging and the factors deemed important by states. This article seeks to explore the nature of the leveraging decision as a two-part process and to discern the factors important at each stage of the decision process. We thus turn our attention to a discussion of the methods employed in this research.

Methods and Data

Is leveraging a response to the needs and demands for funds found within states or is it the product of other factors? To answer this question we analyze the CWSRF program for the years 1993-1999. The year 1993 is used as the starting point for two reasons. As noted earlier, the CWSRF program did not begin until 1988, and the first few years of the program were spent establishing the basic infrastructure for full implementation, a process that was not consistent in timing across states. Also, 1993 is used as the starting date because certain variables are unavailable prior to this year. A portion of our data comes from surveys conducted in 1994 and 1996 of all 50 CWSRF programs in the country. Similarly, 1999 concludes the available data. Even with limiting the years under study, we are still faced with data limitations. Thus, we include only the 43 states for which complete data are available. Data not received through the state surveys were collected from the Office of Wastewater Management at the EPA during the summer of 2000. Other variables used to examine leveraging are from other published research. Table 1 reports the simple correlations among all variables as well as listing the means and standard deviations of the continuous and the distribution of the discrete variables.

In selecting this time period, we do not use the data on a yearly basis. Rather, we pool our data into two time periods, 1993-1996 and 1997-1999. A variety of factors explain this decision. First, there is the problem of the federal and many individual state fiscal years being out of sync; the federal fiscal year begins on October 1, but many states begin their fiscal year on July 1. second, the nature of the federal appropriation contributes to this decision, because the federal government, after making its annual CWSRF appropriation, allows states up to two years to raise its 20% state match. Many states take advantage of this provision and follow a nonyearly process of receiving and matching CWSRF funds. Thus, state budget years often exhibit yearly fluctuations in CWSRF funds that are a function of this matching process. By averaging over a longer period rather than using yearly data, we are better able to estimate the true level of funds devoted to CWSRF appropriations and leveraging.

Finally, our decision to break the data into two time periods rather than average over the whole time period is impacted by the EPA's process of calculating the level of need of each state for CWSRF funds. The EPA only calculates need on a quadrennial basis with the two most recent needs surveys being completed in 1992 and 1996. Because we use level of need as one of our independent variables and because, based on the EPA needs surveys, states have the same level of need for the 1993-1996 and 1997-2000 cycles, we opt to take the average of our yearly data over the 1993-1996 and 1997-1999 periods.5 The two time periods, combined with our 43 states, give us a total of 86 observation periods.

In examining the data, we are also struck by the fact that many states never choose to leverage, and among those states that do leverage, they tend to leverage differing amounts of funds. For instance, of the 43 states, only 17 leverage during both time periods, whereas two more states leveraged during only one period, leaving us 19 leveraged states. This occurrence is reminiscent of findings in other policy-related decision-making arenas. In the foreign aid budgeting literature, several authors have identified a two-stage budget decision making process (Cingranelli & Pasquarello, 1985; Poe, 1991a, 199Ib). The basic contention of this two-stage decision-making process is that during the first stage, the United States must decide to whom to allocate assistance, and during the second stage it must consider how much aid to allocate. In the political action committee (PAC) formation literature, a similar phenomenon is observed in which the decision of a business or other interest group to form a PAC is viewed as being distinct from the decision of how much money to contribute to the political process via a PAC (Grier, Munger, & Roberts, 1994; Hart, 2001; Mitchell, Hansen, & Jepsen, 1997). The results for the PAC literature indicate that the determinants of the first, or PAC formation, stage are different than the determinants of the second, or PAC size, stage.

In keeping with this distinction, we argue that decisions to leverage CWSRF funds can be viewed as a two-stage process. The first stage can be characterized as a simple yes-no decision about pursuing a leveraging strategy. We classify all states into either a leveraged or nonleveraged state, where a state that leveraged in at least one of the time periods is defined as a leveraged state. Then, we create values of 1 and O that correspond to the leverage status of a state. For those 19 states that leveraged during at least one time period, the second stage consists of deciding on a dollar amount to leverage. The process of selecting these 19 states based on the prior decision to leverage or not leverage leads to the potential of selection bias. Selection bias occurs when certain observations are systematically included in the second stage based on the first stage selection process. This can lead to biased and inconsistent estimates at the second stage because the analysis is conducted on a censored sample. A standard solution to this problem is the Heckman two-step selection model (Little & Rubin, 1987). In the first stage, probit analysis is used to predict whether or not states choose to leverage CWSRF funds. This first stage is also used to produce an Inverse Mills Ratio, a measure of the selection bias that is inversely related to the probability of being observed. This selection bias term is then used as a continuous term in the second stage to produce a modified ordinary least squares (OLS) regression model which provides consistent estimates. Still, potential problems loom from this correction method including the production of inefficient estimates and the introduction of measurement error problems (Kennedy, 1998). To increase confidence in our results, we report second-stage results for both OLS and Heckman-corrected OLS models. We use STATA 7.0 to estimate all models for the 43 states included at the first stage and the 19 states included at the second stage. The Appendix lists the states included in both the first- and second-stage analysis.

One other potential problem with the use of the Heckman two-step model is the estimation of models that include either an identical set of regressors in each step or that include only variables in the selection model that are a subset of the variables found in the second stage of the model (Little & Rubin, 230). To counter this problem, and to further set the decision-making context, we employ a variable in the first stage that examines state enabling legislation to determine if this legislation provided specific legal authority to the implementing agency to engage in leveraging. To examine the legal and constitutional support for leveraging in the CWSRF program, we conducted a content analysis of the enabling legislation creating the program in each state.6 Thirty-one states provide explicit authority for leveraging, whereas 19 states do not mention the financial option of leveraging in the CWSRF legislation; that is, they neither provide nor deny the ability to leverage. None of the states explicitly prohibit leveraging in their enabling legislation. We would expect that states that granted explicit authority to leverage had more understanding in the CWSRF program and how their state would pursue implementation.

A Test of EPA's Model of Leveraging

Our first hypothesis examines the relationship between needs of each state and the likelihood/level of leveraging that occurs in the CWSRF program. According to the EPA, the higher the level of need identified the more funds that state will be appropriated for water quality projects. As noted above, though, the federal appropriation often does not cover a significant percentage of the cost of state's needs; therefore, leveraging may be seen as a viable option to address the immediate environmental needs of communities. Even though tension exists between maintaining the long-term financial viability of the fund and meeting short-term needs, we expect the amount of funds leveraged by states to be higher in states with greater environmental need. Need is an interval level measure of the costs states can expect to incur to address water quality needs in the next twenty years. These data are collected in the CWSRF surveys and from the EPA Office of Wastewater Management.

In the CWSRF program, a greater demand from communities for loans will increase pressure on the state programs to provide loans. In states where communities are applying for loans in high numbers and demand is high, it is expected that states are compelled to leverage their program appropriation to create more funds to meet the existing demands, and needs, in their state. During interviews conducted with EPA officials from the Office of Water in 1991 and 1997, nearly all respondents indicated that the decision by states to leverage is a product of needs and demands for assistance.7

Measuring demand for CWSRF assistance is fraught with difficulty. An obvious first choice might be to use the number of applicants for assistance on a state's priority list each year. However, two factors mitigate against this. First, a priority list is as much a measure of need as it is a measure of demand. Communities are placed on a priority list because of their severity of need, not because they desire CWSRF assistance. Second, inclusion on a priority list may occur years before a community is actually ready to apply for CWSRF assistance. States typically do not keep accurate historical records of those applicants actually ready to proceed with a project in a given year or the number of formal inquiries for CWSRF assistance, making time series analysis difficult at best. Other measures, such as the number of potential applicants in a state or the size of competing state programs, are also replete with potential validity problems.

To this end, demand for the CWSRF loans is measured from a question found in the 1996 CWSRF survey sent to the state CWSRF coordinators.8 The question asks, "How would you characterize the demand for SRF assistance over the past five years?" The measure categorizes demand to be very low, low, high, and very high. In our study, we treat this variable as an interval level variable.9 Case studies confirm that some states with a low demand for the CWSRF loans have chosen not to leverage their funds (Morris, 1994). However, for states that have chosen to leverage, the higher the demand from communities, the more funds should be leveraged to meet the demand, the EPA vision of the CWSRF program. Finally, we include the variable representing the legal authority given states to leverage.

The first model that we examine includes only these three independent variables, NEED, DEMAND, and LEGISLATION, which are purported to be directly responsible for the decision to leverage. In Table 2 we notice that the variable measuring explicit legal authority to leverage is a significant determinant of the decision to leverage. Clearly, it seems that many states, in writing their enabling legislation, had already determined that the leveraging of funds was either necessary or desirable and had made sure that such authority existed. We find no evidence, however, to support the hypotheses that either needs or demand are related to the decision of whether to leverage or not. Yet, in Model 2, we do find evidence that need is positively related to the decision of how much to leverage. Finally, the nonsignificant selection bias term in the Heckman corrected equation indicates that this model does not suffer from selection bias. On the basis of these results we are left with a bit of a mystery as to what does explain the decision to leverage. A model based solely on the EPA vision of the leveraging decision is susceptible to problems inherent from model underspecification. Thus, the need to find a better explanation of the driving forces behind the leveraging decision remains.

A Test of Lester's "Capacity/Commitment" Model

James Lester (1994) provides a useful summation of much of the work done to explain state variation in the implementation of environmental policy. The severity argument is predicated on the differences in the degree of environmental problems within states. States with more pressing environmental problems will move quickly, whereas states with less severe problems lack the motivation to move quickly. However, Lester also states that previous research indicates that "the correlation of severity to protection is not clear, and that more refined indicators of pollution severity . . . are needed. . . . Factors other than problem severity (such as politics or economics) appear to be affecting states' behavior in this area" (1994, p. 59). On the other hand, Lester argues that the wealth argument can explain a significant amount of the variation in state protection efforts. The partisanship argument, according to Lester, is the most common argument found in the literature (1994, p. 60)-that ideology or political party identification can explain implementation variation. Calvert (1989) has found some evidence for this explanation. Finally, the organizational capacity argument suggests that factors such as gubernatorial control, professional legislatures, and consolidated environmental bureaucracies (Lester, 1994, p. 62) can be used to explain state variation in implementation, and there is some evidence to support this view.

However, Lester also argues that none of these arguments are sufficient to explain "the multiple forces that influence the states' commitment to environmental quality" (1994, p. 62). Instead, state policy activity is influenced by a rich intergovernmental framework that includes these state factors, as well as federal-level inducements and constraints (see also Goggin, Bowman, Lester, & O'Toole, 1990). Given the movement toward devolution and decentralization of federal environmental policy, he argues that a state's willingness to implement federal policy is fundamentally the result of two factors: the state's overall commitment to the environment, and the institutional capability of the state to implement complex federal programs (Lester, 1994, pp. 62-63). By measuring these two factors for each of the 50 states, Lester creates a typology to describe the ability of states to respond to federal environmental programs. "Progressive" states are those with a high commitment to environmental protection and a strong institutional capacity; they include states such as California, Florida, New Jersey, and Wisconsin. The "struggler" states have a strong commitment to environmental policy but lack a degree of institutional capacity. States such as Delaware, Maine, and North Carolina are found in this category. The "delayer" states have strong institutional capacity but lack a commitment to environmental protection. This category includes states such as Alabama, Illinois, Texas, and West Virginia. Finally, the "regressive" states lack both a strong commitment to the environment and the institutional capacity to carry out federal policies; states in this category include Arizona, Indiana, and Wyoming (1994, pp. 63-64).

We may thus use the leveraging decision within the CWSRF program as a test case to determine the usefulness of Lester's typology. Because leveraging requires significantly greater institutional capacity to implement (see Heilman & Johnson, 1991; Morris, 1997), we should expect to find states with stronger institutional capacity to be more likely to leverage. By the same token, the decision to leverage may threaten the long-term viability of the fund. States with a commitment to environmental protection may thus be less willing to risk the long-term solvency of the fund. We might then expect states more committed to environmental protection to be less likely to leverage.

In this model, institutional capacity is measured using the interval-level data from Bowman and Kearney (1988), as suggested by Lester (1994). The score for each state's INSTITUTIONAL CAPACITY was calculated by adding together the factor scores for the four dimensions measured by Bowman and Kearney10 (staffing and spending, accountability and information management, representation, and executive centralization). State commitment to environmental protection is operationalized using two measures: a measure of a state's overall commitment to environmental protection (Davis & Lester, 1989), which we label DAVIS/LESTER, and a measure of commitment to clean water developed by the Fund for Renewable Energy and the Environment (EREE), as reported in Lester and Lombard (199O).11

Our results, displayed in Table 3, indicate that Lester's contention that environmental commitment and institutional capability will mark state's response in the 1990s to new federalism has little ability to explain the yes/no leverage decision. At the second decision-making stage, however, we do find limited support for Lester's argument. The results in Model 4 indicate states with a greater institutional capacity are more willing to pursue larger leveraging programs. Because of the complexity involved in leveraging, this capacity further enables these states to pursue a leveraging strategy. Because of results reported below in Table 4, where we present a more fully specified model, we do not attribute great importance to this finding.

We also discover that the DAVIS/LESTER index measuring environmental commitment is negatively related to leverage size. Although seemingly odd at first, this negative relationship is consistent with our argument that leverage decisions involve a trade off of meeting short-term needs at the peril of risking the long-term health of the CWSRF program. States that are more committed to environmental protection are less likely to take this risk. In sum, it appears there is some support for Lester's model, but only after the initial decision to leverage has been made. An explanation for the initial yes/no decision making remains problematic, as neither the EPA's vision of the world nor Lester's notions of commitment and capacity explain this decision satisfactorily. We thus turn our attention to a model encompassing additional explanatory capabilities that overcomes any shortcomings due to model underspecification.

An Expanded Explanation of Leveraging

To seek further insight into the leveraging issue we investigate other explanations. An additional explanation is the impact of the size of a state's CWSRF appropriation on leveraging. The total appropriation that states receive in each appropriation cycle is dependent upon an allotment formula found in Section 206 (a) of the Water Quality Act (PL. 100-4). Although the law does not specify the elements of the formula, it extends the existing formula found in the Federal Water Pollution Control Act. The formula, although ultimately arrived at through political negotiation (see Dilger, 1986), takes into account water quality needs and population in the state. Therefore, we expect to see a positive relationship between total appropriation and the choice to leverage program funds in order to address more water quality needs in a state. States with higher total appropriations that have chosen to leverage should also leverage a greater amount of funds. The TOTAL APPROPRIATION variable is measured by combining the eighty percent federal appropriation and the twenty percent state match for the CWSRF program required by law.

We also examine the program emphasis of each state program as an influence on the state's decision to leverage CWSRF funds and the level of leveraging that occurs. States that emphasize the environmental aspects of the CWSRF program may not choose to leverage their funds, even if it means addressing less water quality needs in the short-term, because of the concern that leveraging funds in the short-term means less funds to lend in the long-term. The consequence of this decision means that the same level of funding would not be available in the short-term, but that there would be guaranteed loan funds in the future of the CWSRF program. Conversely, states that have a greater financial emphasis in their program will leverage their CWSRF funds more often. Leveraging provides more funding resources to the CWSRF program to lend to communities in the short term. Therefore, if leveraging is a financial option, one would expect states with a financial emphasis to leverage more of their program funds.

To analyze the influence of program emphasis in the decision to leverage and the level of leveraging, a variable was created from questions in the 1994 and 1996 CWSRF surveys. The survey questions asked, "To what degree have you emphasized FINANCIAL elements of your SRF?" and "To what degree have you emphasized ENVIRONMENTAL elements of your SRF?" Responses to the measure ranged from a major emphasis to a minor emphasis. When state programs determined what factor they would emphasize more, financial or environmental, the programs were choosing to trade off one policy area for another. Environmental emphasis placed the environmental needs of the communities first. In programs with a financial emphasis, loaning funds and community ability to pay the loans back top program emphasis. To measure emphasis in a state's program the financial emphasis response was subtracted from the environmental emphasis score. For this TRADEOFF variable, a positive score indicates a greater emphasis on the long-term environmental aspects of the program with fewer states leveraging and those that do, leveraging less than states with a financial emphasis.

We also examine the role that political culture (Elazar, 1966; see also Wirt, 1991) plays in the leveraging decision process. Miller (1991) finds that traditionalistic states follow different patterns of public expenditures12 at the state and local level than do moralistic or individualistic states, in that traditionalistic states tend to spend significantly less. Following Miller, we expect more traditionalistic states not to leverage their CWSRF fund, since leveraging represents a greater fiscal commitment. Traditionalistic states, with a more conservative view of policy, are less likely to take a risk in leveraging funds to meet water quality needs. The act of leveraging, as stated earlier, creates an imbalance in the interest earned from loan repayment and the market level interest bond repayment that must be paid back to bond holders. Fiscally conservative states are not likely to increase their short-term loan pool for long-term interest rate debt payment that is higher than the interest they are receiving from loan repayment.

Table 4 focuses on the basic decision that states confront when they must decide to either pursue a leveraging strategy or simply resort to the use of federal and state appropriations in an effort to meet needs. Model 5 in Table 4 depicts the relationship between the full complement of independent variables and the simple yes/no decision concerning leveraging. We again find no evidence that the decision to pursue leveraging is related to either the EPA's argument that level of need and demand are the rationalization. Concerning Lester's argument that commitment and capacity are considerations in the decision to pursue leveraging of CWSRF funds, we find mixed results. Similar to the results in Model 4 for the Davis/Lester variable, the evidence for the FREE variable indicates that higher commitments to environmental protection lead states to avoid the long-term risks entailed in leveraging. Yet for the capacity variable we find a negative relationship, which is opposite than expected based on Lester's argument. States with more capacity have more options and are thus less likely to accept the inherent risks in leveraging.

Focusing on the other variables, the findings for the relationships between tradeoffs, total appropriation and traditional political culture, and the yes/no leverage decision are consistent with our expectations. That these three variables are significantly related to the yes/no leverage decision indicates that a variety of state cultural and bureaucratic factors are important in the implementation of certain aspects of the CWSRF program. These findings further support a basic argument that ultimately the impact and implementation of federal programs must be viewed through the lens of federalism, where endemic cultural, institutional, and political factors matter (see Goggin et al., 1990). Finally, we also find that the legislation variable is consistently significant in all Stage 1 models.

The results displayed in Models 6 and 7 focus on the relationships between the independent variables and the actual averaged dollar amounts of leveraged funds for the 19 states identified as leveraging states. Model 6 displays the full complement of independent variables similar to Model 5. Model 7 excludes those variables that are not important in explaining the second stage leveraging decision while including two dummy variables used to account for the influence of two outlying observations in New York for the 1993-1996 period and Texas for the 1997-1999 period.13 In neither of the Heckman corrected models is the selection bias term significant thus indicating that selection bias is not a problem in examining the second stage results.

In both Models 6 and 7, we again find that the dollar amount of leveraged funds is positively related to the level of needs. These results confirm, in part, a basic argument of the EPA that leverage decisions are driven by needs. Among the other independent variables, we see that the level of leveraging is also dependent on the size of total federal and state appropriated funds, as indicated in Model 7, and that the tradeoff and culture variables are not related. These results, coupled with the positive relationships for needs, points to size as an explanation for the level of leveraged funds.14 Although at the first stage of decision making other factors drive the basic yes/no decision, at the second stage the level of funding is a product of size of funds available to leverage and size of need. Yet, concerning Lester's argument about capacity and commitment, support for those suppositions wanes, as only commitment as represented by the Davis/Lester variable is significant. In sum, in these models, when other variables are accounted for, Lester's basic argument retains only marginal importance as an explanation of leveraging.

Conclusions and Implications

The research described in this article supports the definition of the leveraging process as a two-stage decision-the decision whether to leverage, and the decision of how much to leverage. We also find that the factors important at the two decision stages are different. Likewise, the apparent position of the EPA regarding leveraging is not wholly supported by our findings. Although the EPA suggests the decision to leverage is based on a state's environmental need and demand by communities for loans, neither of these variables explain the decision to leverage in our model, although both play a role in the decision of how much to leverage. Indeed, it appears from our findings that the decision of how much to leverage, unlike the initial decision to leverage, is driven largely by measures of size-size of needs, size of demand for loans, and size of federal appropriations.

We find mixed support for Lester's capacity/commitment model. Lester suggests that state response to the devolution of federal environmental programs is a function of institutional capacity and commitment to environmental protection. We find some support for his argument. Depending on the model specification, we find that institutional capacity is significant at both stages of the decision process, a finding consistent with Heilman and Johnson (1991). A state's commitment to environmental protection is also significant, although marginally, at both stages of the leveraging decision. While we are by no means ready to declare Lester's model as fully supported, it is clear further research is warranted to better specify the range of dependent variables useful in this model.

There are at least three important implications to be drawn from this work. First, the decision of a state to leverage, as suggested by Heilman and Johnson (1991), is indeed a complex decision process driven by a range of factors. State discretion, one of the bulwarks of New Federalism, means that states make programmatic decisions on the basis of their understanding of their own particular context, rather than on the assumptions of federal actors. Despite the EPA's understanding and intent, states do not choose to leverage on the basis of environmental needs or demand for loans, but rather, a complex interaction of environmental, financial, and cultural factors. Needs are only important when states decide how aggressively to leverage. In this sense, states are truly "policy laboratories."

Second, our research highlights the tensions in the CWSRF program between environmental outcomes and the financial pressures to maintain the fund in perpetuity. One way to conceive of this problem is the tradeoff between short-term fixes and long-term outcomes. While leveraging may help solve short-term needs or demands, the long-term viability of a leveraged fund raises serious questions about the ability of leveraged states to meet future demand (Holcombe, 1992), and thus to meet federal water quality standards.

Finally, this research points to a larger implication for federalism and the devolution of program authority to states. We find little support for the rather simplistic assumptions of federal actors about the reasons states make the choices they do. The move to devolution was predicated on a belief that states were both able and willing to assume responsibility for the design and implementation of national policy and would make decisions based on their particular circumstances. The experience of the CWSRF program suggests that states have done the latter, sometimes at the expense of the former. In spite of federal pressure to leverage, comparatively few states have chosen to do so, and apparently for reasons other than those espoused by national policymakers. It remains the subject of further research to determine the degree to which this state of affairs is present in other targets of devolution.