A major argument in support of land-value taxation is that it creates no incentives for altering behavior in order to avoid the tax. By contrast, a conventional property tax, levied on buildings, can deter landowners from erecting otherwise desirable structures on their land. For example, homeowners may decide against finishing a basement or adding a second bath because it would increase tax liability. Thus, a conventional property tax can lead to excessively low capital-land ratios and “excess burden”—a cost to taxpayers over and above the actual monetary payments they make to the tax authorities. This article reports on a recent study of excess burden resulting from an early British antecedent of the modern property tax—the 17th-century window tax.

The Case of the Window Tax

In 1696, King William III of England, in dire need of additional revenues, introduced a dwelling unit tax determined by the number of windows in an abode. The tax was designed as a property tax, as described by this discussion in the House of Commons in 1850: “The window tax, when first laid on, was not intended as a window tax, but as a property tax, as a house was considered a safe criterion of the value of a man’s property, and the windows were only assumed as the index of the value of houses” (HCD 9 April 1850).

In its initial form, the tax consisted of a flat rate of 2 shillings upon each house and an additional charge of 4 shillings on houses with between 10 and 20 windows, or 8 shillings on houses with more than 20 windows. The rate structure was amended over the life of the tax; in some cases, rates were raised dramatically. In response, owners of dwellings attempted to reduce their tax bills by boarding up windows or by constructing houses with very few of them. In some dwellings, entire floors were windowless, leading to very serious and adverse health effects. In one instance, lack of ventilation led to the death of 52 people in the surrounding town, as reported by a local physician who called on a house inhabited by poor families:

“In order to reduce the window tax, every window that even poverty could dispense with was built up, and all sources of ventilation were thus removed. The smell in the house was overpowering and offensive to an unbearable extent. There is no evidence that the fever was imported into this house, but it was propagated from it to other parts of town, and 52 of the inhabitants were killed.” (Guthrie 1867)

The people protested and filed numerous petitions to Parliament. But, despite its pernicious effects, the tax lasted more than 150 years before it was finally repealed in 1851.

The window tax represented a substantial sum for most families. In London, it ranged from about 30 percent of rents on “smaller houses on Baker Street” to as much as 40 to 50 percent on other streets, according to a House of Commons debate in 1850 (HCD 9 April 1850). The tax was particularly burdensome on poor families living in tenements, where assessors taxed the residents collectively. Thus, if a building contained 2 apartments, each with 6 windows, the building was taxed at a rate based on 12 windows. By contrast, on very large houses of the wealthy, the tax typically did not exceed 5 percent of the rental value.

The tax schedule underwent several significant changes before it was finally repealed. In 1784, Prime Minister William Pitt raised tax rates to compensate for lower taxes on tea. Then in 1797, Pitt’s Triple Assessment Act tripled the rates to help pay for the Napoleonic Wars. The day following this new act, citizens blocked up thousands of windows and wrote in chalk on the covered spaces, “Lighten our darkness we beseech thee, O Pitt!” (HCD 24 Feb. 1848).

England and Scotland were both subject to the window tax, but Ireland was exempted because of its impoverished state. One member of Parliament quipped, “In advocating the extension of the window tax to Ireland, the Honorable Gentleman seemed to forget that an English window and an Irish window were very different things. In England, the window was intended to let the light in; but in Ireland the use of a window was to let the smoke out” (HCD 5 May 1819).

The window tax, incidentally, was viewed as an improvement over its antecedent, the hearth tax. In 1662, Charles II (following the Restoration) imposed a tax of 2 shillings on every fire hearth and stove in England and Wales. The tax generated great resentment largely because of the intrusive character of the assessment process. The “chimney-men,” as the assessors and tax collectors were called, had to enter the house in order to count the number of hearths and stoves. The window tax, by contrast, did not require access to the interior of a dwelling; the “window peepers” could count the apertures from the outside and avoid invading the privacy of the home.

The window tax, however, created some administrative problems of its own—most notably the definition of a window for purposes of taxation. The law was vague, and it was often unclear what constituted a window for tax purposes. In 1848, for example, Professor Scholefield of Cambridge paid tax on a hole in the wall of his coal cellar (HCD 24 Feb. 1848). In the same year, Mr. Gregory Gragoe of Westminster paid tax for a trapdoor to his cellar (HCD 24 Feb. 1848). As late as 1850, taxpayers urged the Chancellor of the Exchequer to clarify the definition of a window.

Notches and Their Effects on Behavior

Throughout its history, the window tax consisted of a set of “notches.” A notch in a tax schedule exists if a small change in behavior—such as the addition of a window—leads to a large change in tax liability.

Notches are rare (Slemrod 2010) and not to be confused with kinks, which are far more common even today. A kink in a tax schedule exists if a small change in behavior leads to a large change in the marginal tax rate but just a small change in tax liability. The income tax in the United States, for example, has several kinks. Married couples with taxable income from $17,850 to $72,500 are in the 15 percent marginal tax bracket; couples with taxable income from $72,500 to $146,400 are in the 25 percent marginal tax bracket. If a couple with income of $72,500 were to earn an extra dollar, its marginal tax rate would jump to 25 percent, but its tax liability would increase by just $.25.

Microfilm records of local tax data in the U.K. from 1747 to 1830 allow for a more systematic examination of the impact of the window tax and notches. This article draws on a data set from 1747 to 1757, with information on 493 dwellings from Ludlow, a market town in Shropshire, near the border of Wales. Over this period, the window tax schedule included 3 notches. A homeowner in this period paid:

• no tax if the house had fewer than 10 windows;
• 6 pence per window if the house had 10 to 14 windows;
• 9 pence per window if the house had 15 to 19 windows;
• 1 shilling per window if the house had 20 or more windows.

Homeowners who purchased a 10th window thus paid a 6 pence tax on the 10th window as well as on each of their 9 other windows, which previously had been untaxed. Thus the total tax on the 10th window was 60 pence, which was equal to 5 shillings. If the window tax distorted decisions and thus led to excess burden, then one would expect to find many homes with 9, 14, or 19 windows but very few with 10, 15, or 20. A test of this argument is discussed below.

Through the first half of the 18th century, the administration of the tax had been troublesome, as homeowners frequently camouflaged or boarded up windows until the tax collector was gone, or took advantage of loopholes or ambiguities in the tax code. As a result, tax collections were much lower than expected. In 1747, however, Parliament revised the tax by raising rates and introducing measures to improve its administration. Most notably, it prohibited the practice of blocking up and subsequently reopening windows in order to evade assessment; violators had to pay a penalty of 20 shillings (1 pound) for every window they reopened without notifying the tax surveyor (Glantz 2008).

The 1747 act reduced tax evasion significantly, so the data for the following 10 years should provide reasonable estimates of the actual number of windows. If the window tax distorted behavior, one would expect to find spikes in the number of dwellings at the notches, with 9, 14, or 19 windows. And this is precisely what the data demonstrate. Figure 1 is a histogram showing the number of windows for homes in the sample. The pattern is clear; there are sharp increases in the number of homes with 9, 14, or 20 windows:

• 18.4 percent of the homes have 9 windows, 3.9 percent 8 windows, and 4.6 percent 10 windows.
• 16.6 percent have 14 windows, 6.0 percent 13 windows, and 1.8 percent 15 windows.
• 7.1 percent have 19 windows, 3.4 percent 18 windows, and 0.7 percent 20 windows.

Standard statistical tests reject the hypothesis that there are equal numbers of houses with 8, 9, or 10 windows; with 13, 14, or 15 windows; or with 18, 19, or 20 windows. It is manifestly clear that people responded to the window tax by locating at one of the notches so as to minimize their tax liability.

Data on a sample of 170 houses for the period 1761 to 1765 shed light on the response to Parliamentary revisions to the tax in 1761. In addition to rate increases, the 1761 revisions expanded coverage of the tax to include houses with 8 or 9 windows. Under the earlier rate structures, houses with fewer than 10 windows paid no window tax. For this second sample, figure 2 shows a large spike at 7 windows: 28.2 percent of the houses have 7 windows, but only 5.2 percent have 6 windows, and just 2.9 percent have 8 windows. Once again, it’s easy to reject the hypothesis that there were an equal number of houses with 6, 7, or 8 windows.

In summary, the evidence from our two samples makes it quite clear that there was a widespread tendency to alter behavior in order to reduce tax payments. People chose the number of windows not to satisfy their own preferences, but to avoid paying higher levels of taxes. The window tax, in short, generated a real “excess burden.”

How Large Was the Excess Burden from the Window Tax?

As discussed, the window tax was substantial and induced widespread tax-avoiding behavior. Based on some standard techniques of economic analysis, our simulation model generates an estimate of what people would have been willing to pay for their preferred number of windows. The model captures each consumer’s demand for windows with and without the tax, the taxes paid, and the loss of welfare from adjusting the number of windows in response to the tax.

In the sample from 1747 to 1757, the estimated welfare losses were very large for households at one of the notches. For them, the welfare loss (i.e., excess burden) is 62 percent of the taxes they paid. That is to say, for every dollar collected under our simulated version of the window tax, the tax imposed an additional burden or cost of 62 cents on these households. The excess burden, not surprisingly, is particularly large for households that chose 9 windows. One criterion economists use to evaluate a tax is excess burden relative to taxes paid. By this standard, a good tax is one that collects significant revenue buts leads to very small changes in decisions. Consumers who purchased 9 windows are thus the worst possible case. Those consumers paid no tax; so, for them, the entire burden of the tax is excess burden.

For our entire sample of 1,000 simulated households, the excess burden as a fraction of taxes paid is about 14 percent. Thus for each tax dollar raised by the window tax, our simulation suggests an additional cost of 14 cents to taxpayers as a result of their distorted choices.

Some Concluding Remarks

The window tax represents a very clear, transparent case of excess burden—a tax that placed heavy costs on taxpayers in addition to their tax liabilities resulting from tax-avoiding adjustments in behavior. But, as mentioned early on, modern property taxes also create an excess burden, although the consequences are less dramatic than in the case of the window tax.

In designing a tax system, it is important to consider this issue. The ideal, in principle, is a neutral tax that raises the desired revenues but doesn’t distort taxpayer behavior so as to create additional burdens. Such a tax is a pure land-value tax levied on the site value of the land—that is, its value with no improvements. Thus, the assessed value of the land (and hence the tax liability of the owner) is completely independent of any decisions made by the owner of the land parcel. Unlike the window tax, which provides a compelling example of the additional costs that arise when property tax liabilities depend on the behavior of the property owner, a land-value tax creates no incentives for tax-avoiding behavior.

About the Authors

Wallace E. Oates is Distinguished University Professor of Economics, Emeritus, University of Maryland, and University Fellow at Resources for the Future.

Robert M. Schwab is a professor of economics at the University of Maryland.

Resources

Binney, J. E. D. 1958. British Public Finance and Administration, 1774–92. Oxford: Clarendon Press.

Blinder, Alan S., and Harvey S. Rosen. 1985. “Notches.” American Economic Review 78 (September): 736–747.

Dickens, Charles. 1850. Household Words. Vol. 1. London: Bradbury and Evans.

Douglas, Roy. 1999. Taxation in Britain since 1660. London: MacMillan.

Dowell, Stephen. 1884. A History of Taxation and Taxes in England from the Earliest Times to the Present Day. Vols. 2 and 3. London: Frank Cass & Co.

Fielding, Henry. 1975. The History of Tom Jones, A Foundling. Wesley University Press.

George, M. Dorothy. 1926. London Life in the XVIIIth century. New York: Alfred A. Knopf.

Glantz, Andrew E. 2008. “A Tax on Light and Air: Impact of the Window Duty on Tax Administration and Architecture.” Penn History Review 1696–1851 15 (2): 1–23.

Guthrie, Thomas. 1867. “How to Get Rid of an Enemy.” The Sunday Magazine.

HCD (House of Commons Debates). 5 May 1819. Vol. 40 cc 126–148. “Motion for the Repeal of the Window Tax in Ireland.”

HCD. 24 February 1848. Vol. 96 cc 1259–1297. “Lowest Classes Under Assessment.”

HCD. 9 April 1850. Vol. 110 cc 68–99. “Window Tax.”

Kennedy, William. 1913. English Taxation, 1640–1799. London: G. Bell and Sons, Ltd.

Marshall, Alfred. 1948. Principles of Economics, 8th edition. New York: Macmillan.

Neary, J. Peter, and Kevin S. W. Roberts. 1980. “The Theory of Household Behaviour under Rationing.” European Economic Review 13 (January): 25–42.

Sallee, James M., and Joel Slemrod. “Car Notches: Strategic Automaker Responses to Fuel Economy Policy,” NBER Working Paper #16604, 2010. http://www.nber.org/papers/w16604.pdf.

Sinclair, Sir John. 1804. The History of the Public Revenue of the British Empire. London: Strahan and Preston.

Slemrod, Joel. 2010. “Buenas Notches: Lines and Notches in Tax System Design.” Unpublished working paper. http://webuser.bus.umich.edu/jslemrod/pdf/Buenas%20Notches%20090210.pdf.

Smith, Adam. 1937. The Wealth of Nations. New York: Random House.

Walpole, Spencer. 1912. A History of England from the Conclusion of the Great War in 1815. Vol. 5. London: Longmans, Green, and Company.

Weitzman, Martin L. “Prices and Quantities.” Review of Economic Studies 41: 477–491.

The single greatest challenge to any type of land value taxation system is accurate valuation of land on a large scale. In urban areas where nearly all real estate sales data represent transfers of land with improvements, it is difficult to divide prices between land and building components. Although many jurisdictions require a separate listing of land and building values on their tax rolls, these allocations will not affect the final tax bill if the tax rate is the same on both.

Any special tax on land value alone would increase the need to assign more accurate land values to parcels that have been improved over many years. As a result, skepticism as to the feasibility of this process has proven a major stumbling block to serious consideration of two-rate property taxes and other forms of special land taxation. Many observers have concluded that the practical problems of land assessment prevent the realization of the many theoretical benefits it offers.

New advances in computerized approaches to property assessment have important implications for this debate. While land valuation presents special problems in the analysis of sales data for improved parcels, it also can benefit from location analysis and land value mapping techniques. Buildings can and will vary unpredictably in both type or value from lot to lot, but land values for adjoining or nearby parcels should have a more constant relationship to one another. More than 20 years ago, Oliver Oldman of Harvard Law School, considered the implications of this situation for an appeals process under a land value tax, recognizing that a successful challenge to one parcel’s valuation would have implications for many other assessments as well. He wrote, “The key to developing an accurate land-value assessment roll is the process of land-value mapping.” Now the technology is available to achieve this goal.

In a recent seminar at the Lincoln Institute, representatives of the Auditor’s Office in Lucas County, Ohio, which includes the city of Toledo, joined a group of economists, appraisers, lawyers and local officials to examine current methods of land valuation. Lucas County has one of the most sophisticated appraisal systems in the country, with almost 20 years of experience in using computerized methods of spatial data analysis for property taxation. The seminar provided a valuable opportunity to discuss the county’s innovative approaches to the integration of geographic information systems and computer-assisted land valuation to estimate the effect of location on real estate market value.

Traditional Methods of Land Valuation

There are several standard methods of deriving a value for unimproved land, all extremely problematic as the basis for jurisdiction-wide assessment.

Comparable Sales: The most straightforward method is an analysis of sales of comparable unimproved land, adjusting the prices to account for any differences in size, location, and features. Similarly, the capitalization of rental income for comparable vacant land can serve as a basis for estimating its sale price. However, these methods are difficult to apply in densely populated urban areas where sales or rentals of unimproved land are rare. The pool of sales data can be expanded if sales of improved land are followed soon after by demolition of the buildings. In that case, the unimproved land value can be estimated as the purchase price minus the costs of the demolition. Although such sales provide an important check for estimated values produced by other approaches, they do not exist in sufficient numbers over a varied enough geographic range to serve as the sole basis for assessment.

Income Analysis: The land residual method begins with an estimate of the income yielded by the developed property. The building value is then calculated, and from that the income attributable to the building is derived. Capitalizing the remaining income then provides a value for the land. However, even a cursory description of this method suggests the difficulties of its application. In particular, the existence of depreciation, or any deviation from highest and best use that would distort the income available to the unimproved land, can leave the independent value of the improvements extremely uncertain.

Cost Analysis: Similar problems confront a division of value according to the depreciated reproduction cost of the improvements. This method assumes that structures can be worth no more than their cost of construction, and assigns all remaining value in the improved parcel to the land itself. Physical, economic or functional depreciation greatly complicates the attempt to calculate building value, however, so this method requires fairly new construction whose price can be confidently estimated as a measure of value. The financial effect of various forms of obsolescence can only be measured accurately through examination of sales data, which will almost never be available for the building alone.

Cost of Development: A full-scale market appraisal of potential development alternatives provides another basis for estimating the sale price of unimproved land. This is the approach taken by developers considering new uses for land, land trusts seeking to acquire and preserve undeveloped open space, and taxpayers claiming deductions for charitable contributions of development rights. However, it is most suitable for valuing undeveloped land to be used for residential subdivisions. Even in these situations, it requires extensive study of the potential market for such properties, local restrictions on development, and the physical attributes of the land that would affect its building capacity, such as soil and drainage characteristics. This type of exhaustive individual appraisal is appropriate for purchasers or developers of individual parcels, but is not feasible for annual assessments for all parcels in a taxing jurisdiction.

Other valuation methods, such as derivation of typical ratios of site value to total improved property value, are even less useful in the case of densely developed urban property, where buildings of all sizes, ages and utility may be found in close proximity on fairly similar parcels of land.

New Approaches: CAMA and GIS

The greatest change in assessment practice over the past three decades has involved the use of computers and mathematical formulas to establish a relationship between property characteristics and sale prices, thereby permitting an estimate of the market value of other properties not subject to a recent sale. This approach is known as computer-assisted mass appraisal (CAMA). Site characteristics such as size and location are important elements of these mathematical models, raising the possibility of estimating the effect of location on parcel value.

At the same time, the development of computerized geographic information systems (GIS) has permitted assessors to develop location-based property records or cadastres, and to coordinate sales data with location. More sophisticated and less expensive GIS technology now offers the potential for full integration with CAMA for spatial analysis. Initial attempts to quantify location effects faced difficulties not only in defining and maintaining “economic neighborhoods” or zones, i.e., contiguous areas of relatively homogeneous land values, but also in understanding the dynamics of the interactive, elusive locational factor. Some efforts developed different mathematical models for each geographic region or “cluster” of properties with similar characteristics. However, these approaches could not capture the many complex, interrelated and significant micro-variations within any given neighborhood, and could not reduce the determination of location value to an objective process.

Lucas County pioneered a new approach to location value-the use of GIS tools to develop a response surface that represents the effect of location on land value. The response surface is a fitted three-dimensional surface that represents a percentage adjustment to land and/or land and improvements based on a parcel’s geocoded location. Included in the analysis are geographic coordinates and distances from important features, such as other recent sales, institutions, amenities or other “value influence centers.” This analysis results in a three-dimensional representation, with the height of the surface (z) at any specific x-y coordinate indicating the approximated location value of that parcel. This variable is then evaluated with others, such as land and building size, quality, condition and depreciation, to produce a total estimated value for the parcel.

In the Lucas County example, the response surface differs from a mathematical equation in that it is developed through a spatial analysis process available in GIS to estimate the effects of location on value and refine those estimates after comparing them with sales and appraisal data. This approach still relies on an element of appraisal and economic judgment in determining neighborhood boundaries for location effects, but it can be tested and refined by observing the effect of different neighborhood “breaklines” on the resulting three-dimensional value surface.

To be used successfully in mass appraisal, these sophisticated approaches must yield results that are reasonable, understandable and available to typical taxpayers. Lucas County has pioneered this aspect of the assessment process, as well. All real estate records, values and maps are available on a CD with GIS viewing software, priced at its production cost of $10, and online free at all public libraries in the county. Taxpayers can view property records or create customized maps showing the location of multiple parcels and the relationships among their taxable values.

Future Directions

Participants in the Lincoln Institute seminar found great promise in the Lucas County approach to location value, and identified many points for further development and investigation. All agreed that recent decades have seen a literal revolution in assessment practice, with great potential for increasing the feasibility of large-scale land valuation. Among the most important theoretical questions were the “functional form” of this spatial analysis, including the type of effect on value observed with changes in location and distance variables; the identification of omitted variables (those for which data is not available or which have been overlooked in the past); and the relationship between marginal value estimates and the total parcel value needed for assessment. Similarly, the effect of substandard buildings and less than “highest and best use” on values requires further exploration.

Development of these new approaches must be matched by educational efforts to explain their operation to taxpayers, local officials, and the lawyers and judges who will consider their consistency with legal standards for assessment practice. Through its innovative efforts in both of these areas, Lucas County has made an important contribution to the theory and practice of land valuation.

Jerome C. German is the chief assessor for Lucas County, Ohio. Dennis Robinson is vice president of programs and operations at the Lincoln Institute. Joan Youngman is senior fellow and director of the Institute’s Program on Taxation of Land and Buildings.

References

International Association of Assessing Officers. Property Appraisal and Assessment Administration (1990).

Oliver Oldman and Mary Miles Teachout. “Valuation and Appeals Under a Separate Tax on Land.” 15 Assessor’s Journal 43-57 (March 1980).

Richard D. Ward, James R. Weaver, and Jerome C. German. “Improving CAMA Models Using Geographic Information Systems/Response Surface Analysis Location Factors.” 6 Assessment Journal 30-38 (January/February 1999).

Lucas County website: www.co.lucas.oh.us

Editor’s note: The Lincoln Institute published a new report on tax increment financing in September, 2018.

Tax increment financing (TIF) is an alluring tool that allows municipalities to promote economic development by earmarking property tax revenue from increases in assessed values within a designated TIF district. Proponents point to evidence that assessed property value within TIF districts generally grows much faster than in the rest of the municipality and infer that TIF benefits the entire municipality. Our own empirical analysis, using data from Illinois, suggests to the contrary that the non-TIF areas of municipalities that use TIF grow no more rapidly, and perhaps more slowly, than similar municipalities that do not use TIF. An important finding is that TIF has different impacts when land use is considered. For example, commercial TIF districts tend to decrease commercial development in the non-TIF portion of the municipality.

Designating a TIF District

The rules for tax increment financing, and even its name, vary across the 48 states in which the practice is authorized. The designation usually requires a finding that an area is “blighted” or “underdeveloped” and that development would not take place “but for” the public expenditure or subsidy. It is only a bit of an overstatement to characterize the “blight” and “but for” findings as merely pro forma exercises, since specialized consultants can produce the needed evidence in almost all cases. In most states, the requirement for these findings does little to restrict the location of TIF districts.

TIF expenditures are often debt financed in anticipation of future tax revenues. The practice dates to California in 1952, where it started as an innovative way of raising local matching funds for federal grants. TIF became increasingly popular in the 1980s and 1990s, when there were declines in subsidies for local economic development from federal grants, state grants, and federal tax subsidies (especially industrial development bonds). In many cases TIF is “the only game in town” for financing local economic development.

The basic rules of the game are illustrated in Figure 1. The top panel shows a land area view of a hypothetical municipality. The area on the western border is designated a TIF district and its assessed value is measured. The lower panel of Figure 1 shows the base-year property values in the TIF (B) and the non-TIF (N) areas. At a later point in time, assessed property values have grown to include the increment (I) in the TIF district and growth (G) in the non-TIF area of the municipality.

Tax increment financing carves out the increment (I) and reserves it for the exclusive use of the economic development authority, while the base-year assessed value (B) stays in the local government tax base. Thus,

• Before-TIF value = before TIF local government tax base = B + N;
• After-TIF value = B + N + I + G;
• After-TIF tax base available to local governments = B + N + G; and
• TIF district authority’s tax base = I.

Impacts on Overlapping Governments and Non-TIF Areas

The value increment (I) is the tax base of the TIF district. In most states (like Illinois, but unlike Massachusetts) there are multiple overlapping local governments, e.g., the municipality, school district, community college district, county, township, park district, library district, and other special districts. Figure 2 illustrates this situation with the school district representing all the nonmunicipal governments. To understand the economics and politics of TIF, it is crucial to note that while the municipality makes the TIF adoption decision, the TIF area value is part of the tax base of the school district and other local governments as well. Moreover, the TIF district gets revenues from the increment times the combined tax rate for all local governments together. The following hypothetical tax rates for a group of local governments overlapping a TIF district are close to the average proportions in Illinois.

Municipal tax rate = 0.15 %

School district tax rate = 0.60 %

Other governments’ tax rate = 0.25 %

Combined tax rate = 1.00 %

For each 15 cents of its own would-be tax revenues the municipality puts on the line, the school district and other local governments contribute another 85 cents. Thus, there may be an incentive for municipalities to “capture” revenue from growth that would have occurred in the absence of TIF (to collect taxes that would have gone to school districts). Or, municipal decision makers may favor inefficient economic development strategies that do not result in public benefits worth the full cost, since their own cost is only 15 cents on the dollar. TIF proponents would counter that nothing is captured, because the increment to the tax base would not exist “but for” the TIF authority expenditure. That argument, of course, turns on what would have happened to property values in the absence of TIF.

If, as municipalities are often required to assert when they adopt TIF, all of the increment is attributable to the activities of the TIF development authority, then TIF is fair, in that the school district is not giving up any would-be revenues. If, as critics of TIF sometimes assert or assume, none of the increment is attributable to the TIF and all of the new property value growth would have occurred anyway, then the result is just a reallocation of tax revenues by which municipalities win and school districts lose.

The impact of TIF on growth in property values requires a careful reading of the evidence. It is wrong, as those who look only at growth within the TIF district in effect do, to assume to know the answer. Part of the solution is to use appropriate tools to statistically control for other determinants of growth.

It is also necessary to take into account the potential for reverse causality. We want to know the extent to which TIF adoption causes growth. But the causation could go the other way; anticipated growth in property values could lead to TIF adoption if municipalities attempt to capture revenues from overlapping governments. Or there could be reverse causation bias if TIF is adopted in desperation by municipal decision makers in areas where low growth is anticipated. Either way we should ask: Are the municipalities that adopt TIF systematically different from those that do not? If the municipalities are systematically different, we must statistically disentangle the effect of that difference from the effect of the TIF using a technique that corrects for what economists call “sample selection bias.”

Impacts on Growth and Property Values

There are two sides to any government budget: revenues and expenditures. As a revenue-side mechanism, TIF is a way of earmarking tax revenues for a particular purpose, in this case local economic development. The effectiveness of economic development expenditures depends on opportunities, incentives, and planning skills that are specific to each local area and each project. By combining data from a large number of TIF and non-TIF municipalities, we can ask: On average and overall, is TIF adoption associated with increased growth in municipal property values? We have addressed this question in two research studies, both of which use statistical controls for the other determinants of growth and for reverse causation due to sample selection bias.

The first study (Dye and Merriman 2000) uses data from 235 Chicago area municipalities and covers preadoption, TIF adoption (or not), and postadoption time periods. We control for the selection bias (reverse causation) problem by first predicting which municipalities adopt TIF and then using that information (a statistic called the inverse Mills ratio) when estimating the effect of TIF adoption on property values in a second stage. Use of selection bias correction was first applied to the study of TIF by John Anderson (1990) and is now standard practice.

Our estimates of the impact of TIF have a number of additional variables controlling for home-rule status, the combined tax rate, population, income per capita, poverty rate, nonresidential share of equalized assessed value (EAV), EAV per square mile, distance to the Chicago loop, and county of location. We found that property values in TIF-adopting municipalities grew at the same rate as or even less rapidly than in nonadopting municipalities. The study design did not get at this directly, but the offset seemed to come from smaller growth in non-TIF area of the municipality (lower G).

Our findings were a surprise to those, especially nonacademics, who naively had inferred TIF caused growth by observing growth within a TIF district (I) without any statistical controls for the other determinants of growth (in I or G). Our findings were quite threatening to those with an interest in TIF, such as local economic development officers who spend the earmarked funds or TIF consultants who are paid for documenting findings of “blight” or “but for.” Our findings were also at odds with an Indiana study that found a positive effect of TIF adoption on housing values (Man and Rosentraub 1998).

Because our findings were controversial, because the effect of TIF was unsettled in the academic literature, and particularly because we wanted to pursue the possibility of a negative cross relationship between growth in the TIF district (I) and growth outside the TIF district (G), we undertook a second study (Dye and Merriman 2003). In addition we wanted to look at whether there are different TIF effects when more municipalities are included and different types of land uses are considered. We used three different data sets: property value data for 246 municipalities in the six-county Chicago area; less complete property value data for 1,242 municipalities in all 102 Illinois counties; and property value data for 247 TIF districts in the six-county Chicago area.

For the six-county sample (similar to our earlier study, but with more years and more municipalities), Table 1 presents the pre- and postadoption growth rates for the TIF-adopting and nonadopting municipalities. These calculations are from raw data, before any statistical controls for other growth determinants or corrections for selection bias. The first row compares EAV growth rates of the TIF-adopting and nonadopting municipalities in the period before any of them adopted TIF. EAV grew slightly faster for municipalities that would later adopt TIF.

The second row shows that in the period after TIF adoptions took place, gross-of-TIF EAV grew less rapidly for TIF adopters. The last row shows that the net-of-TIF EAV growth rate for TIF adopters was even lower, suggesting that growth (I) in the TIF district may come at the expense of property values outside the development area (G). In summary, if we make no statistical adjustment for the effects of other determinants, TIF adopters grew more slowly than nonadopters.

When we use the more recent six-county data in a multivariate regression model with statistical controls for local characteristics and sample selection, we no longer get the earlier provocative result of a significantly negative impact of TIF adoption on growth, but we still find no positive impact of TIF adoption on the growth in citywide property values. Any growth in the TIF district is offset by declines elsewhere.

The second study was designed with particular attention to land use. The property value data is broken into three land use types: residential, commercial, and industrial. Each TIF district also is identified by one of five development purpose types: central business district (CBD), commercial, industrial, housing, and other or mixed purpose. Thus, we can look separately at growth in municipal EAV by type of land use and type of TIF. Unfortunately, the data do not record EAV by land use within TIF districts, so we must settle for the growth in the tax base that is available to local governments. Most of the estimates of effects by land use type are not significantly different than zero. However, commercial and industrial TIF districts both show a significantly negative impact on growth in commercial assessed values outside the district.

The second study also extends the analysis to all 102 Illinois counties, which results in a much larger sample of municipalities (see Table 2). The TIF-base EAV (B) is unavailable, so we look at growth in available EAV. The simple means from the larger sample again suggest a negative effect of TIF on growth in property values. When we use this all-county sample to estimate the impact of TIF in a multivariate regression with statistical controls for other growth determinants and for TIF selection, there is a significantly negative impact of TIF adoption on growth in overall available (non-TIF) property values. This revives the earlier hypothesis that TIF adoption actually reduces property values in the larger community.

When we run separate regressions for available EAV growth by type of land use for the all-county sample, we see more evidence of a zero or negative impact of TIF on property value growth. Again, there is a significant “cannibalization” of commercial EAV outside the TIF district from commercial development within the TIF district.

The TIF district sample of the second study includes 247 TIF districts in 100 different municipalities in the six-county Chicago area. We match TIF base (B) and TIF increment (I) in each year to information for the host municipality. The key results are:

• Enormous variation in TIF district size, with an average base of around $11 million.
• Enormous variation in TIF district EAV growth rates around an average of 24 percent growth per year.
• TIF districts that start with a smaller base tend to have higher rates of growth.
• Most of the TIF growth occurs in the first several years, and growth rates decline an average of about 1 percent per year after the initial surge.
• Growth rates in the host municipalities are generally much smaller in the TIF district (an average of 3 percent compared to the TIF average of 24 percent).
• The estimated relationship between TIF growth and municipality growth is U-shaped; starting from zero, higher growth in the host municipality means lower growth in the TIF district, but the relationship turns positive at a host municipality growth level of about 6 percent.

Conclusion

Tax increment financing is an alluring tool. TIF districts grow much faster than other areas in their host municipalities. TIF boosters or naive analysts might point to this as evidence of the success of tax increment financing, but they would be wrong. Observing high growth in an area targeted for development is unremarkable. The issues we have studied are (1) whether the targeting causes the growth or merely signals that growth is coming; and (2) whether the growth in the targeted area comes at the expense of other parts of the same municipality. We find evidence that the non-TIF areas of municipalities that use TIF grow no more rapidly, and perhaps more slowly, than similar municipalities that do not use TIF.

Policy makers should use TIF with caution. It is, after all, merely a way of financing economic development and does not change the opportunities for development or the skills of those doing the development planning. Moreover, policy makers should pay careful attention to land use when TIF is being considered. Our evidence shows that commercial TIF districts reduce commercial property value growth in the non-TIF part of the same municipality. This is not terribly surprising, given that much of commercial property is retailing and most retail trade needs to be located close to its customer base. That is, if you subsidize a store in one location there will be less demand to have a store in a nearby location. Industrial land use, in theory, is different. Industrial goods are mostly exported and sold outside the local area, so a local offset would not be expected. Our evidence is generally consistent with this prediction of no offset in industrial property growth in non-TIF areas of the same municipality.

Richard F. Dye is a visiting fellow at the Lincoln Institute of Land Policy in 2005–2006. He is also the Ernest A. Johnson Professor of Economics at Lake Forest College, Lake Forest, Illinois, and adjunct professor at the Institute of Government and Public Affairs, University of Illinois.

David F. Merriman is professor of economics at Loyola University of Chicago and adjunct professor at the Institute of Government and Public Affairs, University of Illinois.

References

Anderson, John E. 1990. Tax increment financing: Municipal adoption and growth. National Tax Journal 43: 155–163.

Dye, Richard F., and David F. Merriman. 2000. The effects of tax increment financing on economic development. Journal of Urban Economics 47: 306–328.

———. 2003. The effect of tax increment financing on land use, in Dick Netzer (ed.), The property tax, land use, and land-use regulation. Cheltenham, UK: Edward Elgar, 37–61.

Dye, Richard F., and Jeffrey O. Sundberg. 1998. A model of tax increment financing adoption incentives. Growth and Change 29: 90–110.

Johnson, Craig L., and Joyce Y. Man (eds.). 2001. Tax increment financing and economic development: Uses, structures and impact. Albany: State University of New York Press.

Man, Joyce Y., and Mark S. Rosentraub. 1998. Tax increment financing: Municipal adoption and effects on property value growth. Public Finance Review 26: 523–547.