Impact of Non-Water/Sewer Impact Fees on Multifamily Housing Construction in Inner Suburbs, Papers of Theatre

Download Impact of Non-Water/Sewer Impact Fees on Multifamily Housing Construction in Inner Suburbs and more Papers Theatre in PDF only on Docsity! THE EFFECTS OF IMPACT FEES ON MULTIFAMILY HOUSING CONSTRUCTION Gregory Burge Keith Ihlanfeldt DeVoe Moore Center and Department of Economics 150 Bellamy Building Florida State University Tallahassee, FL 32306-2220 Prepared for the 2005 Florida State University Critical Issues Symposium “State and Local Government Regulations and Economic Development” March 4-5, 2005. The financial assistance of the Federal Home Loan Bank of Atlanta is gratefully acknowledged. Abstract Development impact fees may create more housing opportunities for lower income households within suburban areas if there is a fiscal incentive behind the adoption of exclusionary land use regulations. Using panel data estimation techniques that allow us to control for unobservable heterogeneity and potential endogeneities, we estimate the effects of different types of impact fees on multifamily housing construction using data from Florida counties. Impact fees earmarked for public services other than for offsite water and sewer system improvements are found to expand the stock of multifamily housing construction within inner suburban areas. Water/sewer impact fees, on the other hand, are found to reduce construction throughout the entire metropolitan area. 1 theoretical perspective, while impact fees may indirectly reduce the total project approval costs imposed on developers by local governments, they also directly increase the developer’s building permit fees. The multifamily housing supply curve may shift inward or outward depending on whether the savings in project approval costs are smaller or larger than the fees themselves. Thus, a priori, the effect of impact fees on low income housing construction in the suburbs is ambiguous. Only empirical evidence can resolve the issue of whether impact fees will help or hurt lower income households find affordable housing in the suburbs. The purpose of this paper is to exploit a unique panel data base on impact fee usage among Florida counties in order to investigate the effects of different types of fees on multifamily housing construction. Separate models are estimated for central cities, inner suburbs, and outer suburbs. The panel nature of the data allow us to exploit a variety of different estimation techniques that control for unobservable factors and possible endogeneities that otherwise may have biased the results. These techniques include fixed effects, random trend, and stock-adjustment models. The results indicate that impact fees earmarked for making offsite water and sewer system improvements generally reduce multifamily housing construction. In contrast, impact fees designated for other types of infrastructure are found to increase multifamily housing construction within inner suburban areas, but have no effect within central cities or outer suburban areas. 4 II. Impact Fees and Multifamily Housing Construction Impact fees affect developers’ costs directly by requiring payment of the fee at the time the building permit is issued. Impact fees may also indirectly lower developers’ costs by reducing the numerous other costs, besides building permit fees, that developers incur in securing the right to develop. The total costs of obtaining project approval from the local government can be broken down into explicit fees (impact fees, environmental permits, plat application fees, etc.), compliance costs, and time costs. Compliance costs may include payments to engineers, surveyors, attorneys, and others in order to satisfy specific rules and regulations that govern the development.3 Time costs are incurred because it generally takes months for local governments to complete their review of project proposals. Multifamily housing projects are widely perceived as imposing a fiscal deficit on the community by increasing public service costs by more than the property tax revenue they generate. The fact that multifamily housing is seen as a free rider may explain why relatively little land is zoned for multifamily housing in the suburbs. This increases the compliance and time costs associated with multifamily housing proposals, because an “up-zoning” to higher density is frequently required. Depending on how high they are set, impact fees shift all or a portion of the financing burden from the overall community to the developer, who in the long run may shift the 3 Compliance costs have received scant attention in studies of the costs of land use regulations, but they may be quite substantial. For example, in Leon County, Florida, the site plan permitting process requires a natural features inventory (which charts all trees and animal features of the site), an environmental impact analysis (which gauges how the proposed development will affect the environment), and a traffic study (which estimates the effect of the proposed development on automobile congestion). Beal (2004) estimates that these requirements add $15,000 in total compliance costs to the cost of obtaining project approval for a limited partition subdivision (i.e., a subdivision of ten lots of less). 5 fees forward to tenants or backward to landowners.4 If the fiscal motivation for exclusion is important, communities with impact fees may zone more land for multifamily housing or make variances/rezonings to higher density easier to obtain. Impact fees may therefore generate savings in the project approval process that more than offset the direct costs of the fees themselves. The above logic, however, applies only to fees that are earmarked by local government for services traditionally funded by property taxes. The major services not funded by property taxes are water and sewer. For these services the financial burden borne by existing residents (in the form of higher base rates) from offsite water/sewer system infrastructure improvements necessitated by new development does not directly depend on the taxable value (per resident) of the development. For example, a high end single-family subdivision and a multifamily housing development may each require an additional pumping station and the same extension of water and sewer lines. Hence, if the goal is to avoid the need for any additional water/sewer infrastructure, all new development would be targeted, not just multifamily housing. However, most communities would not find it advantageous to oppose all new development in order to keep water/sewer user fees low, because development yields benefits (e.g., jobs and shopping opportunities) that exceed the costs of higher water and sewer charges. Water 4 As discussed more fully below, in Florida impact fees have not been set high enough to cover the full marginal cost of the additional public infrastructure needed by new development. 6 III. The Panel Data Set A complete history of impact fee rates was obtained by contacting county planning offices for all Florida counties. Our empirical investigation uses 33 of the 36 metropolitan counties in Florida that have ever imposed impact fees on multifamily housing developments.6 Impact fees in Florida are county-wide, but some cities also impose their own fees on top of those charged by the county. The city fees are in all cases small relative to county totals. Fees are used to fund a wide variety of government services, with fees for water/sewer, schools, and roads being the largest and most popular.7 Impact fees levied on multifamily housing are assessed on each apartment and may increase with the size and number of bedrooms of the apartment. Our impact fee variables – one for water/sewer and one for all other services – are based on a 1,000 square foot apartment with two bedrooms. Table 1 lists per apartment real impact fees (in 2003 dollars) for each county for the first (1996) and last (2003) years of fees used in estimating our models. The largest water/sewer fee is found in Indian River County in 1996 ($4,863 per apartment), while the largest total sum of fees for other services is found in Martin County in 2003 ($5,072 per apartment). Over the years covered by our panel, real water/sewer fees increased in 12 counties, decreased in 19 counties, and did not exist in 2 counties. For non-water/sewer fees, real values increased in 18 counties, 6 Using the most recent census definitions, there are 38 metropolitan counties in Florida. Clay and Jefferson Counties have not charged either type of impact fee and are dropped. Duval is dropped because the County and central city are consolidated governments, making the central city vs. inner suburban area breakdown we describe below impossible. Saint Lucie County is dropped because historical water/sewer impact fee rates are unobtainable. Finally, Alachua County is dropped due to a lack of consistency across tax rolls in the measurement of the square footage of multifamily housing properties. 7 Impact fee ordinances in Florida must satisfy the “rational nexus” test, which requires (1) a clear connection between new growth and the need for new capital facilities, (2) fees that are proportional to the costs of providing the facility, and (3) the payer of the fee benefit from the new public facilities. 9 decreased in 9 counties, and did not exist in 6 counties. As described below, estimates of the effects of impact fees on multifamily housing construction are based only on variation in fees within (and not between) counties; hence, the fact that this variation is nontrivial will help provide more efficient estimates. Although some Florida counties first adopted impact fees back in the 1980s, our panel data base is limited to those years for which we were able to obtain the property tax rolls of the individual counties from the Florida Department of Revenue – 1995 to 2004. From the rolls we were able to calculate the multifamily housing stock in total square footage as of January 1 of each year. The tax roll data also include the two most recent sales prices for each property and the year of each sale. From these data individual county repeat-sales price indexes were estimated for vacant residential land and multifamily housing. The construction of these indexes first involved estimating the standard repeat-sales model: Ln ( n-ti, ti, P P ) = β∑ − T 1k k Di,k + εi,t,t-n (1) Where Pi,t is the most recent selling price of property i at time t; Pi,t-n is the previous selling price of property i at time t-n; βk is the logarithem of the cumulative price index in period t; Di,k is a dummy variable which equals -1 at the time of the initial sale, +1 at the time of the second sale, and 0 otherwise; and εi,t,t-n is the regression error term. The estimated coefficients of (1) were then used to calculate annual appreciation rates. Finally, the nominal price of a square foot of multifamily housing and an acre of vacant residential land were computed for each year by starting with average values in 2003 10 (calculated from the tax rolls) and predicting values for previous years using the estimated annual appreciation rates. The final data item used to complete our panel are the Means City Construction Cost Indexes. These indexes are available annually for 16 Florida cities. For each year of the panel each county was assigned the annual index value of the closest city.8 The 33 metropolitan counties included in our data base are divided into central counties and outer suburban counties (See Table 1). Central counties contain a central city, while outer suburban counties do not. Central counties are further divided into central city areas and inner suburban areas. Central city areas may include more than one central city.9 The models described below investigate the effects of impact fees on multifamily housing construction and are estimated separately for the central city areas, inner suburban areas, and outer suburban areas.10 IV. Estimated Models A change in the multifamily housing stock occurs when there is a difference between the equilibrium stock and the actual stock. If the equilibrium stock is less than the actual stock, the actual stock will shrink over time as rental units are converted to alternative land uses. Alternatively, if the equilibrium stock is larger than the actual stock, the stock will expand due to new apartment construction. Due to adjustment lags, 8 Means includes the cost of materials, labor, and equipment rental costs. 9 Of the 33 counties, 19 are central counties and 14 outer suburban counties. Of the central counties, 6 have more than one central city. 10 According to the 2000 Census of Population and Housing, 21.5% of Florida’s metropolitan population live in central cities, 59.6% live in inner suburban areas, and 19.0% live in the outer suburbs. 11 (1995-2004) yields seven data points for each County. For consistency, these seven data points are used to estimate all three models.12 Our first model is a two way (time and space) fixed effects model: ∆St = αi + γt + β0 * WSIFt-1 + β1 * NWSIFt-1 + εit (2) Where ∆St is the annual change in the total amount of multifamily housing square footage within the area (central city, inner suburbs, outer suburbs) over year t; WSIF and NWSIF are real water/sewer impact fees and real non-water/sewer impact fees on the standard apartment, respectively; αi and γt are fixed effects for area and time, respectively; and εit is the idiosyncratic error term. Area fixed effects account for unobserved heterogeneity across areas related to multifamily housing construction and time effects control for factors that uniformly affect all areas over time. An important advantage of (2) is that impact fees are allowed to depend on area levels of multifamily housing stock changes, thus mitigating the potential for endogeneity bias. Equation (2) is estimated by OLS, after first differencing the variables in order to eliminate the area fixed effect (αi). The second model we estimate adds an area specific time trend (gi) to equation (2): ∆St = αi + γt + git + β0 * WSIFt-1 + β1 * NWSIFt-1 + εit (3) 12 Also in the interest of consistency across models, we use the change in impact fees as our explanatory variable in all three models. While this is the correct variable in our third model (stock-adjustment), it could be argued that in our first two models (fixed effects and random trend) the correct variable is the change in the change in impact fees. This follows because it is changes in impact fees that induce disequilibrium and cause changes in the stock. After differencing, the impact fee variable would be a double difference. However, in our data the difference and the double difference are essentially the same variable, with the intra-county correlation between the two variables generally being .8 or higher. 14 This model, which is referred to as the random trend model, allows each area to have its own time trend in multifamily housing construction. The area-specific trend is an additional source of heterogeneity. The random trend model allows impact fees to depend on area-specific trends in multifamily housing construction, in addition to the level of multifamily housing construction. The model is estimated by first differencing to eliminate αi and then fixed effects are applied to the differences – i.e., n-1 area dummy variables are added to the model. A possible criticism of models (2) and (3) is that there may be other variables, in addition to impact fees, that vary within areas over time that affect multifamily housing construction. If movements in these variables are correlated with changes in impact fees, the estimated effects of impact fees on multifamily housing construction may be biased. However, omitted variable bias is likely unimportant given our data and approach. Bias will only result if the excluded variable commonly varies within areas, this variation is commonly correlated with the variation in impact fees within areas, and the variable has a common important effect on multifamily housing construction across areas, after controlling for fixed effects, random growth trends, and aggregate time effects in the specification. While unlikely, omitted variable bias is still possible. We therefore also estimated a standard stock-adjustment model, including space and time fixed effects: ∆St = αi + γt + β0 * WSIFt-1 + β1 * NWSIFt-1 + β2 * LPt-1 + β3 * HPt-1 + β4* CCt-1 + β5 * St-1 + εit (4) 15 where LP, HP, and CC are land price, multifamily housing price, and construction costs, respectively, as defined in Section III. These variables, along with impact fees, are assumed to determine the equilibrium amount of multifamily housing. The estimated coefficient on the lagged value of the stock (β5, which is expected to be negative) represents the rate at which the stock adjusts to the new equilibrium. The price and cost variables enter as real values, having been deflated by the CPI for the southeast region. Like impact fees, they are measured at the county level. The lagged value of the multifamily housing stock, however, is measured separately for each area type (i.e., for a central county, St-1, is measured separately for the central city and the inner suburbs).13 Once again, equation (4) is estimated in first differences in order to eliminate αi. After differencing, ∆St (= St-1 – St-2) appears on the right hand side of (4). By construction, ∆St-1 is not strictly exogenous, which will result in (4) yielding inconsistent estimates if estimated by OLS. Strict exogeneity is stronger than assuming contemporaneous exogeneity since it implies that explanatory variables in each time period are uncorrelated with the idiosyncratic error (εit) in each time period: E ( εX'is it) = 0, s, t = 1,…, T. Strict exogeneity fails in (4) since ∆St-1 and εit are correlated. A simple approach to consistent estimation is to instrument ∆St-1 with lagged levels of the explanatory variables, beginning with t-2. We choose lagged levels of the stock, housing price, and construction costs as our set of instrumental variables. Two statistics 13 Note that (4) does not include a variable measuring removals from the stock due to depreciation or scrapage, because this cannot be measured. However, if removals are fairly constant within counties over time, they are captured by the area fixed effects. If removals vary within counties over time, we have no reason to believe that they would be correlated with impact fees. 16 The results for non-water/sewer impact fees contrast sharply to those obtained for water/sewer impact fees. For central city and outer suburban areas estimated effects are highly insignificant (with estimated standard errors exceeding estimated coefficients in all cases). However, for inner suburban areas, non-water/sewer impact fees have positive, statistically significant effects on multifamily housing construction across all three estimators. The magnitudes of the estimated coefficients are again highly similar, ranging between 2333 and 2801. The average coefficient indicates that multifamily housing construction will increase by 2581 square feet in the year after a $1 increase in non-water/sewer impact fees. The estimated coefficients on the lagged value of the stock in the stock- adjustment models are all negative and lie between -.2 and -.6, implying dynamic stability. However, these coefficients are not statistically significant. The other variables entering the stock-adjustments models – real values of land price, housing price, and construction costs – are also uniformly insignificant. In the case of the construction cost index, its real value displays little variation within counties over the time period covered by our panel. The real values of multifamily housing price and land price do increase for most counties, but there is high collinearity between these two variables (the simple correlation coefficient exceeds .9 for most counties). The insignificant results obtained for the control variables in the stock-adjustment models are therefore not surprising. The finding, however, that the impact fee results are robust to the inclusion of these 19 variables lends support to our belief that omitted variables have not biased the results yielded by the fixed effects and random trend models.17 To further assess the quantitative impacts of the estimated impact fee effects on multifamily housing construction, we used the average estimated coefficients reported above to calculate short-run elasticities (at the point of means for each area type). These elasticities measure the immediate response of construction to higher impact fees. We also calculated long-run elasticities based upon the estimated impact fee and stock coefficients obtained from the stock-adjustment models. The long-run elasticity measures the percentage change in the equilibrium stock of multifamily housing in response to a one percent change in impact fees. Due to the imprecision in the estimated stock coefficients, these elasticities should only be interpreted as suggestive of the true long-run equilibrium effects. The calculated elasticities are reported in Table 5. The short-run elasticities for water/sewer impact fees range between –6 and –8 across the three area types, indicating that increases in these fees have strong negative effects on multifamily housing construction. As discussed in Section II, these fees impose direct costs on developers 17 Another possible omitted variable is suggested by the possibility that developers anticipate increases in impact fees and attempt to build before the fee increase is imposed. The reduction in construction that we observe in the year after an increase in water/sewer fees may therefore simply represent construction that occurred in the previous year. To investigate this, we added to the random trend models (equation 3) impact fee values for January 1 of the next year. To illustrate, for construction observed during 1999, both the January 1, 1999 and January 1, 2000 fees were included. For central cities and outer suburbs, the leading values of impact fees are not statistically significant and their inclusion has little effect on the lagged value results for either the water/sewer or non-water/sewer variables. However, for inner suburbs there is weak evidence that the anticipation of impact fees does speed up development. The estimated coefficient on the leading value of water/sewer fees is 3582 (t statistic = 1.6). The estimated coefficient on the lagged value of water/sewer fees falls from -4475 (t=4.0) to -3075 (t=3.0). The estimated coefficient on the leading value of non-water/sewer fees is not significant and the lagged value remains positive and statistically significant (β=2061, t=2.34) after the inclusion of the leading values. 20 and probably yield little benefit in the form of project approval cost savings. They therefore act as a tax on development, which unambiguously shifts the supply curve upward. The only short-run elasticity calculated for non-water/sewer impact fees is for the inner suburbs, because it is only for this area type that these fees are found to have a statistically significant impact on construction. Here the elasticity is also large in magnitude (4), indicating that non-water/sewer impact fees have a strong positive effect on multifamily housing construction in the short-run. Apparently, increases in these fees reduce project approval costs by more than the fees themselves, resulting in a downward shift in the supply curve. The long-run elasticities also suggest that changes in impact fees have nontrivial effects on the long run equilibrium stock of multifamily housing. For water/sewer impact fees they range between -.3 (inner suburbs) to –1.2 (central cities). The long-run elasticity with respect to non-water/sewer impact fees within inner suburban areas is .6. Another method for gauging the magnitudes of the estimated effects is to consider what would happen within an actual county if it was to increase its impact fees. The inner suburbs of Pinellas County contain roughly the average amount of multifamily housing square footage found within inner suburban areas throughout the state. On January 1, 2004 suburban Pinellas had 37.7 million square feet of multifamily rental housing. Our panel data show that when counties increase either their water/sewer or non-water/sewer impact fees, in real terms the increase on average is about $250 per apartment. What would happen to the multifamily housing stock in 21 From a policy perspective the implication of our results is clear – if the goal is to increase the stock of multifamily housing within inner suburban areas, states should encourage their communities to adopt non-water/sewer fees but discourage the use of water/sewer fees. Perhaps the best approach would be to adopt non-water/sewer fees but continue to incorporate the costs of offsite water/sewer system improvements within the base of user fees. Of course, there may be other approaches, besides impact fees, toward reducing the fiscal incentive for the exclusion of low income housing from the suburbs. Any approach that reduces a reliance on the property tax as a means of financing the public service costs of new development may work.19 However, because impact fees are specifically intended for new development to pay its own way, alternatives to fees may be second-best in nature. To our knowledge, we have offered the first available evidence on whether impact fees help or hurt lower income households’ quest for affordable suburban housing. Clearly, much more research on this important topic is needed, especially for other states. On our own research agenda is an extension of the present work to determine whether impact fees affect the number of single-family starter homes that get built within suburban areas. 19 For example, as noted by Ladd (1998), court-ordered reforms that force states to provide more aid to equalize education spending may reduce some of the relative disadvantage of admitting poor households. 24 References Altshuler, Alan A. and Jose ́ A. Gomez-Ib́añez. 1993. Regulation for Revenue: The Political Economy of Land Use Exactions. Washington D.C.: Brookings Institution and Cambridge, MA: Lincoln Institute of Land Policy. Beal, Mary. 2004. “A Comparison of Land Use Regulation Between Leon and Wakulla Counties”. Policy Brief #12. DeVoe L. Moore Center. College of Social Sciences. Florida State University. Bobo, B.F. 2001. Locked In and Locked Out: The Impact of Urban Land Use Policy and Market Forces on African Americans.Westport: Praeger. Cervero, Robert. 1989. “Jobs-Housing Balancing and Regional Mobility”. American Planning Association Journa.Spring, 136-150. Danielson, Michael N. 1976. “The Politics of Exclusionary Zoning in Suburbia”. Political Science Quarterly, 91(1), 1-18. Gyourko, Joseph. 1991. “Impact Fees, Exclusionary Zoning, and the Density of New Development”. Journal of Urban Economics, 30, 242-256. Ihlanfeldt, Keith R. 2004. “Exclusionary Land-use Regulations within Suburban Communities: A Review of the Evidence and Policy Prescriptions”. Urban Studies, 41(2), 261-283. Ihlanfeldt, Keith R. and Timothy M. Shaughnessy. 2004. “An Empirical Investigation of the Effects of Impact Fees on Housing and Land Markets”. Regional Science and Urban Economics. 34(6), 639-661. Ihlanfeldt, Keith R. 1999. “The Geography of Economic and Social Opportunity in Metropolitan Areas”. In A. Altshuler, W. Merril, H. Wolman, and F. Mitchell. (eds). Governance and Opportunity in Metropolitan America. Washington D.C.: National Academy Press, pp. 213-252. Ladd, Helen F. 1998. Local Government Tax and Land Use Policies in the United States: Understanding the Links. Cambridge, MA: Lincoln Institute of Land Policy. Lawhon, Larry L. 2003. “Development Impact Fee Use By Local Governments”. Municipal Year Book. 27-31. 25 Schill, Michael H. 1992. “Deconcentrating the Inner City Poor”. Chicago-Kent Law Review. 67(3), 795-853. Wooldrige, J. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: The MIT Press. Yinger, J. 1998. “The Incidence of Development Fees and Special Assessments”. National Tax Journal. 51, 23-41. 26 Table 2 Results for Central Cities (1) (2) (3) Fixed effects Random trend Stock- adjustment Impact fees, not water/sewer -13 -81 74 (497)a (587) (510) Impact fees, water/sewer -1152** -1100** -1356* (492) (526) (822) Land price index -.85 (26) Construction index 132311 (228746) Multifamily price index 32569 (50465) Multifamily stock -.18 (.28) R-square .108 .125 .227 Observations 118 118 118 F-tests on IVsb 4.42 [.001]c Over-id test .37 [.992] aRobust standard errors in parentheses. bIdentifying instruments are lagged values of multifamily stock, land price, multifamily price, and construction cost index. cp-values in brackets. *, **, *** indicate significance at the 10%, 5%, and 1% levels by a two-tailed test, respectively. 29 Table 3 Results for Inner Suburbs (1) (2) (3) Fixed effects Random trend Stock- adjustment Impact fees, not water/sewer 2333* 2801** 2610* (1372)a (1205) (1529) Impact fees, water/sewer -3064*** -4475*** -1137 (820) (1126) (1739) Land price index 32.3 (82.4) Construction index -85976 (282819) Multifamily price index -62542 (70399) Multifamily stock -.27 (.39) R-square .162 .251 .344 Observations 118 118 118 F-tests on IVsb 2.20 [.060]c Over-id test 4.61 [.330] aRobust standard errors in parentheses. bIdentifying instruments are lagged values of multifamily stock, land price, multifamily price, and construction cost index. cp-values in brackets. *, **, *** indicate significance at the 10%, 5%, and 1% levels by a two-tailed test, respectively. 30 Table 4 Results for Outer Suburbs (1) (2) (3) Fixed effects Random trend Stock- adjustment Impact fees, not water/sewer -717 -682 -878 (783)a (948) (1844) Impact fees, water/sewer -1194 -1508* -1250* (907) (864) (697) Land price index 4.7 (80) Construction index -34227 (175483) Multifamily price index -28957 (25832) Multifamily stock -.57 (.86) R-square .111 .116 .474 Observations 111 111 111 F-tests on IVsb 5.35 [.006]c Over-id test 3.12 [.077] aRobust standard errors in parentheses. bIdentifying instruments are lagged values of multifamily stock, land price, multifamily price, and construction cost index. cp-values in brackets. *, **, *** indicate significance at the 10%, 5%, and 1% levels by a two-tailed test, respectively. 31