Diagnosis murder: the death of state death taxes.

Conway, Karen Smith ; Rork, Jonathan C.

Since 1976, more than 30 states have eliminated their "death" taxes and many others have reduced them. This unexplored case of interstate tax competition presents a unique opportunity to develop a new, more satisfying definition of competitor based on historical elderly migration patterns. Using data from 1967 onward, we outline the recent history of state death tax competition and present a spatial econometric analysis. Interstate tax competition is evident and grows stronger when using migration-based definitions of competitors. The article concludes with still more evidence of interstate tax competition the recent movement by states to effectively revive their death taxes. (JEL H7, D7)

I. INTRODUCTION

The federal estate tax has received wide attention especially because its elimination has been a centerpiece of President George W. Bush's tax proposals. (1) Largely overlooked, however, is the quiet revolution that has been taking place at the state level. Since 1976, 31 of the 48 contiguous states have repealed their "death" or estate, inheritance, and gift (EIG) taxes, instead relying only on the pick-up tax whereby states capture a portion of the federal estate tax liability but do not increase the overall liability of the estate. (2) Of the remaining 12 states that still have EIG taxes, 2 have enacted legislation that will eliminate them by 2005 and others are considering doing so. Beyond the outright elimination of EIG taxes, many states have also acted to reduce them in a variety of ways, such as exempting certain beneficiaries, such as the spouse.

This trend is noteworthy for several reasons. Foremost, it appears to us a prime example of intense interstate tax competition due to the growing size and political influence of the elderly population. There is additional political pressure because states may worry that high EIG taxes will drive the high-income elderly to move to bordering states or retirement havens. The stakes may be substantial. For example, Longino and Crown (1989) estimate that Florida had a net gain of $5 billion in income from the elderly migrants it received between 1985 and 1990, and Sastry (1992, 73) estimates that one new job is created for every 2.5 elderly migrants it receives. State EIG tax competition also provides us with a unique opportunity to explore alternative definitions of competitor states beyond simple geography because the movements of the tax base--the elderly--are fairly easy to track via historical migration data. Yet no research to our knowledge has explored interstate EIG tax competition.

Furthermore, these widespread changes in state EIG taxes provide substantial cross-sectional and time-series variation that has been mostly overlooked by researchers interested in the behavioral effects of estate taxes. (3) State EIG tax policy is also still in flux. Current and proposed changes in the federal estate tax have substantial revenue consequences for states, effectively eliminating EIG tax revenues for those that rely solely on the pick-up tax. How will the states react to this change? At present, several states have enacted or are considering legislation that would effectively decouple their EIG taxes from the federal estate tax and so would preserve the revenue source. Indeed, as the federal estate tax is slowly eliminated, EIG taxes in states that continue to use them may be the only remaining taxes on bequests. Yet little is known about them.

Our research seeks to fill this void. Using state-level data from 1967 to the present, we first describe the brief history and geographical pattern of these widespread reductions in state EIG taxes. To our knowledge, ours is the first research to document in a systematic way this phenomenon. We begin with a chronology of the states that eliminated completely their EIG taxes (thereby choosing to rely instead only on the pick-up tax). However, there are many other ways that the states may reduce their EIG taxes, such as increasing exemptions or reducing tax rates. For example, in 1985 legislative sessions alone, nine states increased personal exemptions or fully exempted spouses (Eckl 1986). We therefore also describe this broader phenomena of tax reductions with an eye for evidence of state competition.

Given the many ways that states may reduce their EIG taxes and the subtleties of the timing of the wholesale elimination (many were phased in over several years), focusing only on one policy action is incomplete and likely to lead to erroneous conclusions. Rather, we explore the extent of state EIG tax competition by estimating each state's EIG tax reliance as a function of fiscal and political factors as well as the EIG tax policy of its competitors. Our empirical approach follows that of Case et al. (1993) and includes state and time fixed effects. This research contributes to the tax competition literature both by considering a tax that has been over-looked and by extending the definition of what constitutes "neighbors/competitors" beyond simple geography. Specifically, we use historical information on the migration of elderly individuals between states to construct our weights, which, we argue, should better capture the competitive climate (or mobility of the tax base) among states. Past tax competition research has relied solely on geography, and to our knowledge, no state competition research has used migration as a measure of Tiebout-type pressures. Using our migration-based definitions of competitor states has a substantial impact on our results and suggests a stronger degree of tax competition.

After investigating the presence and determinants of interstate tax competition in this way, we then look at what is currently happening in the state EIG tax policy arena. The current situation is rather remarkable in the swiftness with which the states are reacting to changes in federal estate tax policy that threatens to eliminate their pick-up tax revenue source. It is also strongly suggestive that state competition is again at work, this time however in the reinstitution (or reincarnation) of state EIG taxes.

II. PAST RESEARCH

Our research borrows insights from and has possible implications for at least three bodies of literature--state competition, elderly migration, and federal estate taxation. Our primary contribution is to the state competition literature, as our research first documents the phenomena of widespread reductions of state EIG taxes and then explores it empirically as a classic case of state tax competition. We also bring new insights to this area by considering alternative definitions of neighbor or competitor that go beyond simple geography. Because the tax base that the states may be competing over consists primarily of the elderly, our work intersects with research into Tiebout elderly migration. Finally, both because of the temporal and cross-sectional variability in state EIG taxes we uncover and because reducing EIG taxes may have behavioral consequences other than migration which might benefit the states (e.g., increased savings or labor supply), research on the federal estate tax becomes relevant.

State Tax Competition and Elderly Migration

Although there are many arenas in which states interact with one another, the prevailing literature can be broken down into three major areas--expenditures, program adoptions, and taxation. There are insights to be gained from all three, but the tax competition area is clearly the most relevant for us. The mobility of capital plays an important role in past models of tax competition. In these models, surveyed by Wilson (1999), capital flows between jurisdictions as it searches for the highest after-tax rate of return. A government will attempt to encourage this inflow by lowering the tax it places on capital to a level below that of its neighbors. In response, the first jurisdiction will often lower its tax rate, and a battle of under-cutting will often result. In a world of perfect capital mobility, the equilibrium outcome will be for states to eliminate the tax in question. In reality, capital is not perfectly mobile, because there is often a cost associated with relocation. Hence, one does not witness such extreme results because at some level, the savings received with lower taxation will not offset the additional moving costs incurred by relocating capital.

The spatial econometric techniques used by Case et al. (1993) and Baicker (2001) in their analyses of state expenditures have allowed for a recent wave of empirical research focusing on the dimensions, if any, in which tax competition is taking place. Not surprisingly, those taxes that are considered to have a mobile tax base, such as property taxes (Heyndels and Vuchelen 1998; Brueckner and Saavedra 2001; Revelli 2001), business taxes (Buettner 2001), and excise taxes on motor fuel and cigarettes (Rork 2003) are found to be subject to the forces of tax competition by neighboring jurisdictions. In addition, Case (1993) and Besley and Case (1995) argue for a political component to tax competition, in that voters are likely to look to taxes of neighboring states as a means of assessing economic conditions and hence the performance of elected officials at home. (4)

The lesson from this research appears to be that states are engaged in tax competition with their neighbors, especially when the tax base is thought to be mobile. Just what tax base are the states competing over in the case of EIG taxes? It appears likely that the tax base primarily consists of the elderly, although as we will discuss, eliminating EIG taxes may have other behavioral consequences. States may be competing for wealthy elderly individuals who might choose their residence on the basis of state tax policy. There is some support that elders are considering EIG taxes when they make migration decisions. Using migration flows of elderly individuals, Voss et al. (1988) and Conway and Houtenville (2001, 2003) find evidence that high state EIG tax rates discourage in-migration. However, the one study that examines individual level data and simulates the effects of an actual change in state EIG tax policy finds no significant effect (Dresher 1994). Interestingly, research in this area has also largely overlooked the dramatic changes in state EIG taxes over the past 30 years and instead has focused chiefly on cross-sectional differences. One exception is Bakija and Slemrod (2002), who look at the number and size of federal estate tax returns over time in relation to state EIG and income tax policies; their preliminary evidence suggests that the elderly do respond to state EIG taxes in their location decisions.

Whether attracting or retaining the elderly is a wise policy is also unclear. Though the presumption by many is that elderly migrants positively affect the state (Longino and Crown 1989 call them "pure gold" in estimating that Florida had a net gain of $5 billion in income from the elderly migrants it received during 1985-90), their total effect is likely more complex, especially on the public sector. A comprehensive analysis that weighs all of their possible effects does not exist, a void noted and discussed by Serow (1992).

Regardless of what research has found, it is clear that at least some practitioners believe that EIG taxes are a powerful motivator to move:

 Yes, we're supposed to be called Michiganians,
 but some of prefer the old Michigander. And wise
 old Michiganders with estates had preferred to die
 as Floridians, Texans, etc. Under Michigan's old
 inheritance tax, the deceased's estate would have
 to pay Michigan's inheritance tax even if the
 estate's worth was under the federal estate tax
 exemption for the first $600,000 in property
 interest for most estates. This inheritance tax
 could be quite costly (e.g. $77,400 if single person
 left all $600,000 in assets to niece). Therefore, in
 the past if [sic] made more sense for a Michigan
 resident, who would have to pay a larger
 inheritance tax, to become a Florida resident at
 retirement in order to pay a lower Florida "pick-up"
 tax. However, Michigan passed the Michigan
 Estate Tax Act for persons dying after September
 30, 1993. Now, for Michigan residents with
 Michigan property, the tax is equal to the credit
 allowed for state death taxes paid for federal
 estate tax purposes (i.e. the state "picks-up" the
 amount that would have gone.) Net effect: in a
 typical estate, there will be no out-of-pocket cost
 to Michigan residents because of this new tax
 scheme. Therefore, there is no reason to become a
 Floridian for death tax purposes (though
 warm weather may also have something to do
 with it) (The Gavel, February 1995,
 www.forestlaw.com/gavel1-95.html).

This quote suggests that elderly migration may have been a factor in Michigan's decision to eliminate its estate tax. Perhaps even more important is that it reveals that a state's competitors are not necessarily limited to its geographic neighbors, as is assumed in all of the tax competition studies we know of (and cite). Exploring whether this kind of thinking is typical is one of the purposes of this article.

Federal Estate Taxation

As will be discussed in the next section, state EIG taxes are tightly linked to the federal estate tax, which is currently being reduced and may be eliminated in the near future. In other areas of federal taxation, states have been shown to react to changes in federal taxes. Besley and Rosen (1998) show that states match increases in federal gasoline taxation with an increase in state tax rates. Goodspeed (2000, 2002) discovered a similar responsiveness between national and local income taxes among European states. In our descriptive analysis, we also see a pattern of reducing state EIG taxes during times of major federal tax reforms. In the last section of this article, we report on the way that states are currently reacting to the recent changes in the federal estate tax.

Most work involving federal estate taxation has focused on the behavioral responses of individuals, responses such as the savings and labor supply behaviors of both the individuals making the bequests and the beneficiaries (see Gale and Slemrod 2000 for an excellent review of the literature and the articles in Gale et al. 2001 for recent advances). Past research has also considered the motives under-lying bequests by exploring whether individuals take full advantage of lower gift taxes via inter vivos transfers (e.g., Kuehlwein 1994; Joulfaian 2000a; McGarry 2001; Poterba 2001; Page 2003). Still others have considered the impact on charitable contributions (Joulfaian 2000b; Auten and Joulfaian 1996) and capital gains realization (Auten and Joulfaian 2001).

This line of research demonstrates many other ways that states may be affected by eliminating their EIG taxes beyond simply retaining or building their tax base of elderly individuals. In a recent work, Blumkin and Sadka (forthcoming) go one step further by demonstrating the properties of an optimal estate tax in the presence of a complete set of tax instruments. An interesting extension of our research, then, would be to use their theoretical framework to examine whether states are behaving optimally in their widespread reductions in EIG taxes.

Even more important, a look at past research on estate taxation reveals how the various behavioral responses could be better identified and ultimately understood if researchers took advantage of the variability introduced by the revolution in state EIG taxes. The promise of using state EIG taxes is evident in Page (2003), who uses the variability of state EIG taxes to consider the impact of such taxes on inter vivos giving. He points out that by looking at state EIG taxes, one can investigate a much wider range of estates (because the federal estate tax only taxes very large estates). Joulfaian (2000a, b) likewise uses state variability to help identify the effects of estate taxation on inter vivos giving and charitable contributions, respectively, although his analyses are only limited to federal estate tax returns. To our knowledge, only Bakija et al. (2003) have taken advantage of the variability in state EIG tax policy over time, which as we discuss next, is quite extensive.

III. A PRIMER ON STATE EIG TAXES

State EIG taxes are made up of three different types of taxes, (1) estate taxes, (2) inheritance taxes, and (3) gift taxes. (5) Estate and inheritance taxes are levied on the transfer of wealth that occurs at death and as such are determined by the decedent's state of residence. The primary difference between the two is that because inheritance taxes are levied on the individual beneficiaries of transfers made at death (and within two years of death), the tax rates may differ depending on the type of beneficiary. Typically, those beneficiaries with the closest relationship to the deceased have the lowest rates (e.g., a spouse would pay less than a nephew). Conversely, an estate tax is levied on the net estate of the deceased. The type of beneficiaries of the estate has no effect on the tax imposed unless the beneficiary is exempt (such as a surviving spouse). States impose an estate or an inheritance tax, and in either case the tax rates and exemptions vary widely across states. Gift taxes are not really death taxes per se, but they act to prevent death tax avoidance that could occur by transferring wealth prior to death. They also tend to have similar rates and exemptions as the estate or inheritance tax in the given state.

State EIG taxes have been in existence since 1826, but a permanent federal estate tax was not imposed until 1916 (Eckl 1986). (6) In 1924, Congress provided a credit against the federal tax for any EIG taxes paid to the states. In other words, the state EIG taxes paid are credited dollar for dollar against the federal estate tax liability owed, up to a certain amount. This is known as the pick-up tax, whereby the state picks up some of the federal tax liability without increasing the total tax liability of the estate. All states, with the exception of Nevada prior to the 1980s, take advantage of the pickup tax. We classify a state as eliminating their EIG tax, instead relying only on the pick-up tax, when their tax laws are rewritten such that the state tax liability is not to exceed this credit. We define the date as when those laws become effective. (7) These dates were found from various sources, including Significant Features of Fiscal Federalism, Eckl (1986), and direct correspondence with state officials.

A Quiet Revolution in State EIG Taxes

Prior to 1960, four states chose to levy only the pick-up tax--Alabama, Arkansas, Florida, and Georgia. Not only are all of these Southern states, but also the latter three are known as or are becoming retirement havens for the younger elderly (Conway and Houtenville 2003). The other states also took advantage of the pick-up tax, but they levied additional EIG taxes. Nevada is an exception in that it never enacted an EIG tax, but eventually took advantage of the pick-up tax in the late 1980s.

The year 1976 marks the beginning of a movement by the states to eliminate EIG taxes in favor of only the pick-up tax. Table 1 presents a chronology of the events, and Figure 1 provides information on the geographical pattern in the form of a map. New Mexico was the first to eliminate its EIG tax in January 1976, followed by Utah in January 1977. Four states--Arizona, Colorado, Vermont, and Virginia---eliminated their EIG taxes effective 1 January 1980. More generally, the trend to eliminate EIG taxes has not been steady; rather it has clustered around major changes in the federal tax law. (8) First of all, it is interesting that the movement began in the same year as the Tax Reform Act of 1976 (TRA76). TRA76 repealed the $60,000 exemption on estates and replaced it with a unified credit of $47,000, which is equivalent to an exemption of $175,625; this credit rose to an equivalent of $600,000 in 1981 (as a result of the Economic Recovery Tax Act of 1981) where it remained until increasing to $1 million in 1997 (as a result of the TRA of 1997) (Joulfaian 2000c). Thus, 1976 also marks the beginning of the federal estate tax law exempting a greater number of smaller estates, which perhaps placed pressure on the states to do likewise.

[FIGURE 1 OMITTED]

The early 1980s witnessed many state actions to reduce EIG taxes, with 11 states eliminating them between 1979 and 1983. Five more states eliminated them between 1985 and 1988, with two doing so in the same year as the Tax Reform Act of 1986. No states eliminated EIG taxes (although many acted to reduce them) until 1991, when four states eliminated them between 1991 and 1993. Once again, no actions were taken until 1997; since then, nine states have eliminated their taxes and two more have enacted legislation that will do so in the next few years. This suggests that the movement may be picking up momentum again. It also reveals that states may react to federal tax changes in setting their own EIG tax policy, as found in other policy arenas by Besley and Rosen (1998) and Goodspeed (2000, 2002). At the same time, however, the phasing out of the federal estate tax (and thus the pick-up tax revenues the states receive) has prompted several states to decouple their EIG taxes from the federal tax, which effectively reinstitutes a state tax on the estate. We discuss this development further in section VI.

Figure 1 reveals a limited geographic pattern to the movement. (9) The four states that eliminated EIG taxes prior to 1976 are obviously clustered in the Deep South. Yet Mississippi, which is in the same cluster, just eliminated its EIG tax in 2000, and its neighbor, Louisiana, is slated to do so in 2004. The Southwest, with the exception of Oklahoma, had eliminated EIG taxes by 1983, as did the Mountain states, with the exception of Montana, by 1988. With New Hampshire's and Connecticut's elimination of EIG taxes by 2005, the entire New England region, along with New York, is devoid of EIG taxes, although the actions occurred over a more than 20-year period. The states that continue to have EIG taxes form geographic blocks concentrated in the middle of the country and the Atlantic seaboard.

Of course, part of the problem with this analysis is that we are not controlling for other things. One of those things is the subtleties in the timing of the policy actions. Many were phased in over several years, and there is likely a significant lag between when the decision was made and when the law became effective (witness Lousiana and Connecticut). Another important factor is that the states have many ways of reducing their EIG taxes beyond wholesale elimination. We discuss some of those changes next.

Other Ways that States May Compete

An examination of state EIG tax policy changes reported in various editions of Significant Features of Fiscal Federalism spanning the early 1980s to mid-1990s reveals that states made many different kinds of changes to their EIG taxes during the period. (10) Probably the most common action is to increase exemptions for taxable estates, at times exempting entirely the spouse (e.g., New York as of 1/1/ 84, Connecticut and New Jersey as of 7/1/88, and Iowa after 1987) or any homestead created under the civil code (California, effective 1/1/ 1981) or community property (Washington, effective 1984). A common pattern is for states to begin increasing their exemptions a few years prior to the elimination of their state EIG tax. For instance, Oregon steadily increased the exemption for net tax taxable estates from 1980 to 1986 before eliminating the estate tax entirely for estates of decedents dying on or after 1 January 1987. Texas, Washington, South Carolina, and Massachusetts behaved similarly in increasing the exemption in the years leading up to the elimination, whereas Rhode Island and Wisconsin had policies that phased out the percentage owed in each year for their tax such that it approached the pick-up tax. Michigan exempted family businesses and farms from its inheritance tax in 1992, one year before it eliminated the tax altogether.

During the same period, however, states made other changes to their laws, some of which effectively increased their EIG taxes. For example, Connecticut placed a 10% surcharge on inheritance taxes in 1983 and shortened the due date for inheritance tax payments from nine to six months after date of death in 1990 (while increasing the exemptions during the same period). Likewise, in 1990 New York both expanded the estate and gift tax base to include out-of-state property and, similar to its neighbor Connecticut, accelerated both taxes by shifting their administration to the Department of Taxation and Finance. (11) In the year before they eliminated their EIG taxes, both Rhode Island and Michigan like-wise accelerated the collection of these taxes. Other changes include an increase in inheritance tax credits (North Carolina from 1987 to 1989), a reduction in inheritance tax rates (Tennessee, which by reducing its rates on all other beneficiaries to match those of the lowest category effectively switched to an estate tax after 1989), and a change in rates and brackets for the estate tax (Mississippi after 10/1/89). States therefore have a multitude of ways to adjust their state EIG tax policy, some subtle, such as collection and administration policies, and others more obvious, such as eliminating their tax to rely only on the pick-up tax.

Are there obvious patterns in these smaller changes? Our limited inquiry leads us to believe so. The vast majority of the changes during this period effectively reduced taxes, either by exempting entire classes of beneficiaries or estates or by raising exemptions or lowering tax rates. As mentioned earlier, many of these reductions were precursors to the wholesale phase-out of the tax. Those actions that increased taxes tended to be more subtle (e.g., accelerated collection) or temporary (e.g., Ohio's delay of making spouses exempt). Likewise, the same states tend to appear in these report summaries, such as Connecticut and New York. Geography also appears to play a role as revealed by the flurry of changes taking place in Connecticut, New York, Rhode Island, and New Jersey in the late 1980s and early 1990s, and in North Carolina, South Carolina, and Tennessee during the same period. (12)

These reports also make clear that focusing on one policy change--the date on which the elimination of state EIG taxes in favor of the pick-up tax became effective-is missing other important avenues for state competition and also the subtleties of how it took place. For instance, Wisconsin eliminated its tax effective on 1 January 1992. However, it began phasing out its tax in 1987--the year after its neighbor Minnesota eliminated its tax and three years after another neighbor, Illinois, eliminated its tax. Likewise, although Wisconsin's neighbor Iowa continues to have an EIG tax, it completely exempted spouses from inheritance taxation after 1987, the year in which Wisconsin began phasing out its tax. Similar interactions occurred between other neighboring states.

Ideally, one would study the individual policy changes, big and small, and the interactions between states; however, quantifying and distilling those changes into a variable to analyze is very difficult. We therefore focus on the EIG tax revenues generated by each state as the best summary measure of their policies. What has happened to state EIG tax revenues during this period? Figure 2 reports the average proportion of state tax revenue that came from state EIG taxes in every year from 1968 until 1999. Interestingly, we see that the prominence of state EIG taxes began to fall well before New Mexico kicked off the pick-up revolution in 1976. In fact, it fell most steeply between 1968 and 1976, and continued its steep descent into the mid-1980s where it has held fairly steady except for an uptick in the late 1990s (which is likely due to large increases in estate sizes due to the economic boom). This figure highlights that the competition on state EIG taxes is not just limited to wholesale elimination; it also emphasizes the need to control for the level of wealth in estimating state EIG tax revenues, which we are able to do using state and time fixed effects.

[FIGURE 2 OMITTED]

IV. DEFINING COMPETITORS AND OTHER MODEL SPECIFICATION ISSUES

The data used in our analyses encompasses 32 years of history from 1967 to 1999, because 1999 is the last year for which information on some of our variables is available. We follow the practice of Case et al. (1993) and others of limiting our analysis to the 48 contiguous states. Alaska and Hawaii are typically eliminated both because they have no geographic neighbors (and thus in typical tax competition research are assumed to have no competitors) and because their economies and fiscal policies are so different from the other 48 states. Annual state financial data, including tax revenues, expenditures, and debt issues, is compiled from various issues of State Government Finances. Political variables along with state unemployment rates and the amount of money transferred from federal to state government were compiled from the Statistical Abstract of the United States. Series P-25 of Current Population Reports, in conjunction with the Statistical Abstract of the United States, was used to gather breakdowns of state population by age including those aged 65 and older as well as 85 and older. (13) We also use data on elderly migration. The elderly migration flows during 1965-70 that is, how many elderly moved between each pair of state--is obtained from the 1970 Public Use Microdata Samples (PUMS) using Micro-Analyst software.

The Econometric Model

We follow Case et al. (1993) by assuming that our EIG tax reliance measure is a function of state characteristics ([X.sub.it]), state and time fixed effects, and the tax measure of its 'competitors'. Denoting Tit to be the tax measure in state i at time t yields a linear relationship of

(1) [T.sub.it] = [X.sub.it][beta] + [theta][T.sub.jt] + [[delta].sub.t] + [[zeta].sub.i] + [u.sub.it]

where [T.sub.it] represents the competitor's tax measure; [[delta].sub.t] and [[zeta].sub.i] are year and state fixed effects, respectively; and [u.sub.it] is a mean zero, normally distributed random error. Because states have more than one competitor, Tit is replaced by [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where [W.sub.ij] represents the weight assigned to state j. We discuss in the next section the four different sets of weights we use.

Each state, therefore, has a set of weights (which can be thought of as a 48 x 1 vector) signifying how important another state's tax measure is in its own determination. Equation

(1) can be written as

(2) [T.sub.it] = [theta][W.sub.i][T.sub.t] + [X.sub.it][beta] + [[delta].sub.t] + [[zeta].sub.i] + [u.sub.t],

where [W.sub.i] is the ith row of a weighting matrix, W. W is 48 x 48 matrix that assigns competitors to every state. This is known as a spatial lag model.

The inclusion of [T.sub.t] on the right-hand side of equation (2) means that estimation by ordinary least squares (OLS) will be inconsistent, due to correlation with the error or simultaneity bias. Intuitively, if tax competition is taking place, then the EIG policies of one's competitors are going to be simultaneously determined with one's own--for example, Alabama's policies are affected by what Georgia is doing, and Georgia in turn will react to what Alabama does. To deal with this endogeneity, we estimate equation (2) using the instrumental variables approach outlined in Kelejian and Prucha (1998) in which weighted values of the exogenous variables (WX) are used as instruments. (14) This approach estimates a first-stage equation in which competitor EIG tax reliance (WT) is regressed on the spatially weighted values of the other characteristics (WX), which identifies the structural equation. Intuitively, the model is identified with the assumption that the other characteristics (e.g., state debt per capita) of one's competitors do not directly affect EIG reliance once the competitors' EIG reliance has been controlled for. This assumption seems reasonable to us, that economic and demographic variables of competitor states will not have an effect on EIG reliance over and above their effect on the competitor's EIG reliance. As a check of this approach, we perform three separate tests of these overidentifying restrictions (i.e., that WX is uncorrelated with u in equation [2]) required for our instruments, the Hausman, Sargan N*[R.sub.2] and Basmann tests, and our model passes all three. We also verify that WX is jointly statistically significant in the first-stage regression (such that it is correlated with WT).

A final consideration is the likelihood that the error term, u, in equation (2) is both heteroscedastic and serially correlated. The heteroscedasticity may arise because of the sampling variability of our dependent variable--one would expect a small state to have a larger variance than a large state, because the former is going to be more strongly affected by the number and type of people who die each year. The correlation may arise because we are following the same states across time and there may be serial correlation within a state over time that is not captured entirely by the state fixed effects. In addition, other types of correlation may exist (e.g., across states at a point in time) that are not captured by our model. For all of these reasons, we calculate our standard errors via the bootstrap method in which 500 random samples of our original sample size are drawn with replacement and then used to estimate 500 tax reliance equations. These 500 sets of estimated coefficients are used to construct the bootstrap estimates of the coefficient covariance matrix (standard errors and covariances). Covariance estimates calculated via bootstrap are robust to correlation among the errors and heteroscedasticity.

Constructing the Spatial Weights

Because the purpose of our analyses is to discover the factors that push a state toward reducing its EIG taxes, defining which states are competing with one another (i.e., defining W in equation [2]) is critical. Accordingly, we use four different weighting matrices in exploring which states are competitors, or neighbors, with one another.

For comparison with past studies of state competition, we begin with two typical geography-based weights before moving on to two other weights that use real economic information in the form of elderly migration patterns. The first set of geographic weights used is known as contiguity weights. States are considered competitors if they share a common geographic border. In the case of New Hampshire, for instance, Massachusetts, Maine, and Vermont would be considered competitors, whereas Alabama would not. Each neighboring state would be given equal weight, as they are assumed to exert equal influence. It is common for weights to be row-standardized, meaning that the weights sum up to one. Continuing with the New Hampshire example, Massachusetts, Vermont, and Maine would each be given a weight of 1/3, whereas Alabama would be assigned a weight of zero.

In this example, it may be the case that Massachusetts wields more influence on New Hampshire than Maine, as it has a larger population. Thus, we create a second set of geographic weights known as population-contiguity weights. States must still share a common geographic border to be considered a competitor, but now the weight varies on the basis of population. To create the weight for Massachusetts, the population of Massachusetts would be divided by the entire bordering population (i.e., the population of Massachusetts, Vermont, and Maine). Note that in this case the weights are row standardized. A non-row-standardized weight for Massachusetts in this case would be just its population.

Our other two measures are in line with the aforementioned quote from The Gavel, which alluded to how states may consider the policies of other states to which they are losing or gaining large numbers of elderly individuals. Thus, we construct two other sets of weights, referred to as flow and corridor, which are based on historical elderly migration flows using 1965 70 migration flow data from the 1970 PUMS. State j will be considered a flow competitor of state i should it be the case that there are elderly migrants moving from state j to state i, that is, state i is receiving elderly migrants from state j. Given that most interstate elderly movement is across borders, this measure will still be heavily based on geography, but this measure allows us to also account for Michigan-Florida or Michigan-New York competition as well. Similar to the way the population-contiguity weights are created, the flow weight for state i (with state j) is simply the total elderly immigrants into state i that originated from state j divided by all the elderly in-migrants into state i, regardless of state of origin. Thus, if 30% of New York's elderly in-migrants came from Connecticut, Connecticut would be assigned a weight of 0.3 in being defined as one of New York's competitors.

States may also be concerned about losing their elderly residents, however, as evidenced by The Gavel quote. Our preferred and final measure is therefore one that captures the entire corridor of migrants between the two states--that is, the number of elderly individuals who are moving between the two states. State j is a corridor competitor of state i should it be the case that state j is either sending or receiving elderly migrants to/from state i. This measure assumes that a state views itself as competing with another state if it is either losing elderly individuals to it or luring them from it. Note that any states that were flow competitors are automatically corridor competitors, although the weight will differ. The weights are calculated similar to the flow weights, with the exception that the weight is based on the percentage of the total corridor a state accounts for, rather than the percentage of the total flow. This will place a greater weight on those states for which there is a relatively large amount of movement of the elderly, regardless of direct. (15)

In constructing the flow and corridor weights, we use historical migration flow data from 1965-70 to avoid endogeneity. State actions to eliminate EIG taxes may affect elderly migration patterns; indeed, one hypothesis maintains that it is the goal of these actions. We therefore use the same weights for the entire period the data cover. One alternative is to update the measure every 10 years as it becomes available through the census. However, this would lead to discrete changes at 10-year intervals, some of which may be due to sampling error as the flow estimates are based on a sample of the total population. We can also take comfort in the fact that primary elderly migration patterns, especially corridors, have changed fairly slowly over time. (16) Finally, it is important to note that because of the prevalence of border moves, all four measures place some weight on geographic neighbors. The latter two measures augment geography by taking account of movements beyond the bordering states and are based on what we hypothesize the states are competing over.

Choice of Variables

The means of our dependent and explanatory variables are listed in Appendix Table A. 1. Although imperfect, we believe that state EIG tax revenues are the best summary measure of a state's approach to EIG taxation. It should capture all of the deliberate changes that a state makes to its EIG tax laws and distill them into a measure that the state itself should care about: the revenues produced. It is also a measure that is easy for taxpayers to understand. In addition, although state EIG tax revenues are going to be affected by the levels and types of individual wealth, the richness of the data allows us to control for these confounding effects with state and time fixed effects, as well as our explanatory variables.

Our primary measure of state EIG tax policy, T in equations (1) and (2), is the tax share measure, calculated by dividing state EIG tax revenues by overall state tax revenues. Taxpayers and policy makers should be considering the entire tax structure of the state when making their decisions, and this measure allows for that. Furthermore, it is not confounded by the size of the public sector. In our sensitivity analyses, we consider several alternative measures, including per capita state EIG taxes and state EIG taxes per death. One might argue that taxpayers would use these measures as a best guess of their approximate burden from the tax; however, we argue that they suffer from being an absolute rather than relative measure. (17) Other possible measures include marginal or average tax rates for a certain size estate; however, as the discussion in section III reveals, a state's EIG tax system has many components beyond the tax rate for a given estate, and so such a measure also suffers from being incomplete. Rather, as one of our sensitivity checks we construct a pseudo-average tax rate using information about total federal gross estates in each state, which is available for a small number of years.

To isolate evidence of tax competition, we must include any other economic or political factors that we believe might affect a state's EIG policy over time, written as X in equations (1) and (2); however, we must also include any variables that help capture variation in our EIG tax reliance measure (T) that is not due to deliberate state EIG policy, such as a change in the number and size of the estates in the state. One variable that belongs for both reasons is the per capita income in the state. A wealthier state may choose a different EIG policy both for economic and political reasons, but it is also more likely to have a greater number of large estates, which will generate larger EIG revenues, regardless of the state's actual EIG policy. Per capita income also helps control for the tendency of states that have experienced the most in-migration (from all age groups) to also have high growth in income. Another variable with dual roles is the percentage of the population that is over age 65 and over age 85. We include both over 85 and over 65 because we hypothesize that the two age groups exert different sorts of pressures. The younger elderly may be a more politically active group than the older elderly, and there is evidence that their migration decisions are more sensitive to state fiscal policies (Conway and Houtenville 2003). The size of the older elderly population is more likely to approximate for the potential tax base; however, both variables may have confounding effects on state EIG revenues because they may capture both political pressures and the tax base.

To capture the state's fiscal situation (and thus the ability to reduce EIG taxes), we include the per capita federal transfers received by the state, the state debt per capita, and the unemployment rate. (18) The political environment is reflected by whether it is an election year, and dummy variables for whether both houses and the governor are Democratic or Republican.

We also include the percent of state tax revenue that is raised through personal income taxes and, separately, sales taxes. These variables capture the availability of other sources of revenue, and so a strong presence of these taxes may make it easier to eliminate EIG taxes. In addition, potential elderly migrants likely consider the entire tax climate of the state in the decisions to move, and there is some evidence that they are discouraged by an income tax (Cebula 1990). Thus, a state with a heavy income or sales tax burden may feel more pressure to reduce their EIG taxes to compete for wealthy elderly migrants. However, one might argue that states make decisions about all of their tax policies simultaneously, so that including these two alternative tax sources could lead to bias. We first address this by including the previous year's values for these taxes in all specifications. In addition, as a robustness check, we explore the impact of excluding these variables from the model.

Finally, we include state and year fixed effects in all specifications. The state effects should capture any unobserved state characteristics that are stable over time and would subsume any regional effects. The year effects control for changes in the federal tax code that may influence a state's decision; they also help capture business cycle effects, although not perfectly because states may be affected differently. However, including the state unemployment rate should help capture these differences.

V. EMPIRICAL RESULTS

Our main model therefore includes the aforementioned explanatory variables plus state and time fixed effects, uses four different weighting schemes and our tax share measure of state EIG tax policy, and uses annual data from 1967 to 1999. However, as will be discussed further, we subject this baseline model to several model specification checks. These checks include excluding the income and sales tax variables, using alternative measures of state EIG reliance, taking account of the fact that the states' ability to tinker with their EIG taxes has decreased over time because many now have only the pick-up tax, and using a random weighting scheme to explore whether our model is picking up only spurious correlation rather than state competition.

Main Model Results

Table 2 shows the results for the revenue share using the four different weighting schemes and annual data from 1968-99. Recall that all models include state and time fixed effects to help control for unobservables. For our key variable, the competitors' EIG tax reliance, the results are unequivocal; there is a substantial, statistically significant effect of the EIG tax reliance of competitor states, no matter how we choose to define competitors. Furthermore, the strength of this competition grows as our definition of competitor expands to include elderly migration patterns. The size of the effect is striking as well. Our estimates suggest that a 10% decrease in the EIG tax share of a competitor leads to between a 2.2% to 4.3% decrease in the state's own EIG tax share. These estimates are consistent with what has been found in the literature for other taxes with mobile tax bases. (19) Our empirical results therefore strongly suggest that states are paying attention to what their competitors are doing when they devise their EIG tax policy.

The other variables are also fairly stable across the different definitions of competitor. We find that personal income taxes are very important, suggesting that a heavy reliance on income taxes either enables or puts pressure on the state to rely less on EIG taxes. It appears as if states are avoiding having their constituents undergo double jeopardy in terms of wealth taxation. Sales taxes, on the other hand, do not appear important to a state's reliance on EIG taxation. The impact of having an older population leads to an increase in our EIG reliance measure. This is not surprising given that the elderly are the most likely group to be leaving an estate; as discussed earlier, these variables play dual roles making interpretation difficult. We return to this issue when we discuss our alternative specifications. The only other variable that has a significant influence is the unemployment rate, which has a positive effect on a state's reliance on EIG taxes.

Robustness Checks

Next we subject our baseline model to several robustness checks, the key results of which are summarized in Table 3. (20) First, we explore the impact of excluding the states' (lagged) reliance on personal income and sales taxes. The resulting estimated competitors' E1G tax reliance coefficients are reported in the second row of Table 3. Excluding these two variables has little impact on the results except to slightly decrease the magnitude of the effect. The other variables' coefficients (not reported) are likewise unaffected. For this reason and because personal income taxes are statistically significant, we continue to include these two variables in the rest of the analysis.

Second, we explore several alternative measures of EIG tax reliance to make sure that our results are robust to our choice of measure. The first two measures use an absolute rather than relative measure of the EIG tax burden. One could argue that the EIG share of state tax revenue, our main measure, could fluctuate because of changes to other sources of revenue--which could have little to do with tax competition. We therefore use, alternately, EIG taxes per capita and EIG taxes per death. The competitors' tax reliance coefficients are reported in the third and fourth lines of Table 3. In general, using either of these measures instead of the state EIG tax share yields similar results and even stronger evidence of interstate EIG tax competition. The estimated competitors' tax reliance coefficients increase in magnitude and range from 0.38 to 0.64. Most of the other coefficients also behave similarly across the two measures with a couple of exceptions. Personal income taxation appears less important and now has a positive effect, although it is not statistically significant. This is likely due to our now using an absolute measure. More important is that using EIG taxes per death diminishes the effect of having a very old population (percent over 85 years old). This lends additional support to our argument that including the age variables is capturing the fact that a larger elderly population is going to generate more EIG tax revenue. When one redefines to be EIG per death, this relationship is understandably weakened, as expected.

However, even these measures could fluctuate for reasons other than tax competition. Although including the percent of the population over age 65 and over age 85 helps control for the fact that states with big elderly populations will generate more EIG tax revenues, they do not capture the differences within the elderly. The rich elderly generate most of the EIG tax revenues, and the percentage of the elderly that are rich may vary a great deal from state to state. They may also react to state EIG tax policy so that reverse causality may exist. (Low competitors' EIG tax reliance may lure the rich elderly and erode the tax base, making the state's reliance also appear lower.) Likewise, asset values and wealth accumulation behaviors may vary across states and over time--and perhaps be influenced by state EIG tax policy.

Our third measure seeks to address all of these issues by using information about the aggregate value of gross estates reported on federal estate tax returns. (21) This variable should capture the geographic and temporal differences in the wealth of the elderly residing in the state, as well as any random variation in the types of people who die in a given year (e.g., a billionaire dies in a small state). Unfortunately, however, this variable is only available for a limited number of years--1970, 1977, 1983, 1986, 1987, 1990, 1993, and 1998. As noted in section III, the bulk of reductions in state EIG taxes came in the early years of our sample, which makes the missing early years of this data especially troubling. Given the volatile nature of the data and the large number of holes in the data, we believe we must drop the missing years from this particular analysis, which reduces our sample size by approximately 75%. For this reason, we use this information as a sensitivity check rather than include it in our main specification.

Using this information we construct our third alternative measure, which redefines state EIG tax reliance to be total state EIG tax revenues divided by the aggregate value of gross estates reported on federal estate tax returns. This measure can be viewed as roughly approximating an average tax rate on estates. As the federal estate tax exempts small estates, this measure will be larger in a state if it (1) taxes small estates not subject to federal taxation, and/or (2) levies high taxes on those estates subject to federal taxation. The results of this exercise are reported in the fifth line of Table 3. Again, the results are quite similar to our main specification and our other exercises. The estimated magnitude of tax competition is quite similar and tends to be even larger than in our baseline model, especially when migration-based weights are used. The statistical significance of these coefficients is diminished somewhat, which is not surprising given the much smaller number of observations and the loss of so many of the early years of apparent EIG tax competition. The other coefficients are affected in a manner similar to when we use per capita or per death measures in that personal income taxes are no longer important. This specification, however, completely eliminates the effects of having an elderly population on EIG revenues, further suggesting that the positive effects found in our baseline model are due to a larger EIG tax base. (22)

A related issue is that EIG taxes have the potential to be extremely volatile--for example, the death of one billionaire in a small state can greatly skew the numbers for that year. To make sure that such volatility is not driving our results and to smooth out the volatility, we reestimate the model using three-year moving averages of all of the variables to see if there is a difference. (23) The key results for this exercise are reported in the sixth row of Table 3. Here we see that the estimated degree of interstate tax competition grows, if anything, when moving averages are used. The other variable coefficients are similar with the two interesting exceptions that per capita federal transfers and election year, which were both negative before, are now also statistically significant. Therefore, our evidence of interstate EIG tax competition is not due to random volatility in EIG tax revenues and appears to be quite robust to alternative measures of EIG tax reliance.

Our next set of exercises recognizes that once a state abolishes its EIG tax and relies solely on the pick-up tax, it has also greatly reduced its ability to further tinker with its EIG tax policy. Rather, its revenues are completely determined by the federal estate tax code and the estates of its taxpayers. Because the number of pick-up states grew steadily during the period our sample covers, one would expect to see a weakening of interstate tax competition--evidenced by a shrinking competitors' EIG tax reliance coefficient--over time. This presents an interesting opportunity to test the validity of our results and perhaps dispel the concern that we are simply picking up spurious correlations in state EIG tax revenues. To accomplish this, we split our sample into the 1970s, the 1980s and the 1990s and then reestimate the model for each time period. (24)

The next three rows of Table 3 report the estimated competitors' EIG tax coefficients for each of the four weights across the three time periods. Immediately evident is how our estimates of the degree of interstate tax competition steadily decline over time. In the 1970s, the response is always statistically significant and ranges from 0.42 to 0.93, the latter of which is about double the response found for other taxes (see note 19). In the 1980s, the response is smaller (ranging from 0.22 to 0.371) but still significant. In the 1990s, however, the response is either zero or even negative. We are somewhat puzzled by the negative response that appears with our migration-based weights for the 1990s. Given that there is a fairly small number of states who can still compete during this period (25 at the beginning and 17 by the end of the 1990s), it is not clear what this measure is capturing, especially for the majority of states in the sample who had little control over their state tax revenues. If it is a measure of the underlying spurious correlation between state tax revenues, then it makes the strong, positive responses we find in the rest of the analysis all the more remarkable. More generally, the explanatory power of the model (as measured by the [R.sup.2] and the joint significance of the explanatory variables other than the year and state effects) steadily declines as one moves from the early years to the later years. This combined with the dramatic fall in the estimated responses over time is further support for our interpretation of these effects as interstate tax competition.

Another way to explore the issue is to allow the competitors' tax reliance coefficient to vary between pick-up and non-pick-up states. Because pick-up states no longer have any way to compete, we would expect their competitors' reliance coefficients to be zero. We accomplish this by adding an interaction term--whether a state is a pick-up state in the prior year multiplied by the competitors' tax reliance variable--which we also treat as endogenous. If our hypothesis is correct, then the coefficient on the interaction term should be of approximately equal magnitude and opposite sign, so that the sum is zero. We perform a t-test of this restriction--that is, whether the competitors' reliance coefficient for pick-up states (which is the sum of the main coefficient and the interaction term) is zero. Given that the majority of the tests are supportive, we also reestimate the model imposing this restriction by interacting a nonpick-up dummy variable with the competitors' tax reliance variable.

The results of these exercises are reported in the tenth through twelfth lines of Table 3. In the first exercise, strong evidence of our hypothesis exists as the coefficients on the interaction term are always negative, statistically significant, and of near equal magnitude as the primary coefficient. Furthermore, in three out of four cases we fail to reject the hypothesis that they sum to zero, suggesting that the effect of competitors' EIG reliance is zero for pick-up states. (In the fourth case, it is negative rather than positive suggesting that any spurious correlation is negative, if anything, and is working against evidence of tax competition.) Imposing this restriction typically leads to the estimated magnitude of tax competition to increase over our baseline model. Dealing with pick-up states in this way therefore strengthens our evidence of state EIG tax competition.

Our last robustness check attempts to verify that we are indeed capturing state tax competition and not spurious correlation or simply the random spread of a good idea (although the latter should be captured with our time dummies). Following Case et al. (1993), we reestimate our model using weights based on the alphabetical order of the states, which clearly should have nothing to do with competitive forces. Specifically, two states are defined as neighbors or competitors if they are adjacent to each other alphabetically (e.g., Arizona has Alabama and Arkansas for competitors; Alabama has Wyoming and Arizona for competitors). The estimated competitors' EIG tax coefficient from this exercise is reported at the bottom of Table 3. Using these nonsensical weights completely eliminates any evidence of competition, both in terms of magnitude and statistical significance, and confirms that our other analyses are not simply picking up spurious correlation among the states.

Our spatial analysis of state EIG tax reliance, then, provides strong evidence that interstate EIG tax competition has been quite intense over the past 30 years, especially in the earlier part of the period when most states had not yet finished their race to the bottom. The strength of this competitive response grows when we use a definition of competitor that goes beyond simple geography. It is also robust to using alternative measures of state EIG tax policy and to different treatments of the pick-up tax phenomena. Perhaps the most compelling evidence that states are competing in the EIG tax arena, however, is revealed with a quick look at what the states are doing now.

VI. THE RENCARNATION OF STATE DEATH TAXES?

Are state EIG taxes really dead? Apparently not. The federal tax cut package adopted in June 2001 is phasing out and will eventually eliminate the federal estate tax. This change will effectively eliminate the pick-up tax revenues for the states. It also reduces revenues for those states that have not eliminated their EIG taxes because they still take advantage of the pick-up tax. Specifically, the pick-up tax credit for state estate taxes paid will be reduced by 25% each year beginning in 2002 and is eliminated completely by 2005. The revenue consequences for the states is substantial; McNichol et al. (2003) report that states would have received approximately $6 billion from the tax credit in fiscal year 2003 alone had the law not changed. How are the states reacting to the change? As it turns out, many states are now effectively bringing their death taxes back to life.

As documented thoroughly in McNichol et al. (2003), states have the option of decoupling from the federal estate tax to preserve these tax revenues. Decoupling requires either changing (or, interestingly enough, failing to change) state law so that the automatic connection between the state estate tax credit in the federal code and the amount of state estate taxes owed is broken. In some states, state law references the federal estate tax law as of a specific date. For these states, all that is required to decouple is failing to update that date. For others, state law makes no reference to a specific date and so a state must initiate changes to its law to preserve their EIG tax revenues. For example, Rhode Island decoupled in this way during its 2001 legislative session by setting the state's estate tax equal to the amount of the federal credit as determined by IRS code as of 1 January 2001. Wisconsin acted similarly. Of course, states could also decouple by enacting a new, separate estate or inheritance tax. In all cases, however, the state EIG tax liability net of the federal tax credit will increase as a result.

According to an August 7 report by McNichol (2003), 12 states have taken action to decouple, and another 6 plus the District of Columbia will remain decoupled unless they take action. Figure 3 displays on a map these 18 states and also shows which states have or have not eliminated their EIG taxes in favor of the pick-up tax. (25) A strong geographic pattern is immediately obvious, especially along the Eastern seaboard, where the line of decoupling is virtually unbroken (except for New Hampshire, Connecticut, and Delaware) from Maine to North Carolina. Likewise, there are geographic clusters in other parts of the country (e.g., Washington-Oregon and Minnesota-Wisconsin-Illinois). When states that still have EIG taxes are included, one sees an even stronger geographic pattern to state EIG taxation. The Eastern seaboard, Mid-Atlantic, and Midwest are much more likely to impose a state EIG tax--either by continuing to retain their tax, decoupling from the federal law, or both. In contrast, the Deep South, Southwest, and Mountain states are so far completely devoid of decoupling, and all have eliminated their EIG taxes. It therefore appears that a state is more likely to have decoupled if its geographic neighbors also have decoupled or still have a EIG tax, which is compelling evidence of interstate competition. Three states with a long history of no EIG taxes--Alabama, Florida, and Nevada, two of which are heavy net importers of the elderly--would likely need to alter constitutional provisions on the amount of estate taxes levied (McNichol 2003). California would require a vote of the people.

[FIGURE 3 OMITTED]

What this illustrates is that state EIG tax policy is still very much in flux. States appear to be paying attention to one another, or at the very least are behaving in geographic clusters. It has taken 25 years for 30 states to completely eliminate state EIG taxes in favor of the pick-up tax (and many other states have likewise reduced their taxes during the period). In contrast, within the past two years, 18 states, 13 of which had eliminated their EIG taxes, have effectively increased their EIG taxes in response to the threat of losing their pick-up tax revenues. Why this sudden reversal in state EIG tax policy? A possible political explanation is that decoupling is a very subtle and largely invisible way of increasing a tax. However, a recent article in the Boston Globe suggests the contrary with its quote from estate planning attorney, Stephen Ziobrowski, "You don't see much advice suggesting people leave Massachusetts any more, but maybe attorneys will just re-draft those old letters and suggest that their clients move to Florida.... That's where people used to go to avoid Massachusetts's estate tax, and it may be what they start doing again." (26) Perhaps it is the current economic downturn that is forcing states to take any actions possible to retain and build up their tax revenues. It might even be a combination of factors--the reductions in EIG taxes have not produced the results (in either migration or other behavioral responses) that states had hoped for, and this presents a perfect opportunity to bring the tax back at a small political cost and a large fiscal benefit. In any case, it appears that the state EIG tax revolution is far from over.

VII. CONCLUDING REMARKS

This research takes a close look at what has happened to state death or EIG taxes over the past 30 years. We note that more than 30 states have eliminated their EIG taxes (instead relying only on the pick-up tax) since 1976 and that there have been many other state policy changes that likewise reduce their burden. Yet this unique arena of potential interstate tax competition with a potentially mobile group of affluent elderly as the prize has gone largely unexplored. We investigate this phenomena by first exploring in a descriptive way the historical patterns of EIG tax elimination and reduction. This analysis reveals the complexity of the EIG itself and the difficulty of measuring efforts to eliminate or reduce it in a specific way. We then perform a spatial empirical analysis of the state's overall reliance on EIG taxes, and in so doing extend the usual definition of competitors beyond geography by using historical information on elderly migration patterns, as presumably states should be watching the policies of states to which they are losing (or from which they are gaining) their tax base--the elderly. Our results provide very strong evidence of interstate tax competition, and the intensity of the competition grows as our migration-based definitions of competitors are used and the time period is limited to an era in which states had maximum ability to change their policies.

Now the movement to reduce or eliminate EIG taxes has reached the federal government as well. Interestingly, the movement at the federal level appears to be inspiring the states to effectively bring back their EIG taxes through decoupling, and those actions have been swift and geographically clustered. The decoupling movement is too recent and ongoing for us to thoroughly examine its causes, but the emerging pattern provides additional, compelling evidence that interstate tax competition is once again playing a role.

What happens in the state EIG tax arena in the coming months and years is quite relevant for researchers interested in either state tax competition or the effects of estate taxation. The current movement to decouple is remarkable as one would expect a movement to increase taxes to be much slower than the usual race to the bottom, yet the opposite is true here. Simply tracing out this movement to its logical conclusion will likely yield important insights into the nature of state policy competition and the politics of increasing taxes. As the federal estate tax is gradually eliminated, state EIG taxes will be the only remaining taxes on bequests. Looking backward, one can see that there has been a great many changes to these taxes. These changes provide a multitude of natural experiments with which researchers can study the consequences of EIG taxation on individual behaviors and state budgets and thereby draw inferences about optimal EIG tax policy. Looking forward, one sees a state tax policy that is still very much in flux and still apparently subject to the forces of state policy competition.

TABLE 1
Chronology of State Actions to Eliminate Their EIG Taxes

Year State Date (If Tied)

1976 New Mexico
1977 Utah
1978
1979 North Dakota
1980 Arizona, Colorado, Vermont, Virginia 1 January
1981 Missouri
1982 Washington 1 January
 California 1 July
1983 Illinois, Wyoming 1 January
 Texas 1 September
1984
1985 West Virginia
1986 Minnesota 1 January
 Maine 1 July
1987 Oregon
1988 Idaho
1989
1990
1991 Rhode Island
1992 South Carolina, Wisconsin 1 January
1993 Michigan
1994
1995
1996
1997 Massachusetts
1998 Kansas
1999 Delaware, North Carolina 1 January
2000 Mississippi 1 January
 New York 31 January
2001 Montana 1 January
 South Dakota 1 July
2002
2003 New Hampshire

States with Elimination Upcoming: Lousiana (2004),
 Connecticut (2005)
States that Still Have EIG Taxes: Indiana, Iowa, Kentucky,
 Maryland, Nebraska, New
 Jersey, Ohio, Oklahoma,
 Pennsylvania, Tennessee
States that Eliminated Prior to 1976: Alabama, Arkansas,
 Florida, Georgia, Nevada

Notes: Effective June 2002, Kansas reenacted a succession tax, which is
an inheritance tax exempting spouses, siblings, lineal ancestors, and
lineal descendants.

TABLE 2
Spatial 2SLS Results for State EIG Tax Reliance

 (1) (2) (3)
Weight Measure Contiguity Population Flow

Independent variable

Competitors' EIG tax 0.28936 *** 0.22137 *** 0.42688 ***
reliance (0.01615) (0.04837) (0.07032)
Last year's measure of 0.00037 0.00075 0.00217
sales taxation (0.00400) (0.00420) (0.00392)
Last year's measure of -0.01535 *** -0.01661 *** -0.01743 ***
personal income taxation (0.00429) (0.00431) (0.00400)
Percent of population 85 0.40586 ** 0.40288 ** 0.39632 **
years or older (0.18207) (0.19311) (0.18894)
Percent of population 65 0.07495 ** 0.07535 ** 0.06799 *
years or older (0.03457) (0.03376) (0.03603)
State per capita income -0.00002 -0.00002 -0.00003
(in 1000s) (0.00017) (0.00024) (0.00021)
State per capita federal -0.00320 -0.00330 -0.00304
transfers (0.00258) (0.00277) (0.00272)
State per capita debt -0.00016 -0.00028 -0.00033
 (0.00056) (0.00063) (0.00065)
State unemployment rate 0.00049 *** 0.00050 *** 0.00052 ***
 (0.00011) (0.00011) (0.00011)
Same party-Democrat -0.00008 -0.00010 -0.00007
 (0.00037) (0.00037) (0.00038)
Same party-Republican 0.00006 0.00005 0.00018
 (0.00048) (0.00053) (0.00052)
Election year dummy -0.00055 -0.00052 -0.00052
(Yes = 1) (0.00033) (0.00033) (0.00033)
Constant -0.00473 -0.00283 -0.00610
 (0.01088) (0.01229) (0.01214)

 (4)
Weight Measure Corridor

Independent variable

Competitors' EIG tax 0.38913 ***
reliance (0.08743)
Last year's measure of 0.00109
sales taxation (0.00410)
Last year's measure of -0.01897 ***
personal income taxation (0.00446)
Percent of population 85 0.39058 **
years or older (0.18583)
Percent of population 65 0.06623 *
years or older (0.03662)
State per capita income -0.00003
(in 1000s) (0.00027)
State per capita federal -0.00345
transfers (0.00284)
State per capita debt -0.00035
 (0.00069)
State unemployment rate 0.00055 ***
 (0.00012)
Same party-Democrat -0.00014
 (0.00041)
Same party-Republican 0.00326
 (0.00048)
Election year dummy -0.00053
(Yes = 1) (0.00034)
Constant -0.00504
 (0.01225)

Notes: All regressions included state and year fixed effects.
Bootstrapped standard errors in parentheses. *** significant at the
99% level, ** significant at the 95% level, and * significant at the
90% level.

TABLE 3
Sensitivity Checks on the Coefficient for Competitors' EIG Tax Reliance

 (1) (2) (3)
Weight Measure Contiguity Population Flow

Baseline model--tax 0.28936 *** 0.22137 *** 0.42688 ***
share (Table 2) (0.01615) (0.04837) (0.07032)
Without alternative 0.28357 *** 0.19331 *** 0.35814 ***
tax measures (0.06095) (0.04978) (0.07563)

Changes in measuring
dependent variable

Per capita EIG tax 0.40168 *** 0.40697 *** 0.53736 ***
revenue (0.13266) (0.09947) (0.13424)
Per death EIG tax 0.38120 *** 0.40427 *** 0.53872 ***
revenue (0.14056) (0.11029) (0.14881)
EIG revenues/total 0.28530 * 0.31532 ** 0.61368 **
estate value (0.17465) (0.13782) (0.26588)
3-year average of EIG 0.33053 *** 0.23850 *** 0.46353 ***
revenues (0.06246) (0.04440) (0.06255)

Changes in time period

Just 1970s 0.46708 *** 0.42693 *** 0.90656 ***
 (0.09616) (0.09759) (0.16749)
Just 1980s 0.23942 * 0.22046 *** 0.31516 ***
 (0.12770) (0.07506) (0.10975)
Just 1990S -0.09770 -0.27369 ** -0.65466 ***
 (0.09991) (0.11323) (0.23625)

Treating pick-up states
differently

Competitors' effect 0.37698 ** 0.26909 *** 0.43395 ***
 (0.05637) (0.04728) (0.07000)
Competitors' effect -0.35977 *** -0.40514 *** -0.37274 ***
interacted with pick-up (0.02894) (0.02989) (0.02898)
dummy
Impose zero restriction 0.36409 *** 0.35322 *** 0.38409 ***
on pick-up states (0.02858) (0.02856) (0.02847)

Using random 0.02199
(alphabetical) weights (0.03234)

 (4)
Weight Measure Corridor

Baseline model--tax 0.38913 ***
share (Table 2) (0.08743)
Without alternative 0.26349 ***
tax measures (0.08158)

Changes in measuring
dependent variable

Per capita EIG tax 0.63528 ***
revenue (0.14618)
Per death EIG tax 0.63825 ***
revenue (0.17937)
EIG revenues/total 0.50566 *
estate value (0.27013)
3-year average of EIG 0.45245 ***
revenues (0.07727)

Changes in time period

Just 1970s 0.92801 ***
 (0.20476)
Just 1980s 0.37085 ***
 (0.12437)
Just 1990S -0.80789 ***
 (0.28767)

Treating pick-up states
differently

Competitors' effect 0.39871 ***
 (0.08614)
Competitors' effect -0.39857 ***
interacted with pick-up (0.03049)
dummy
Impose zero restriction 0.39859 ***
on pick-up states (0.02849)

Note: All regressions included state and year fixed effects.
Bootstrapped standard errors in parentheses. *** significant at the
99% level, ** significant at the 95% level, and * significant at the
90% level.

ABBREVIATIONS

EIG: Estate, Inheritance, and Gift OLS: Ordinary Least Squares PUMS: Public Use Microdata Samples TRA: Tax Reform Act

* We thank Deniz Arslan, Stephen Fink, and Liping Zheng for their able research assistance; the numerous state officials who helped provide the necessary data; Meg Stewart for help creating our maps; and Jan Brueckner, Jim Hines, Robert Mohr, Jon Skinner, and two anonymous referees for helpful comments. This research has also benefited from the comments received at the National Tax Association Meetings, the Winter Meetings of the Econometric Society, Eastern Economic Association Meetings, and the UNH Economics Seminar.

(1.) See the work of Gale and Slemrod (2000), Poterba (2001), and the articles in Gale et al. (2001).

(2.) The pick-up tax, sometimes referred to as the soak-up tax or gap tax, arises from federal tax law whereby state EIG taxes are credited dollar for dollar against the federal estate tax liability owed, up to a certain amount. The state therefore picks up some of the federal estate tax revenue without increasing the total tax liability. All states take advantage of this tax credit. We use the term EIG taxes throughout to denote any estate, inheritance, or gift tax rather than using the cumbersome label "estate, inheritance, and gift" or politically charged "death". Five states eliminated their EIG taxes prior to 1960 or never had them.

3. Rather, most emphasize only the federal estate tax. Four notable exceptions are Page (2003), Bakija et al. (2003), and Joulfaian (2000a,b), which we discuss shortly. Only Bakija et al. (2003) take advantage of the temporal variation in state EIG tax policy.

(4.) Beyond taxation, neighboring jurisdictions have also been found to influence the adoptions of lotteries (Alm et al. 1993), tax abatements (Anderson and Wassmer 1995), growth controls (Brueckner 1998), and environmental regulations (Fredriksson and Millimet 2002).

(5.) For further discussion of the history and specifics of state death taxes, see Eckl (1986).

(6.) It is interesting to note that federal estate/inheritance taxes had been enacted and subsequently repealed several times prior to 1916, often to support a war. See Gale and Slemrod (2000) and Joulfaian (2000c) for a brief history and thorough discussion of the workings of the federal estate tax.

(7.) There are obvious limitations to this measure. A preferable measure might be the date that the law was proposed or was voted on. However, the fact that laws must pass various legislative bodies and the governor makes this difficult. In addition, the fact that many states have phased their laws in slowly suggests that the effective date may have been an important part of the decision.

8. We thank Jim Hines for first pointing out how strong this relationship is.

(9.) Note that we are limiting our discussion--and our analyses to the continental48 states. It is a common practice to exclude Alaska and Hawaii from studies of state policies because they have no geographic neighbors and are different in many respects from the other 48 states.

(10.) Two types of tables are often reported in these documents. One summarizes major state tax changes in 19--, and the other reports the exemptions, credits, and rates for state estate and inheritance taxes and includes notes that document changes (and the dates) to these policies. This information is not reported in a consistent way every year during the period, so our summary of it should not be viewed as a complete accounting of events. In addition, the reports end in 1994 because of the elimination of the Advisory Commission on Intergovernmental Relations (ACIR).

(11.) This lead to accelerated collections of the estate tax, from 80% to 90% of estimated payments, and the gift tax from no estimated payments to 90% of estimated payments, according to Significant Features of Fiscal Federalism, volume 1, 1992, Table 17.

(12.) In chronological order, Connecticut placed a 10% surcharge on inheritance taxes in 1983, followed by New York, who allowed unlimited deductions for spouses. In 1985, Connecticut raised inheritance tax exemptions, and New Jersey also reduced its inheritance tax. In 1986 Connecticut began phasing out its inheritance tax on spouses. In 1987, Rhode Island began phasing out its estate tax altogether, and New Jersey exempted spouses effective 1 July 1988. In 1990, Connecticut, New York, and Rhode Island all accelerated the collection of tax payments, and New York expanded the tax base to include out-of state property. For North Carolina, South Carolina, and Tennessee, for example, in 1987 alone South Carolina began phasing out the estate tax (completed by 7/1/91), North Carolina increased its inheritance tax credits, and Tennessee increased exemptions for two types of beneficiaries.

(13.) Population breakdowns by state for the 85+ category were first reported in 1980. For years prior, the number was estimated based on (1) the difference in overall population growth between the state and the nation, (2) the difference in population growth for the 65+ group between the state and the nation, and (3) the difference between the growth rate of each state for 1980-99 in the 65+ and 85+ categories. The exact algorithm is available on request.

(14.) Note that an alternative approach is to invert the system and estimate via maximum likelihood. However, should spatial error dependence also be present, maximum likelihood may provide false evidence of strategic interaction. (see Brueckner 2003 for a discussion). Although results remain unchanged using maximum likelihood, in the article we present the instrumental variable technique because it has been shown (Kelejian and Prucha 1998) to yield consistent estimates, even in the presence of spatial error dependence.

(15.) In creating the flow and corridor weights, we also tried limiting ourselves to high-income elderly migrants (those with annual income of more than $25,000 in 1970) but found that it made little difference in the results.

(16.) See Conway and Houtenville (2003) for a discussion of the trends they uncover contrasted with earlier studies, and Rogers and Raymer (1999) for a descriptive analysis of the internal migration behavior of the elderly since 1950.

(17.) For example, two states could have the same per capita or per death state EIG tax revenues, but EIG taxes could be a much smaller portion of the overall budget in one than the other if the size of their public sectors differed. We believe that a taxpayer would treat the two states differently for this reason. The per capita measure is also somewhat confounded by swings in other population groups (recall that we are controlling for the older age groups) because it is state tax revenues divided by total population.

(18.) We define federal transfers and state debt in per capita terms as opposed to their revenue share because we believe that is a more accurate measure of a state's fiscal climate or income. For example, a state that receives $500 per capita is being subsidized much more heavily by the federal government and therefore should face fewer fiscal pressures than a state receiving $250 per capita. The same is not necessarily true for a comparison of tax shares because the variables are confounded by the size of the public sectors; two states could receive the same level of transfers per capita, but the one with the smaller public sector would appear better off if the share were used. A similar logic applies to state debt.

(19.) For example, Rork (2003) found a response rate of 4-6% for cigarette taxation and Revelli (2001) found a response rate of 4-5% for property taxation in the United Kingdom.

(20.) The full results from any of the estimated models discussed in this section are available on request.

(21.) This information can be found in the Compendium of Federal Estate Tax and Personal Wealth Studies, along with various issues of the Statistics of Income Bulletin, published by the IRS Statistics of Income.

(22.) An alternative way of using this information is to include it as an explanatory variable. We reestimated our baseline model including the aggregate value of gross estates and the per capita estate value, alternatively. Neither variable is ever statistically significant and the estimated competitors" tax reliance coefficients are quite similar and grow in magnitude, if anything. The only substantial difference was that the effect of having an elderly population was again diminished.

(23.) We calculated these averages two ways. One takes the moving average of the final variable--that is, tax share--whereas the other takes the moving average of the numerator (state death taxes) and of the denominator (total state tax revenues) separately and then calculates the shares. The two methods yielded nearly identical results, so we report the results from the first method, which we find more intuitive.

(24.) The results we report use annual data. However, using the three-year moving averages did not substantively alter our results. Likewise, we split the sample into four periods--1968-75, 1976-83, 1984-91, 1992-99--and then into two 15-year periods and find very similar results.

(25.) The 12 states that took action to decouple (as classified by McNichol 2003) are IL, ME, MD. MA, MN, NE, NJ, NC, PA, RI, VT, and WI. The six states that will remain decoupled unless they take action are KS, NY, OH, OR, VA, and WA, plus the District of Columbia. Oklahoma is the only state that has no need to decouple because it has a separate tax that is designed such that it does not suffer a loss from the federal changes. Kansas is also an unusual case in that after eliminating its inheritance taxes in 1998, it adopted a succession tax effective 6 June 2002. A succession tax differs from an inheritance tax in that it exempts spouses, siblings, lineal ancestors, and lineal descendants from taxation.

(26.) "Residents Face Higher Estate Taxes," Boston Globe, 15 November 2002.

REFERENCES

Aim, J., M. McKee, and M. Skidmore. "Fiscal Pressure, Tax Competition, and the Introduction of State Lotteries." National Tax Journal, 46(4), 1993, 463-76.

Anderson, J., and R. Wassmer. "The Decision to 'Bid for Business': Municipal Behavior in Granting Property Tax Abatements." Regional Science and Urban Economics, 25(6), 1995, 727-38.

Auten, G., and D. Joulfaian. "Charitable Contributions and Intergenerational Transfers." Journal of Public Economics, 59(1), 1996, 55-68.

--. "Bequest Taxes and Capital Gains Realizations." Journal of Public Economics, 81(2), 2001, 213-29.

Baicker, K. "The Spillover Effects of State Spending." National Bureau of Economic Research Working Paper no. 8383, Cambridge, MA, 2001.

Bakija, J., and J. Slemrod. "Evidence on the Impact of Progressive State Taxes on the Locations and Estates of the Rich." Mimeo, 2002.

Bakija, J., W. Gale, and J. Slemrod. "Charitable Bequests and Taxes on Inheritances and Estates: Aggregate Evidence from across States and Time." American Economic Review Papers and Proceedings, 93(2), 2003, 366-70.

Besley, T., and A. Case. "Incumbent Behavior: Vote-Seeking, Tax-Setting, and Yardstick Competition." American Economic Review, 85(1), 1995, 25-45.

Besley, T., and H. Rosen. "Vertical Externalities in Tax Setting: Evidence from Gasoline and Cigarettes." Journal of Public Economics, 70(3), 1998, 383-98.

Blumkin, T., and E. Sadka. "Estate Taxation with Intended and Accidental Bequests." Journal of Public Economics, forthcoming.

Brueckner, J. "Testing for Strategic Interaction among Local Governments: The Case of Growth Controls." Journal of Urban Economics, 44(3), 1998, 438 67.

--. "Strategic Interaction among Governments: An Overview of Empirical Studies." International Regional Science Review, 26(2), 2003, 175-88.

Brueckner, J., and L. Saavedra. "Do Local Governments Engage in Strategic Property-Tax Competition?" National Tax Journal, 54(2), 2001, 203-29.

Buettner, T. "Local Business Taxation and Competition for Capital: The Choice of the Tax Rate." Regional Science and Urban Economics, 31 (2-3), 2001, 215-45.

Case, A. "Interstate Tax Competition after TRA86." Journal of Policy Analysis and Management, 12(1), 1993, 136-48.

Case, A., J. Hines Jr., and H. Rosen. "Budget Spillovers and Fiscal Policy Interdependence." Journal of Public Economics, 52(3), 1993, 285-307.

Cebula, R. "A Brief Empirical Note on the Tiebout Hypothesis and State Income Tax Policies." Public Choice, 67(1), 1990, 87-89.

Conway, K. S., and A. Houtenville. "Elderly Migration and State Fiscal Policy: Evidence from the 1990 Census Migration Flows." National Tax Journal, 54(1), 2001, 103-23.

--. "Out with the Old, In with the Old: A Closer Look at Younger versus Older Elderly Migration." Social Science Quarterly, 84(2), 2003, 309-28.

Dresher, K. "Local Public Finance and the Residential Location Decisions of the Elderly." Ph.D. diss., University of Wisconsin, 1994.

Eckl, C. "Death Taxes," in Reforming State Tax Systems, edited by S. Gold. Denver, CO: National Conference of State Legislatures, 1986, 291-306.

Fredriksson, P., and D. Millimet. "Strategic Interaction and the Determination of Environmental Policy across U.S. States." Journal of Urban Economics, 51(1), 2002, 101-22.

Gale, W., and J. Slemrod. "Life and Death Questions about the Estate and Gift Tax." National Tax Journal, 53(4,1), 2000, 889-912.

Gale, W., J. Hines Jr., and J. Slemrod. Rethinking Estate and Gift Taxation. Washington: Brookings Institute, 2001.

Goodspeed, T. "Tax Structure in a Federation." Journal of Public Economics, 75(3), 2000, 493-506.

--. "Tax Competition and Tax Structure in Open Federal Economies: Evidence from OECD Countries with Implications for the European Union." European Economic Review, 46(2), 2002, 357-74.

Heyndels, B., and J. Vuchelen. "Tax Mimicking among Belgian Municipalities." National Tax Journal, 51(1), 1998, 89-101.

Joulfaian, D. "Choosing between Gifts and Bequests: How Taxes Affect the Timing of Wealth Transfers." Office of Tax Analysis Paper #86, 2000a.

--. "Estate Taxes and Charitable Bequests by the Wealthy." National Bureau of Economic Research Working Paper no. 7663, Cambridge, MA, 2000b.

--. "A Quarter Century of Estate Tax Reforms." National Tax Journal, 53(3-1), 2000c, 343-60.

Kelejian, H., and 1. Prucha. "A Generalized Spatial Two-Stage Least Squares Procedure for Estimating a Spatial Autoregressive Model with Autoregressive Disturbances." Journal of Real Estate Finance and Economies, 17(1), 1998, 99-121.

Kuehlwein, M. "The Non-Equalization of True Gift and Estate Tax Rates." Journal of Public Economics, 53(2), 1994, 319-23.

Longino, C., and W. Crown. "The Migration of Old Money." American Demographics, 11(10), 1989, 28-31.

McGarry, K. "The Cost of Equality: Unequal Bequests and Tax Avoidance." Journal of Public Economics, 79(1), 2001, 179-204.

McNichol, E. "Many States Are Decoupling from the Federal Estate Tax Cut." Center on Budget and Policy Priorities, revised 7 August 2003.

McNichol, E., I. Lav, and D. Tenny. "States Can Retain their Estate Taxes even as the Federal Estate Tax Is Phased Out." Center on Budget and Policy Priorities, revised 4 February 2003.

Page, B. "Bequest Taxes, Inter Vivos Gifts and the Bequest Motive." Journal of Public Economies, 87(5-6), 2003, 1219-29.

Poterba, J. "Estate and Gift Taxes and Incentives for Inter Vivos Giving in the US." Journal of Public Economics, 79(1), 2001, 237-64.

Revelli, F. "Spatial Patterns in Local Taxation: Tax Mimicking or Error Mimicking?" Applied Economies, 33(9), 2001, 1101-7.

Rogers, A., and J. Raymer. "The Regional Demographics of the Elderly Foreign-Born and Native-Born Populations in the United States Since 1950." Research on Aging, 21(1), 1999, 3-35.

Rork, J. "Coveting Thy Neighbors' Taxation." National Tax Journal, 56(4), 2003, 775-88.

Sastry, M. L. "Estimating the Economic Impacts of Elderly Migration: An Input-Output Analysis." Growth and Change, 23(1), 1992, 54-79.

Serow, W. "Unanswered Questions and New Directions in Research on Elderly Migration: Economic and Demographic Perspectives." Journal of Aging and Social Policy, 4(3-4), 1992, 73-89.

Voss, P. R., R. J. Gunderson, and R. Manchin, "Death Taxes and Elderly Interstate Migration." Research on Aging, 10(3), 1988, 420-50.

Wilson, J. "Theories of Tax Competition." National Tax Journal, 52(2), 1999, 269-304.

KAREN SMITH CONWAY and JONATHAN C. RORK*

Conway: Professor, Department of Economics, University of New Hampshire, Durham, NH 03824. Phone 1-603-862-3386, Fax 1-603-862-3383, E-mail ksconway@cisunix.unh.edu

Rork: Assistant Professor, Department of Economics, Vassar College, Poughkeepsie, NY 12604. Phone 1-845-437-7398, Fax: 1-845-437-7576, E-mail: rork@vassar.edu