February 1996 € Volume 23 € No. 1

Why Are Public Budgets Such a Mess?

While Congress and President Clinton mud wrestle in Washington, the author outlines three principles that help explain why budgetary problems are so severe. There is no separating politics from administration, but managers and politicians can still improve things by considering some of the underlying problems that lead to inefficiency.

By Harold Pollack

No issue subjects politicians to greater scorn than their seeming inability to prevent chronic budget deficits and cost growth. As a fraction of gross national product, U.S. public debt is as large in 1995 as it was in 1960 when we were still paying the huge bill required to defeat the Nazis. With the exception of World War II, public debt ratios are higher than at any point since the founding of the republic. Everyone knows that the federal government accumulated large deficits over the last two decades. Many state and local governments have also experienced severe fiscal trauma in recent years. Efforts to confront such deficits have caused almost as much disruption as the debt itself. As this article is written, thousands of federal employees deemed "inessential" remain idle as the two parties bicker in Washington about how to balance the budget by 2002.

Why are public budgets such a mess? Ross Perot and connoisseurs of conspiracy would have us believe that lobbyists and political hacks are to blame for our fiscal woes. Events such as the savings and loan debacle indicate that the charge is not wholly without merit. Yet our analytic training should lead us towards less personal explanations. On the whole, deficits and inefficient public expenditure do not arise from the misconduct of elected politicians, as incompetent and corrupt as these men and women can occasionally be. Such problems recur because of partisan conflict inherent to democratic government, and because it is hard to manage the large, decentralized organizations we need in today's world. We can put our fiscal house in better order, but the first step is to stop blaming the messengers and face up to deeper concerns.

Alas, I can provide no magic bullet to solve our fiscal woes, but the three principles below help to explain why budgetary problems are so severe. 1. Budgeting is most important in the presence of large fixed costs.This is exactly when the problems associated with fixed budgets are most severe. We drum into our students that the efficient allocation of scarce resources requires prices that accurately capture marginal costs. This strategy works great when a large organization wants to limit overuse of the Xerox copier. One doesn't really need budgets if it is possible to capture the true opportunity costs of scarce resources. The municipal sanitation department can charge $40 for bulk trash removal. It can then serve whatever local customers are willing to pay the fee. Such approaches don't work so well in deciding tuition and class size at Yale University, or how many MRI machines to locate in Boston.

The problem here is not that MRI machines are inherently more complicated than Xerox copiers or the typical garbage truck. The problem is not even that patients are over-insured and therefore face strong incentives to over-utilize medical resources. The central quandary is caused by large fixed costs that drive a huge wedge between marginal and average costs. The big expenses for such a technology is on the research, development and production of the machine, not to mention the costs to train men and women to operate the thing. Once the MRI is turned on, it is pretty cheap to scan another high school basketball player who may have ligament damage in his knee.

To be more concrete, suppose that (at reasonable usage) the average cost of an MRI scan is maybe $2,000, but that the true marginal cost is only $500. How many MRI machines should there be, and how much should patients or insurance companies be charged to use them? Who should bear the fixed costs of the machines?

It should be obvious that no one "price" is capable of getting things right. At the same time, hospitals must recover at least their average costs to stay afloat. Yet if they charge that much per visit, they fail to serve many patients who are willing to pay the marginal cost of the scan. This is bad for patients, and it is bad for the hospital itself. As long as there is over-capacity in the system, everyone would benefit if one could sneak a few patients through who are only willing to pay $750. If only hospital administrators could be sure that their more lucrative patients never find out that they are paying more for the same service!

Sometimes one can finesse the problem by offering low-quality services to patients who do not pay the full freight. One might impose annoying delays, or only offer discount services during inconvenient, off-peak times. Some clerk in the waiting room might defend such policies by arguing that discount patients should not be allowed to disrupt services provided to other patients. The real purpose, however, is to make sure that everyone doesn't start demanding to pay the lower rate.

Smart undergraduates might respond to this scenario by saying that the price should be set to $500. An exceptionally naive policy analyst might say, "Of course this system is inefficient. The problem is that you are using price as your only instrument to allocate resources and recover your fixed costs. You should simply charge everyone marginal cost. Let the government or someone else subsidize the fixed costs."

This proposed solution has the virtue of simplicity, but too many MRI machines will be built. To see why, consider what happens if one starts by charging $501 for an MRI, and then one reduces the price to $500. Utilization will slightly rise because one will attract some patients at the margin between getting an MRI or pursuing other treatments. These new patients, however, gain virtually nothing from the discount because they value the MRI at almost exactly the marginal cost. Only now, the government is stuck subsidizing the fixed costs of whatever new machines are required to meet the increased demand.

There is no elegant or easy solution to such concerns. Within health care, public and private organizations have developed health systems agencies and an alphabet soup of commissions, think tanks, and consulting firms to manage chronic over-capacity and rapid technological change. Within any given setting, some combination of subsidy and price discrimination is required to efficiently recover costs. Some form of explicit or implicit rationing is often required when prices fall far below average cost. 2. The size of government is a function of ideology. Large deficits and blatant inefficiency are usually products of political uncertainty or severe partisan disagreement. You might think that budget deficits and blatantly inefficient government spending are the result of ideology. Democrats blame the deficit on the deep tax cuts and bloated military procurement of the 1980s. Republicans cite rising demand for government welfare services and adverse incentives associated with heavy taxes. Neither theory fits the data very well.

A recent survey by economists Alberto Alesina and Roberto Perotti compared the public debt burdens of the advanced industrial democracies. As a fraction of gross national product, Belgium, Ireland, Italy and Greece have all accumulated far larger public debts than the United States. Meanwhile, such diverse nations as Spain and Denmark have accumulated very low levels of public debt.

The best predictors of such debt burdens is not the per-capita incomes or political ideologies of these nations, but the state of their political systems. Precarious coalition governments tend to run large deficits because they are sluggish to enact controversial policies that respond to sudden demands on public resources. Polarized societies run large deficits as an outgrowth of partisan conflict over time. Many Republicans were willing to countenance the large deficits of the Reagan era because they thought this would constrain social expenditures in future Democratic administrations.

Some of the same political dynamics explain pervasive and sometimes wasteful micro-management of public organizations by their elected overseers. Such micro-management is most likely in decisions that politicians would have the greatest difficulty reversing later on. For example, legislators rarely grant agencies broad discretion for long-term capital expenditures that would be difficult to abandon in future years. Intrusive micro-management may also be more common when the legislature is controlled by a slim majority. Politicians will be especially vigilant in exercising their own prerogatives when they know they are especially vulnerable to political gaming by their nominal subordinates. 3. Fixed budgets create changing incentives over the fiscal year. Resulting inefficiencies have proved surprisingly difficult to eliminate. Friedrich Hayek noted long ago that command economies were bound to fail because central planners could not possibly collect and analyze all the information reflected in local market prices. This is no less true of large private sector organizations: top executives inevitably lack critical information held by their subordinates. Hence, they decentralize. (This section draws upon joint research with Richard Zeckhauser, "Budgets as Dynamic Gatekeepers," Management Science, forthcoming.)

While such decentralization is unavoidable, it inevitably brings incentive problems and other sources of agency loss. Economic theorists and others have developed complex, sometimes elegant proposed solutions to these difficulties. In practice, however, fixed budgets predominate because they provide a realistic and workable compromise between the need for decentralized decision-making and centralized control.

Although we usually think about resource allocation as static, budgets are actually determined for use over a longer time. Budgets often create dynamic incentive problems because all or most of an organization's unspent funds are generally lost at the end of the fiscal year.

This is not a serious problem when an organization is large enough so that it faces a small proportional variance in its optimal expenditure due to unexpected contingencies. Fixed budgets work less well when organizations face large or highly correlated risks, or when the fiscal year is too short to smooth out the natural variability due to repeated chance occurrences. When such uncertainties emerge, the timing of decisions within the fiscal year takes on added importance.

Fixed budgets also provide obvious incentives for over-spending near the end of the fiscal year. These adverse incentives are then aggravated by the complicated signaling games that take place between central authorities and their subordinates. Ironically, underspending may be taken as a sign of incompetence, that needed work has not been done. Thus, one official, frustrated that her department had failed to exhaust its full budget authority during the fiscal year, admonished her staff: While I know that it is difficult to spend every last penny and some lapses are always going to occur, the magnitude of these lapses makes me extremely uncomfortable for three reasons:

Some astute budget analyst is going to notice a pattern and reduce our appropriations, figuring that if we don't spend it we must not need it.

Significant underspending makes it difficult to justify any increases in these appropriations.

3. I am concerned that clients needing services are not getting them.
This problem illustrates many strengths and weaknesses of common budgetary schemes. From an analytic perspective, the problem also illustrates how one would model issues of timing and uncertainty within a dynamic setting. I first became interested in these problems through joint work with Richard Zeckhauser concerning the analytics of medical "gatekeeping" -- widely employed administrative schemes that seek to reduce over-use of costly services.

Gatekeeping schemes
The underlying logic of gatekeeping is that physicians should be provided some incentive to reduce inefficient and expensive specialty care. For example, the British National Health Service during the latter Thatcher years became concerned that general practitioners were referring too many patients for costly specialty care. Patients faced virtually no charge and generally derived at least some positive benefit from these referrals. Not surprisingly, patients encountered long waiting times and much inefficient specialty care was dispensed.

One proposed solution was to give practitioners a fixed number of referrals over the entire fiscal year. Central authorities would not second-guess practitioners' decisions in particular cases. However, a practitioner would face some administrative or financial penalty for referring more than her allotted number of patients over the fiscal year. In effect, practitioners were asked to be "gatekeepers," to restrict low-priority uses of scarce medical resources. Similar practices are widespread outside the medical arena. University admission, bank loans and many social services are implicitly or explicitly rationed by low-level decision-makers who enjoy broad discretion as long as they respect overall spending constraints.

This scheme eliminates one source of inefficiency, but it creates distortions of its own. Because of random fluctuations, gatekeepers are uncertain about future needs. Near the beginning of the fiscal year, they will therefore be overly reluctant to refer patients. Because the gatekeeper has no emergency reserve, she fears that she will be unable to serve especially needy patients later on. Then, when the end of the fiscal year arrives, the gatekeeper has strong incentives to refer marginal patients she would have rejected earlier on. (Operations researchers should file this insight for later use. If you are applying for a research grant, apply early if you are a superstar, but don't lose heart if your research is pretty lame! You may still sneak through if the funding agency has a little extra money at the end of the year.)

As this process unfolds, different gatekeepers within the same organization will face dramatically different shadow prices for their remaining referrals. Ideally, one would then reallocate referrals to the gatekeepers serving the most needy patients over the remainder of the fiscal year. Yet, such reallocation is rarely possible within the internal politics of most large organizations. Moreover, the prospect of such reallocation would dilute the incentives for prudence contained in the original scheme. Suppose that a gatekeeper expends 70 percent of her referrals during the first six months and claims to have seen unusually sick patients. If central planners cannot verify the truth of these claims they would be foolish indeed to provide additional resources.

Dynamic optimization model
Technical aficionados might note that the basic features of the problem are easily illustrated using a dynamic optimization model. Suppose that a gatekeeper can provide a total of B referrals over the fiscal year. The fiscal year consists of T periods, t1..tT. At time t, she examines a patient, and either makes a referral (providing some benefit zt) or provides no further care and the patient is never seen again. One might therefore think of zt as a cost-effectiveness measure. We also assume that the collection {zt} are independent, identically distributed draws from some cumulative distribution F(z). There is also no discounting. The gatekeeper's only goal is to maximize the well-being of her patients. She does not worry about the well-being of the taxpayer or the patients of other doctors who might be affected by her referral.

One can then formulate her referrals using a stochastic dynamic program. Suppose at some time t-1 the gatekeeper has b remaining referrals. She will refer as long as the benefit to the patient exceeds some cutoff C(x,t-1). Notice that this cutoff may be interpreted as the shadow price of spending one unit now, rather than waiting to expend optimally later on.

Notice that with probability 1-F(C), a referral is made providing expected benefit E[Z|Z>C] and leaving b-1 units for future spending from t on. On the other hand, with probability F(C) the patient is sent home. So if the gatekeeper seeks to maximize her expected benefit V(b,t-1), one can therefore write the Bellman recursion: V(b,t-1)=[1-F(C)]E[Z|Z>C]+[1-F(C)]V(b-1,t)+F(C)V(b,t) One finds the optimal cutoff by differentiating with respect to C. Some algebra yields the optimal shadow price relation: (C(b,t-1) = V(b,t) - V(b-1,t) So a patient is referred whenever the benefit exceeds the option value of being able to treat an additional patient later in the fiscal year. Notice that at the end of the year, the gatekeeper will seek to use up whatever referrals she has; so C(b,T) = 0. Given this boundary condition, the model is easily calculated using a rudimentary computer program or spreadsheet.

This latter condition reflects the inefficiency inherent to any budget schemes that provide no incentive for restraint near the end of the fiscal year. One obvious response is to lengthen the fiscal year, or to allow some carry-overs of unspent funds into the next fiscal year. Vice President Gore's National Performance Review proposed that federal agencies be allowed to carry over half of their unspent funds into the next fiscal year. One might also allow agencies to "borrow" against future budgets when they encounter exceptional needs.

Given these compelling arguments for such policies, it is surprising that they are rarely employed. One reason is that such strategies provide increased opportunities for wasteful gaming. Lengthening the budget year or otherwise "softening" the budget constraint lessens the credibility of efforts to enforce fiscal discipline. Within government, a frequent legislative complaint is that agencies exploit contingency funds and other off-budget mechanisms to increase spending. Studies of state pending patterns also find that biennial budgets are associated with increased expenditure. Apparently, the increased monitoring and agency costs outweigh the efficiency gains associated with such measures.

Politics and administration
So what answers can we provide to the budget dilemma? Mismatched incentives and coordination problems inherent to decentralized organizations almost invariably create efficiency losses. Within any organization, the most subtle and carefully implemented budgetary scheme creates opportunities for wasteful gaming. A company that is internally divided, whose management does not or cannot trust their subordinates, will be inefficient because it cannot effectively use local knowledge possessed by men and women near the bottom of the organization chart. A government agency that faces conflicting mandates and preferences from its elected overseers is likely to endure costly micro-management that reduces the cost-effectiveness of resulting expenditures.

There is no separating politics from administration. Managers and politicians can still improve things by considering some of the underlying problems that lead to inefficiency. For example, decision-makers might consider the optimal size of organizational subunits. Maybe gatekeeping physicians should operate in small group practices that help to smooth out local variability while maintaining the advantages of peer monitoring in a small group. Procedural "gimmicks" such as balanced budget requirements also deserve a closer look. Research on state governments suggests that such requirements may be helpful in reducing adverse political incentives associated with chronic deficits. Operations researchers and management scientists can make a lasting contribution by developing empirically useful models that illuminate the strengths and weaknesses of such budgetary approaches.

NOTE: To access a related article, "Budgeting for Planning and Control," click here.

Harold Pollack is a Robert Wood Johnson Health Policy Scholar at Yale University. He earned a Ph.D. from Harvard University's Kennedy School of Government in 1994. His research interests include the provision of health services to under-served populations and public policies towards teen mothers.

E-mail to the Editorial Department of OR/MS Today: orms@lionhrtpub.com

OR/MS Today copyright 1997, 1998 by the Institute for Operations Research and the Management Sciences. All rights reserved.

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com

Web Site Copyright 1997, 1998 by Lionheart Publishing, Inc. All rights reserved.
Web Design by Premier Web Designs, e-mail lionwebmaster@preweb.com