The cost of the war: A comment on Stiglitz-Bilmes

April 8, 2008

The U.S. military invaded Iraq in March 2003, and the conflict there has continued for the ensuing five years. In September 2002, CBO published its first projection of the costs associated with a U.S. invasion of Iraq. CBO has subsequently provided the Congress with numerous updates of funding provided to date for that conflict, as well as projections of future costs under several alternative scenarios. Indeed, in part because the Defense Department has never published its own long-range timetable for future U.S. troop levels in the region, the CBO troop scenarios have been used as a basis for cost estimates and other discourse by many other organizations and individual researchers.

A new book about the cost of the war, The Three Trillion Dollar War: The True Cost of the Iraq Conflict, by Nobel-prize winning economist Joseph Stiglitz of Columbia University and his colleague Linda Bilmes of Harvard Universitys Kennedy School of Government, has received significant attention. (To disclose any potential conflict, I am a friend and coauthor of Stiglitz.) In contrast to the cost figure in the title of the Stiglitz-Bilmes book, CBOs most recent testimony suggests that cumulative budgetary costs for the combined conflicts in Iraq and Afghanistan might total between $1.2 trillion and $1.7 trillion through 2017.

A natural question (and one that several people have indeed asked me) is why such a large disparity exists between CBOs projections and those of Stiglitz and Bilmes.

Most importantly, CBO restricts its estimates to Federal budgetary costs, whereas Stiglitz and Bilmes include many other types of costs borne by veterans, their families, and U.S. taxpayers. The CBO testimony notes that the Congress had provided total funding of $604 billion through October 2007 (in then-year or nominal dollars without making any adjustment for inflation), not only for military operations but also for indigenous security forces, diplomatic operations and foreign aid, and medical care and disability compensation for veterans of the conflicts in Iraq and Afghanistan. Stiglitz and Bilmes report a comparable total of $646 billion when inflated to 2007 dollars. So there's not that big a difference when the two sets of estimates are put on a comparable basis.

Below, I go into more detail about all this.

Military costs (not including veterans benefits)

CBOs analysis suggested that, depending on the rates at which U.S. troops are withdrawn from the region under two illustrative scenarios, costs of future military operations (including costs such as combat pay for deployed troops, fuel and transportation, and contractor support in the wartime theater, but excluding costs on medical care and disability benefits provided by the Department of Veterans Affairs, which were tallied separately) through 2017 could total between $485 billion and $966 billion. Using essentially the same CBO scenarios, Stiglitz and Bilmes project future costs of between $521 billion and $913 billion. The similarity of these figures is somewhat misleading, since we are measuring costs in different units. Again, CBO sums up then-year or nominal dollars without making any adjustment for inflation or the time value of money. Stiglitz and Bilmes instead use a present-value approach, in which they effectively discount the flow of future costs at a nominal interest rate of 4.5 percent to reflect the time value of money. Most economists prefer present-value figures, which is why Stiglitz and Bilmes undertook their analysis that way; the Federal budget process, however, is based on nominal dollars, which is why CBO undertook our analysis in the manner we did.

Notwithstanding those differences, the estimates of past and future costs of military operations that CBO has published are in rough agreement with Stiglitz and Bilmes estimates, which account for about half of their $3 trillion estimate of the total cost of the war.

Other costs

Beyond the costs of military operations, the remaining types of costs that Stiglitz and Bilmes consider can be classified into three broad categories: other government spending that could be attributed to the military operations; interest payments made by the Treasury to finance war-related debt; and costs outside the federal budget, which are borne by military veterans who served in Iraq or Afghanistan, their family members or, in the case of reservists, their employers, or U.S. citizens more broadly. Stiglitz and Bilmes track the costs of military operations through 2017, but track veterans disability and medical costs through 2046.

1. Other non-interest government spending Within the category of other government spending, perhaps the most notable differences between CBOs estimates and those from Stiglitz and Bilmes involve the cost of medical care and disability compensation provided by the Department of Veterans Affairs (VA). For example, Stiglitz and Bilmes estimate that monthly compensation paid by the VA to disabled veterans of the conflicts in Iraq and Afghanistan could total between $277 billion and $388 billion over the next 40 years. (CBO truncates its corresponding estimates after 2017, but I will use the 40-year horizon throughout this discussion to maintain consistency with Stiglitz and Bilmes basic framework.)

  • Stiglitz and Bilmes base their estimate on the experience of veterans from the first Gulf War, about 40 percent of whom applied for VA disability compensation since that war ended in 1991 and had their applications approved -- that is, they were rated." One-quarter of those veterans, however, were rated for conditions that are zero-percent disabling; Stiglitz and Bilmes assign those veterans the average benefit level even though they are not currently drawing any benefits. (VA sometimes uses a zero-percent disability rating as a placeholder to acknowledge that a disability first manifested during a period of active military service, although the disability does not yet adversely affect the veterans earning potential.)
  • Furthermore, it is unclear whether Stiglitz and Bilmes adjusted for the number of veterans of the current conflict who might have become disabled as a result of their military service even if they had never deployed to the war zone; those veterans should be subtracted out when estimating the incremental cost of the war. Their footnote 48 (p. 259) outlines a method for making that adjustment, though they begin that discussion by stating, We have not adjusted for incremental cost that is, the number of veterans who may have claimed disability even during peacetime. Even if they did make the indicated adjustment, their method would not allocate enough of the disabilities to the peacetime baseline, and would instead attribute too many disabilities to the war.
  • In summary, we estimate that the incremental number of veterans who are or will draw disability benefits may be only half of what Stiglitz and Bilmes estimate, in which case their costs would be too high by between about $150 billion and $200 billion.

Differences also arise between CBOs and Stiglitz and Bilmes estimates of VA medical costs (as opposed to disability benefits) to treat veterans of the conflicts in Iraq and Afghanistan.

  • In their higher case, Stiglitz and Bilmes apply to veterans of those recent conflicts the $5,765 average amount that VA spent in 2006 to treat veterans of all eras, including the aging Vietnam-era population. Stiglitz and Bilmes then increase the larger amount in subsequent years at a medical-specific inflation rate equal to twice the rate of general inflation. The VA, though, estimates that it spent less than half that initial amount an average of $2,610 in 2006 to treat veterans returning from Iraq and Afghanistan. It may appear surprising to some readers, but veterans of the recent conflicts on average require less medical care from VA than veterans of earlier conflicts such as the Vietnam War do today. (By contrast to recent veterans, many Vietnam-era veterans are now age 60 years or older and are being treated for age-related chronic conditions that demand more resources, on average, from the VA.)
  • It is difficult to project how rapidly VAs medical costs for recent veterans will grow in the future. The growth in costs will depend on how many veterans enter or remain in the VA medical system, as opposed to going elsewhere for their care, and on the change in the average annual cost for veterans who continue to use the system. Some veterans may migrate to other sources of medical coverage, such as employer-sponsored health insurance, for some or all of their care when they assimilate back in to civilian society. Others may find that their medical conditions are resolved satisfactorily with the care that they received at the VA and the passage of time; while they remain reliant on the VA, their demand for medical services will decline. Indeed, some data that we have obtained from the VA indicate that utilization of a variety of medical services tends to decline after a veterans first 3 to 6 months in the VA system (though a few medical services, such as residential rehabilitation for post-traumatic stress disorder or PTSD, show an increasing trend as discussed below). On the other hand, the medical costs for veterans who remain in the VA system or who enroll later will continue to grow as those veterans age, as has been the case for Vietnam-era veterans. It is uncertain how much of the long-run growth in cost should be attributed to the recent veterans service in Iraq or Afghanistan as opposed to aging.
  • Because we believe that the short-term average annual cost ($5,765) in Stiglitz and Bilmes high case is about twice the appropriate value, and in light of the uncertainty in projecting both the average cost and number of veterans of the current conflicts who will utilize VA medical care in the future, Stiglitz and Bilmes projection of up to $285 billion for VA medical costs may be overestimated by at least $100 billion.

Before leaving the topic of VA medical care, I need to touch upon the incidence of PTSD among recent veterans.

  • It is not clear how many veterans suffer from PTSD. DoD has collected data on the incidence of PTSD from the post-deployment health assessment (PDHA) and post-deployment health reassessment (PDHRA) surveys it administers to troops immediately upon return from deployment and again about 6 months later. A series of studies by Army Colonel Charles Hoge, M.D. and his colleagues have used data from those surveys to study the effects of deployments on the incidence of PTSD. Those studies are reporting positive screens for PTSD among 17 percent of active Army soldiers and 25 percent of Army National Guard and Army Reserve personnel. (One of the studies found higher PTSD screening rates for soldiers who had deployed more than once. The rates just cited include some such soldiers: CBO estimates that 27 percent of current active Army soldiers have deployed more than once and 7 percent more than twice. Some 9 percent of current Army reservists have deployed more than once and 2 percent more than twice).
  • Two points should be noted when interpreting the screening rates for PTSD. First, those rates are particular to the Army; while we might expect similar screening results for Marines, we would not for Air Force or Navy personnel who do not participate in direct combat to the same degree as soldiers and Marines. As a lower bound, if all soldiers and Marines (the majority of the deployed forces) experienced the rates quoted above but the rates were zero for Air Force and Navy personnel, the overall screening rate among military personnel who have deployed would be about 10 percent. Second, a positive screen for PTSD on the PDHA or the PDHRA survey does not necessarily imply a diagnosis of PTSD; treatment is largely optional and not all soldiers who screen positive either seek or receive any treatment. Psychiatrists at DoD and VA have advised us that survey responses do not constitute a diagnosis PTSD is not definitive until it is confirmed at two consecutive medical visits.
  • VA, too, collects data on the incidence of PTSD. Their data through October 2007 indicate nearly 50,000 initial diagnoses of PTSD among the 264,000 veterans of Iraq and Afghanistan who have been treated at VA medical facilities an apparent incidence rate of 18 percent. However, their report notes that, up to one-third of coded diagnoses may not be confirmed when initially coded because the diagnosis is rule-out or provisional, pending further evaluation. Removing up to one-third of provisional diagnoses yields an adjusted PTSD incidence rate perhaps as low as 12 percent, roughly consistent with the DoD figures cited in the previous paragraph. VA estimates that it spent $35 million in 2007 to treat PTSD (including provisional cases) among veterans of Iraq and Afghanistan (an average of $700 per patient). That sum represents one-half of one percent of total VA medical costs for veterans of those conflicts.

Debt service costs

Another category of budgetary costs that Stiglitz and Bilmes examine is the cost of debt service. In part to separate the costs of government activities from the financing mechanism (issuing debt versus raising taxes versus crowding-out other programs), CBO does not typically include debt service in its cost analyses. However, at the specific request of the House Budget Committee, CBO calculated the debt-service costs of the war on terrorism under the assumption that all spending for those operations, both past and present, was financed with federal borrowing. Under that assumption, CBO estimates that interest payments on spending thus far would total $415 billion over the 20012017 period. The path of spending generated by CBOs first scenario would contribute an additional $175 billion in interest payments from 2008 through 2017. Under the second scenario, interest outlays would increase by a total of $290 billion over that 10-year period. CBOs range of $590 billion to $705 billion (in undiscounted dollars) over the entire period contrasts with Stiglitz and Bilmes estimate of $613 billion to $816 billion (over the same time period but expressed in discounted present-value dollars their undiscounted totals would be higher).

Other costs beyond the Federal budget

As noted at the beginning of this post, Stiglitz and Bilmes do not limit themselves to federal budgetary costs -- and that explains a substantial amount of the difference between their estimates of the cost of the war and CBO's.

For example, in an attempt to measure the costs of deaths or injuries not compensated by any government agency, Stiglitz and Bilmes apply the economic construct of value of statistical life (VSL). Although the VSL calculation is a legitimate attempt to monetize one type of cost, none of the estimates that CBO has published or on which CBO has testified have included the VSL. The reason is that CBOs charter generally restricts our cost estimates of legislative provisions to the Federal budgetary costs (except that, in accordance with the Unfunded Mandates Reform Act of 1995, we also estimate any mandates that legislation reported from a Congressional committee would impose on state, local, or tribal governments or on the private sector).

Stiglitz and Bilmes assign a VSL of $7.2 million (in 2007 dollars) to each death of a U.S. service member, and a prorated fraction of that amount to each serious injury. Although CBO does not promulgate or endorse any particular value for the VSL, I would note that the Stiglitz and Bilmes estimate is based on guidance that the Environmental Protection Agency promulgated in 2000. However, the EPA selected that value from a survey of 26 studies that estimated VSLs ranging from $0.7 million to $21 million in 2007 dollars; the bulk of those estimates fell within the narrower range of about $4 million to $8 million.

Finally, Stiglitz and Bilmes estimate macro-economic costs such as the effects on the U.S. economy of increased oil prices (to the extent those are a result of the war). The addition of those costs in their most-inclusive accounting, with all of their other assumptions set to high, leads to their estimate of $5 trillion for the total cost of the war. One could debate at length the causality between the war in Iraq and oil prices on the world market; Stiglitz and Bilmes attribute between $5 per barrel and $10 per barrel of the price increase to the war. They translate those price increases into total oil-related macro-economic costs to the U.S. economy of $187 billion and $800 billion, respectively. CBO does not include these types of macro-economic effects in any of our formal cost estimates.


The bottom line, as it were, is that the war in Iraq and Afghanistan has clearly involved significant costs for the federal budget, as well as for soldiers and their families. (The potential benefits are for others to debate.) The Stiglitz-Bilmes estimates of the budgetary costs are likely at least somewhat high, but the biggest difference between our estimates and theirs is that they go beyond the cost to the federal government.

Much of CBOs work in this area has been undertaken by outstanding analysts in our National Security Division (including Matthew Goldberg, the deputy director of that division, and Heidi Golding) and the national security team within our Budget Analysis Division (including Sarah Jennings, who runs that unit, and David Newman).