What Goes Up May Not Go Down
State Medicaid Decisions in Times of Plenty

Papers and Studies

Download file This paper is also available in Adobe Acrobat PDF format.

1. Introduction

A debilitating fiscal crunch has forced many states to consider painful budget cuts. Rapidly growing Medicaid expenditures, which now constitute over twenty percent of state budgets, are a a particularly urgent item for state legislatures. Many states are considering cutbacks; all demand increased federal funding.

State officials attribute escalating Medicaid expenditures to factors ostensibly beyond their control--federal “mandates,” “exploding health care costs” (especially pharmaceuticals) that have hit the states like a force of nature, and (in consequence of the lackluster economy since 2001) rising levels of poverty. Even at first glance, though, the states’ representations are open to considerable doubt. For example, while most Americans still think of Medicaid as a “mandatory” federal safety net for poor families with children, two-thirds of Medicaid spending is now devoted to constituencies and services that the states may, but need not, cover as a condition of receiving federal Medicaid reimbursements. Similarly, Medicaid expenditures rose throughout the 1990s, an economic boom period when poverty rates fell substantially. Thus, neither mandates nor poverty seems likely to explain the Medicaid cost explosion. (As further discussed below, “health care inflation” fares little better.)

The AEI Federalism Project has conducted an analysis of state Medicaid decisions and spending between 1994 and 2000. Our study shows an enormous divergence in Medicaid spending trends among the individual states. That observation strongly suggests that Medicaid’s seemingly irresistible growth is largely attributable to the states’ policy choices, rather than exogenous factors that affect all states. While our study permits no definitive conclusions about the factors that generate particular state Medicaid choices, state revenue growth appears to be a driving force behind Medicaid growth.

2. Medicaid Meets the Boom

We examined state Medicaid spending from 1994 to 2000 inclusive. The time span covers all but the first and last fiscal years of an unprecedented boom in state finances. FYs 1993 and 2001 were omitted because state-by-state data were unavailable.

State by state, we examined the change in the following variables: state revenues; Medicaid spending (state, federal, and combined); the federal matching ratio (“FMAP”) that determines the federal reimbursement of state Medicaid expenditures; the proportion of state funds dedicated to Medicaid; and changes in poverty. Revenue and spending figures were adjusted for inflation, using the general CPI deflator. Since we are primarily interested in state-to-state differences rather than aggregate Medicaid trends, the national averages for all states are not weighted by state size.

State-by-state data are shown in Appendix A, along with the sources from which we derived the data. Table (1) below shows the 1994–2000 pattern in comparison to GDP growth. State revenues grew faster than GDP. At the same time, poverty dropped sharply. Even so, Medicaid spending increased at a faster rate than either GDP or state revenues.

Table 1: Real Change (in percent), 1994–2000

Per Capita
State revenues (Mean)
Poverty Population
Poverty Rate
Medicaid Expenditures (State and Federal)

Source (GDP): Statistical Abstract of the United States: 2002, p. 417, Table No. 631. All other sources shown in Appendix A.

State-by-state data further illustrate the pattern:

  • All states except Hawaii (where revenues were flat) experienced substantial real revenue growth. In thirty-one states, real revenue growth exceeded real GDP growth for the period.
  • The economic boom, coupled with the effects of the 1996 welfare reform, sharply reduced poverty rates. In eleven states, the rate dropped by over 30 percent; another fourteen experienced a reduction of 20 percent or more. Only eleven states suffered a poverty rate increase. In 35 states, the absolute number of persons in poverty declined over the period.

Also reflected in the data is the dramatic growth of Medicaid expenditures in virtually all states:

  • Between 1994 and 2000, only two states registered an unambiguous decline in Medicaid expenditures, and only one of the two (New Hampshire) can lay claim to having realized credible spending reductions commensurate with declining poverty.
  • Despite booming revenues, Medicaid expenses consumed a slightly larger portion of state budgets at the end of the period than at the beginning. Medicaid’s “take” of the budget rose in 25 states but fell in only nine. (For further analysis see below.)

How did this happen? The conventional explanation, as noted, attributes Medicaid’s stupendous growth phenomenon to exogenous factors.

One of the proffered factors, federal mandates, can safely be excluded. During the period under consideration, the federal government imposed no significant new Medicaid mandate. Rather, the 1994–2000 period was characterized by an increasingly aggressive use of statutory waivers that are calculated to enhance state flexibility and autonomy. To the extent that Medicaid spending growth is attributable to government decisions, those decisions are almost exclusively the states’.

Other exogenous explanations advert to a supposed “medical cost inflation” and poverty. We consider--and exclude--those explanations in turn.

3. Health Care “Inflation”?

The “rising costs of health care” are variously attributed to changing demographics, medical progress, and the pharmaceutical industry’s monopolistic pricing practices. Several considerations, however, render those explanations suspect.

First, Medicaid spending is largely a function of government decisions concerning benefits and eligibility. To the extent that governments expand benefits, “rising health care costs” is not an explanation of Medicaid spending growth but a tautology. The figures displayed in Table (2) below illustrate the significance of this point.

As shown, aggregate Medicaid expenditures rose much faster during the 1994–2000 period than either national Medicare expenditures (which are entirely federal) or national health care spending, exclusive of Medicaid. (GDP growth figures are shown for comparison.) Expenditures per Medicaid recipient, however, rose more slowly than per-recipient expenditures under Medicare and only half as fast as national per-capita health care expenditures (excluding Medicaid). This observation is highly consistent with the hypothesis that Medicaid growth has been driven by factors other than, or in addition to, uncontrollable “health care inflation.”

Table 2: Real Growth, 1994–2000

Per Capita
Medicare Spending
National Health Care Spending
(excluding Medicaid)

Sources: GDP: Statistical Abstract of the United States: 2002, p. 417, Table No. 631. All other sources shown in Appendix A.

Second, while “health care inflation” explains some portion of the aggregate spending trend, it cannot explain the astounding differences among the states. By way of impressionistic examples: among the states where poverty rates fell by over 30 percent, New Hampshire’s Medicaid expenditures dropped by 25 percent. In Minnesota, where the poverty rate was cut almost in half, Medicaid expenditures increased by 24 percent. Similarly, among the states where poverty rose, Montana (36.5 percent poverty growth) managed to keep Medicaid growth at 9.3 percent, while Vermont (48.7 percent poverty growth) boosted Medicaid expenditures by 46.3 percent.

Variations of such magnitude cannot be ascribed to differing demographic changes (see below), the dispersion of expensive medical technologies (as if all the innovations should occur in Minnesota), or a pharmaceutical industry conspiracy against Vermont (but not Montana). The differences have to do with the states’ policy choices.

4. Poverty?

As noted, Medicaid is increasingly serving constituencies that are not poor. Conversely, not every poor person is eligible for Medicaid (in all states), and not all those who are eligible do in fact enroll. Changing utilization rates or eligibility rules may (and do) have powerful financial effects even if poverty remains constant.

That said, poverty remains an important factor affecting Medicaid spending--but not in the sense that is commonly suggested. All else being equal, higher poverty levels mean higher Medicaid spending. All else, however, was not equal during the boom years. Poverty fell dramatically in most states; and yet, Medicaid spending failed to decline.

Figure (1) plots the 1994–2000 change in state and federal Medicaid expenditures (y-axis) against the change in the poverty rate for that period (x-axis). Predictably, higher levels of poverty are associated with higher Medicaid expenditures. The regression line represents the correlation (corr.= .380).

Download file Figure 1

To account for the impact of demographic changes (including migration), we also correlated the 1994–2000 in the poverty population (not the rate) with the change in per capita Medicaid spending. The result is shown in Figure (2). The correlation is somewhat weaker (.310). In a substantial number of states, per capita spending growth was (predictably) lower than aggregate spending growth. Some states shift along both axes, and a few do so in significant and interesting ways. The general pattern, however, remains substantially unchanged. Note the following:

Download file Figure 2

  • The dispersion along both axes is enormous. Change in the poverty rate ranges from -48.7 percent (Minnesota) to +46.4 percent (Vermont). Medicaid expenditure changes ranged from 20-percent-plus declines (New Hampshire, Louisiana) to around 60 percent growth in five states. In other words, states differ dramatically in their “poverty experience”--and in their policy responses.
  • The high intercept on the y-axis indicates a strong underlying expenditure increase, irrespective of variations in poverty rates or populations. Rising poverty levels are associated with higher Medicaid expenditures. But then, so are falling poverty levels, as reflected in the startling concentration of states in quadrant (I).

5. Substitution

Those quadrant (I) denizens are the interesting cases: states where lower poverty levels failed to translate into reduced Medicaid expenditures. The boxplot Figure (3) below plots four categories of states, grouped in accordance with their change in the poverty population, against the change in state Medicaid expenditures within each group. The shaded box shows the range of the middle fifty percent of states in each group; the line inside the box, the median; and the thin lines outside the box, the entire range. Outliers and extremes (Louisiana and New Hampshire, respectively) are shown separately.

Download file Figure 3

Groups (III) (modestly falling poverty) and (IV) (rising poverty) show widely differing policy responses—or perhaps (since Medicaid eligibility decisions are made with less than perfect knowledge about the state’s economic future), widely differing degrees of realism about poverty trends. The ranges and medians, however, behave vis-à-vis one another as one would expect: all are higher for the poverty-increasing group (IV) than for Group (III).

Groups (I) and (II), with sharp drops in poverty, do not behave as expected. The range of Medicaid growth is much tighter for both groups--and, with the noted exceptions of Louisiana and New Hampshire, there is nothing but growth. Moving from right to left, moreover, the median growth rate refuses to budge as we go from (III) to (II)--and then rises for (I). Likewise, the lower end of the range ascends from (II) to (I). Loose translation: between 1994 and 2000, state Medicaid expenditure growth hit a “bottom” (of around 20 percent). It refused to drop below that bottom no matter what happened to the poverty population and in fact rose in states where that population decreased most sharply.

A few modest assumptions generate an exceedingly plausible story to explain the distribution just described. Suppose that state politicians by and large have a reasonably good sense of their state’s economic future (poverty rates, revenues, etc.). Further assume that they view Medicaid not as a countercyclical poverty program, but as a fixed budget item with some built-in growth rate. In that frame of mind, politicians will tend to “substitute” eligibility, benefit, and coverage expansions for declining poverty populations. In so doing, though, they will try to stay within an “affordable” Medicaid growth range. This substitution mechanism would explain both the Medicaid growth “floor” and the comparatively narrow range of the Medicaid growth in Groups (I) and (II).

This account, to repeat, is a plausible story, not a tested (let alone proven) explanation. It is, however, far more plausible than a naked “health care inflation” tale, and it is consistent with the evidence further sketched below.

6. An Excursion into Red America

A casual inspection of the Group (I) states--where poverty fell most precipitously--reveals an intriguing fact: with the lone exception of Connecticut, states in this group are geographically cohesive. They form a corridor from Mississippi and Louisiana northward through Kentucky, Missouri, and Kansas and on to Indiana, Iowa, Minnesota, and South Dakota. America’s Heartland states experienced particularly dramatic poverty declines--and Medicaid expenditure increases above those experienced by less fortunate (or policy-wise) states.

Table 3 below shows several indicators for the eight individual states in this group (exempting Connecticut and Louisiana, a statistical and real-world outlier). On average, Medicaid expenditures for these states grew at about the national average rate, despite their exceptional poverty “losses.” Revenues, in contrast, grew at a somewhat slower rate than the national state average (21 percent versus 24.4 percent). The growth of Medicaid as a percent of state budgets exceeds that of the average state (8.0 percent versus 5.1 percent).

Table 3: State Revenues and Medicaid Spending

Revenue Growth
State Medicaid
State Medicaird as %
Expenditure Growth
of Budget-Change
South Dakota
National Mean
National Median

Sources: State Revenues: Census Bureau, “State Government Finances,” available at http://www.census.gov/govs/www/index.html. Medicaid Expenditures: Centers for Medicare and Medicaid Services, “Net Reported Medicaid and SCHIP Expenditures,” (1997–2000) available at http://cms.hhs.gov/medicaid/mbes/sttotal.pdf.

Perhaps, the “corridor” states--many of them accustomed to persistent poverty in their midst--responded to their sudden good fortune by increasing utilization rates or by extending, with particular aggressiveness, Medicaid benefits to constituencies that had long been declared eligible in richer sister states. Equally plausible is the interpretation that the Group (I) states are actually pretty average citizens, Medicaid-wise; they are distinguished--and lumped together--simply by their successful anti-poverty policies. The simple fact, though, is this: the states whose poverty declines should have created room for Medicaid expenditure reductions—at least relative to less fortunate sister states—by and large failed to avail themselves of that option.

7. Revenues

Having argued that the conventional explanations for Medicaid expenditure growth fare poorly (at least for the period under consideration), we are left with an obvious question: what does explain that growth?

Among the plausible candidates is state revenue growth. As revenues grow, states feel free to increase Medicaid expenditures (along with expenditures on everything else)--by expanding eligibility, adding benefits, or increasing utilization (or some combination thereof).

At first inspection, state revenue growth appears only weakly correlated with Medicaid expenditure growth. More careful inspection, however, reveals a very different picture.

We conducted a regression analysis with four independent variables: (1) the change in the federal matching formula (“FMAP Growth”), which affects each state’s Medicaid share; (2) the change in the poverty population over the entire period; (3) change in state revenue; and (4) change in the proportion of state revenues dedicated to Medicaid.

Legislative Medicaid decisions have a certain lag time; their fiscal effects are felt in later years. At the same time, state legislatures make Medicaid choices in anticipation of future economic developments. To mimic these real-world conditions, we considered changes in variables (3) and (4) only for the first three years of the period (that is, FY 1994–1997). The results for state Medicaid expenditure change are shown in Table 4.

Table 4: State Medicaid Growth, 1994–2000

Unstandardized Coefficients
Standardized Coefficients
FMAP Growth
Poverty Growth
Revenue Growth, 1994–97
Medicaid Growth (as % of
state revenue, 1994–97)
Significant F
n = 50 states

We replicated the regression for total (state and federal) Medicaid expenditures, with similar results:

Table 5: Total Medicaid Growth, 1994–2000

Unstandardized Coefficients
Standardized Coefficients
FMAP Growth
Poverty Growth
Revnue Growth, 1994-97
Medicaid Growth (as % of
state revenue, 1994-97)
Significant F
n = 50 states

Both analyses explain a large portion of the variation (two-thirds and over one-half, respectively). The constants are very similar--and high, reflecting the high growth of Medicaid expenditures regardless of the considered variables. The FMAP variable changes from negative to positive--as one would expect: a higher federal reimbursement rate reduces state expenditures (all else being equal) but increases combined federal and state expenditures. That latter influence, however, is not statistically significant.

Unsurprisingly, growth in state Medicaid spending is positively associated with poverty increases. More strongly, however, growth in (state) Medicaid spending is positively associated with early revenue growth and, yet more strongly, with the early growth of state Medicaid expenditures in proportion to state budgets. Each of these factors easily swamps the poverty factor.

Our findings may appear to border on the trivial: when states decide, early on, to devote a larger share of growing resources to Medicaid, Medicaid spending grows. A growing Medicaid budget share, however, may reflect any number of decisions or considerations—for example, a deliberate decision to reduce non-Medicaid budget items (in which case Medicaid expenditures might remain constant), or a decision to increase Medicaid expenditures now in the hope or expectation of reductions down the road.

Similarly, growing revenues typically coincide with improved economic conditions, including declining poverty. If state legislatures thought of Medicaid as a poverty program, there should be no reason for revenue improvements to be associated with higher future Medicaid expenditures; the opposite might be the case. The results strongly support an explanation suggested above: state legislatures tend to sustain and expand Medicaid expenditures at a rate that they deem “affordable,” regardless of changing poverty conditions.

8. Affordable Medicaid Growth?

Health care experts have occasionally suggested that states can and should make Medicaid decisions in accordance with “affordability” criteria. While those criteria are naturally open to debate (“affordable” compared to what? For whom?), even very generous measures demonstrate the hazards of an affordability criterion--or, perhaps, the unwillingness of some states to pay much heed to affordability.

A plausible measure of “affordability” is an implicit judgment that Medicaid expenditures should keep pace with rising state budgets: a rising tide should lift all boats. Upon inspection, that measure (however well-intended) implies an irresponsible assumption of perpetual budget growth. (No one believes that an occasional ebb tide should sink all boats.) Putting that aside, even under the assumption that states wanted to include distressed constituencies in their good fortune, many of them overshot that target.

Figure (4) below plots state revenue growth against the proportion of state expenditures devoted to Medicaid. The states between the dotted lines kept Medicaid growth, as a proportion of their budgets, within a generous 10 percent range of error (such that a 20 percent state Medicaid share in 1994 translated into no more than 22 percent and no less than 18 percent by 2000). One-third of all states, however, overshot and made Medicaid less affordable, relative to (growing!) budgets. Only six states, including the Louisiana aberration and a marginal case (Arkansas), made Medicaid more affordable in the sense explained. Even among the states that grew faster than the national average (indicated by the vertical line), five managed to render Medicaid substantially less affordable to their citizens.

Download file Figure 4

It bears emphasis that this thumbnail analysis gives the states every benefit of the doubt. It implies that the states’ 1994 Medicaid expenditures, which were mostly made in 1992 and 1993, reflect a considered judgment that the then-existing Medicaid commitments were reasonable. In fact, those baselines likely reflect a Medicaid spending bump as a result of the 1991–1992 recession. It likewise implies that states somehow foresaw or anticipated the future revenue growth (such that the Quadrant (I) denizens may have been “done in” by lower-than-expected growth). In fact, they could not, and did not, foresee the unprecedented boom ahead.

9. Conclusion

Between 1994 and 2000, falling poverty was accompanied by escalating Medicaid expenditures. That escalation is largely attributable to state policy choices under loosening budget constraints. While higher poverty levels are associated with higher expenditures all else being equal, state revenue growth is a much stronger predictor of future Medicaid growth.

Medicaid has grown much faster (both during the observation period and in years since) than Medicare, a federal program that does not entail state decisions concerning eligibility or benefit levels. While our analysis does not prove that state decisions account for Medicaid’s growth in excess even of the unsustainable Medicare growth rate, that hypothesis is entirely consistent with our findings.

It is consistent, moreover, with the theoretical expectation that an uncapped federal matching grant--such as Medicaid--will generate asymmetrical state responses to changing budget constraints. In an economic downturn, rising poverty pushes Medicaid spending--state and federal--upward. (State attempts to control costs will fail to offset that effect, both because they are politically difficult and because a $1 Medicaid cut translates at most into a fifty-cent “savings” for the state.) Improved economic fortunes and falling poverty, however, do not produce the same effect in reverse. Under those conditions, state policymakers substitute benefit and eligibility expansions for falling poverty rates, and improved revenues act as an independent spur for program and expenditure expansions. With the lone exception of New Hampshire, no state has unambiguously “pocketed” a poverty windfall; the general tendency is to spend those funds on benefit or eligibility expansions.

Medicaid was designed as a countercyclical program: as economic recessions deprive states of the ability to help the poor, the federal government would step in and provide funding. In actual operation, Medicaid is a spiral program. In bad economic times, it stabilizes state expenditures; in good times--even the best of times--it provides an incentive for program expansions that would prove unsustainable even under the most optimistic state revenue scenarios. With the inevitable bump in the road, the wheels come off.

Appendix A

Download file 1994-2000 Changes in State Revenues, Medicaid Spending, and Poverty


Medicaid (Total and Federal shares by state)
Centers for Medicare and Medicaid Services, “Net Reported Medicaid and SCHIP Expenditures,” (1997–2000) available at http://cms.hhs.gov/medicaid/mbes/sttotal.pdf. CMS recently removed the 1994–1996 data, which had been available at http://cms.hhs.gov/medicaid/mbes/m64.asp.

National Health Care Costs (Medicare, Medicaid, and other expenditures)
National Health Expenditures 1965–2012, available at http://www.cms.hhs.gov/statistics/nhe/default.asp (as nhe65-12.zip).

Census Bureau and the Bureau of Labor Statistics, Annual Demographic Survey, Table 25 (1994–2000) available at http://www.census.gov/hhes/poverty/prevdetailtabs.html.

Census Bureau and the Bureau of Labor Statistics, Annual Demographic Survey, Table 25 (1994–2000) available at http://www.census.gov/hhes/poverty/prevdetailtabs.html.

State Revenues
Census Bureau, “State Government Finances,” available at http://www.census.gov/govs/www/index.html.

Michael S. Greve is AEI’s John G. Searle Scholar and director of its Federalism Project. Jinney Smith is a doctoral candidate in Political Science, Northwestern University. We are indebted to Kate Crawford for tireless research assistance and to Joseph Antos for helpful comments.

Also Visit
AEIdeas Blog The American Magazine
About the Author


Michael S.

What's new on AEI

AEI Election Watch 2014: What will happen and why it matters
image A nation divided by marriage
image Teaching reform
image Socialist party pushing $20 minimum wage defends $13-an-hour job listing
AEI on Facebook
Events Calendar
  • 20
  • 21
  • 22
  • 23
  • 24
Monday, October 20, 2014 | 2:00 p.m. – 3:30 p.m.
Warfare beneath the waves: The undersea domain in Asia

We welcome you to join us for a panel discussion of the undersea military competition occurring in Asia and what it means for the United States and its allies.

Tuesday, October 21, 2014 | 8:30 a.m. – 10:00 a.m.
AEI Election Watch 2014: What will happen and why it matters

AEI’s Election Watch is back! Please join us for two sessions of the longest-running election program in Washington, DC. 

Wednesday, October 22, 2014 | 1:00 p.m. – 2:30 p.m.
What now for the Common Core?

We welcome you to join us at AEI for a discussion of what’s next for the Common Core.

Thursday, October 23, 2014 | 10:00 a.m. – 11:00 a.m.
Brazil’s presidential election: Real challenges, real choices

Please join AEI for a discussion examining each candidate’s platform and prospects for victory and the impact that a possible shift toward free-market policies in Brazil might have on South America as a whole.

No events scheduled this day.
No events scheduled today.
No events scheduled this day.
No events scheduled this day.