On December 8, 2003, President George W. Bush signed the Medicare Prescription Drug, Improvement, and Modernization Act of 2003, Public Law 108-173, and Medicare Part D entered the statute books. For the first time, Medicare would cover outpatient prescription drugs for its roughly 40 million beneficiaries. The benefit was voluntary, carried a monthly premium, and did not take effect until January 1, 2006, a full two years after enactment. Lawmakers had debated adding drug coverage to Medicare for decades. The 2003 law ended the debate by writing a new Part D into title XVIII of the Social Security Act. The more consequential decision, though, was not the decision to add a drug benefit. It was the decision about how to deliver it.

Congress did not build Part D the way it had built the rest of Medicare. Instead of paying pharmacies and reimbursing costs the way the program paid hospitals under Part A and physicians under Part B, Congress routed the entire drug benefit through competing private plans. No beneficiary receives drug coverage directly from the government. A beneficiary who wants the benefit must enroll in a prescription drug plan offered in her region. The plan bargains with drug makers and pharmacies. Medicare pays the plan. The statute goes further than permitting private administration. It forbids the Secretary of Health and Human Services from entering the price negotiations that plans conduct. That forbidden step is the noninterference clause, and it makes Part D the clearest statutory statement of the model this series traces: Congress creates an entitlement, then contracts out both its administration and its price negotiation.

Medicare Part D and the Architecture of a Delegated Benefit

Section 101 of the 2003 law, the opening section of its first title, added the new benefit to title XVIII of the Social Security Act as Part D, the Voluntary Prescription Drug Benefit Program, codified at 42 U.S.C. 1395w-101 and following (Social Security Act sections 1860D-1 and following). The placement matters. Parts A and B, hospital insurance and supplementary medical insurance, are administered by the government and pay providers directly, as explained in this series’ account of the Medicare statute (/2010/12/01/medicare-parts-a-b-statute-explained/). Part D sits beside them in the same title but works on opposite principles: the government writes the rules, private organizations bear the insurance risk, and the beneficiary shops among plans. Eligibility is broad and simple. Anyone entitled to Part A or enrolled in Part B may enroll in Part D, regardless of income or health status. Enrollment is voluntary, premiums are subsidized, and a late-enrollment penalty discourages healthy beneficiaries from waiting until they need drugs to sign up.

Enactment did not mean immediate coverage. The drug benefit took effect on January 1, 2006, and Congress filled the two-year gap with a Medicare-endorsed prescription drug discount card and a transitional assistance program that subsidized drug purchases for low-income beneficiaries without coverage. The delay was deliberate. The Centers for Medicare and Medicaid Services needed time to write the contracting rules, divide the country into regions, and run the first bidding cycle before a single plan could enroll a single beneficiary. Grassley, selling the conference report on the Senate floor, named the two groups the delay was meant to serve: low-income seniors too often forced to choose between necessities and prescriptions, and beneficiaries with catastrophically high drug costs. Both groups, he argued, would feel the new benefit first and hardest.

How does Medicare Part D actually deliver drug coverage?

Medicare Part D delivers drug coverage through competing private prescription drug plans, not through the program itself. Beneficiaries eligible for Medicare enroll in a plan offered in their region, the plan contracts with drug makers and pharmacies, and Medicare subsidizes the plan’s bid and pays premiums, cost sharing, and reinsurance.

The delivery system has two doors. Beneficiaries in traditional fee-for-service Medicare enroll in standalone prescription drug plans, known as PDPs, offered by private sponsors under contract with the Department of Health and Human Services. Beneficiaries in Medicare Advantage receive drug coverage through Medicare Advantage prescription drug plans, or MA-PD plans, integrated with their Part C coverage. In both cases the government never buys a pill. It contracts with organizations that do.

Geography organizes the market. The country is divided into 34 PDP regions, and within each region sponsors offer competing plans from which beneficiaries choose. Choice is not decorative; it is the mechanism. Each year, sponsors submit bids to the Centers for Medicare and Medicaid Services based on their expected cost of providing the basic benefit to an enrollee of average health. A national average of those bids sets two numbers: the share of the cost the government subsidizes and the monthly premium the beneficiary pays. A sponsor that bids below the benchmark can attract enrollees automatically, because beneficiaries receiving the low-income subsidy who do not pick a plan are assigned to a benchmark plan in their region at no premium cost to them. Medicare thus pays plans in three ways: a direct subsidy tied to the bid, premium and cost-sharing subsidies for low-income enrollees, and reinsurance payments that protect sponsors against catastrophic losses. Risk corridors in the early years gave sponsors further shelter against unpredictable variation in costs, so that private firms would actually enter a market no federal program had ever tried to create.

The statute defines a standard benefit, and every plan is measured against it. In 2006, the defined standard benefit carried a $250 annual deductible, 25 percent coinsurance on drug costs up to $2,250, no coverage on the next $2,850 of spending, the famous gap in coverage, and catastrophic protection once out-of-pocket spending passed $3,600, with the average monthly premium estimated at $35. Plans could offer this exact benefit, an actuarially equivalent alternative with a different cost-sharing shape, or enhanced coverage with richer benefits for a higher premium. The standard benefit gave CMS a yardstick for judging bids and gave beneficiaries a baseline for comparing plans. It was a floor for comparison, not a ceiling on what sponsors could sell.

Plans compete on more than premiums. Each sponsor designs its own formulary, the list of covered drugs, and typically sorts those drugs into cost-sharing tiers that steer enrollees toward generics and preferred brands. Each sponsor also assembles its own pharmacy network and negotiates its own rebates with manufacturers, which means the price of the same pill can differ between two plans in the same region. CMS reviews and approves the bids and the plan designs each year, but within the statutory guardrails the sponsor, not the agency, decides what the benefit looks like. That discretion is the point. The statute treats plan sponsors as the actors best positioned to bargain, and it treats the beneficiary’s annual right to switch plans as the discipline that keeps sponsors honest.

The low-income subsidy shows how far the delegation runs. Beneficiaries with limited income and assets receive premium subsidies up to a regional benchmark amount, plus reduced cost sharing. Those who do not choose a plan are automatically enrolled in a benchmark plan in their region, which means sponsors compete fiercely to bid at or below the benchmark and capture that assigned enrollment. The government sets the benchmark formula. The market fills the plans. Even the safety net for the poorest beneficiaries operates through private contracts rather than through direct public provision.

Congress also paid employers to stay in the game. Rather than letting the new benefit crowd out existing retiree drug coverage, the statute offered subsidies to sponsors of employer retiree plans that provided drug coverage at least as generous as the Part D standard. The choice reflected the same philosophy as the rest of the design: the government would rather subsidize private coverage that already existed than replace it with a public program. The subsidy was meant to keep employers from dropping drug coverage and shifting their retirees onto the new benefit, extending the delegated model beyond Medicare’s own enrollees and into the employer market.

Congress also built a backstop for markets where competition failed to show up. The statute authorizes government fallback plans for areas without sufficient plan choices, and it requires the Secretary to report to Congress on the use of limited-risk and fallback plans, with recommendations for maximizing the assumption of financial risk by private sponsors. The fallback was a confession and a reassurance at once: a confession that private plans might not appear everywhere, and a reassurance that the benefit would reach every region even if they did not. Students working through these mechanics may find it useful to build a running comparison in a legislation study notebook, because the design only makes sense when each moving part is held next to the alternative Congress refused.

The Noninterference Clause: Three Prohibitions and Their Boundaries

The design described above would not hold without the sentence that locks the government out of the bargaining room. The noninterference clause, section 1860D-11(i) of the Social Security Act, codified at 42 U.S.C. 1395w-111(i), added by section 101(a)(2) of Public Law 108-173, reads in full:

(i) Noninterference. In order to promote competition under this part and in carrying out this part, the Secretary: (1) may not interfere with the negotiations between drug manufacturers and pharmacies and PDP sponsors; (2) may not require a particular formulary, except as provided under section 1395w-104(b)(3)(l) of this title; and (3) may not institute a price structure for the reimbursement of covered part D drugs, except as provided under part E of subchapter XI.

Three prohibitions, one purpose. The Secretary may not insert the government into the price talks between manufacturers, pharmacies, and plan sponsors. The Secretary may not dictate which drugs the plans must cover, apart from the formulary standards the statute sets elsewhere. The Secretary may not set the reimbursement prices the plans pay. Read together, the three provisions make the federal government a market steward rather than a market participant. It approves bids, polices plan conduct, and pays the subsidies. It does not buy.

What the clause does not prohibit is as instructive as what it does. It does not bar plan sponsors from negotiating; on the contrary, the entire design assumes they will, and assumes their negotiations will produce the discounts the government then helps beneficiaries capture. It does not strip the Centers for Medicare and Medicaid Services of its formulary authority under section 1860D-4, the carve-out referenced in the second prohibition. Codifiers have flagged that cross-reference as apparently mislabeled in the original text, but its intent is readable: the agency keeps the power to set formulary standards, and the clause only forbids a government-imposed list of covered drugs. The clause also does not require the Secretary to accept every bid a sponsor submits. In 2010, Congress clarified that bid-rejection authority for contract years beginning on or after January 1, 2011, amending subsection (d) of the same section while leaving the noninterference language untouched. And the clause guarantees no particular price level; it is a rule about who negotiates, not about what the negotiation yields.

The government’s retained powers show what noninterference did not take away. The Secretary writes the contracting rules, approves every plan bid and every plan design each year, and enforces the statute’s beneficiary protections, including the formulary standards and coverage-determination rules of section 1860D-4. CMS can reject bids, police plan conduct, and demand the annual reports on fallback plans. What it cannot do is the one thing the clause names three times in different words: substitute its own judgment for the market’s on what drugs should cost. The line the statute draws runs between regulating the marketplace and participating in it.

Through 2010, no later statute altered the clause. The text stands exactly as enacted on December 8, 2003, which is why the 2003 debate over it still reads as current legislative history rather than as archaeology.

The clause survived because the coalition behind the bill needed it to. The conference agreement nearly collapsed over premium-support provisions that would have forced traditional Medicare to compete against private plans, and the final deal admitted only a limited demonstration program. Senators Max Baucus of Montana and John Breaux of Louisiana broke with most Senate Democrats to join the Republican conferees, supplying the bipartisan cover the bill required. The House approved the agreement 220 to 215 on November 22, 2003, after leadership held the vote open for nearly three hours; the Senate followed 54 to 44 on November 25. The Congressional Budget Office estimated the drug benefit would add $410 billion in direct federal spending over 2004 to 2013. The noninterference language was part of the price of that coalition. The sponsors could not sell a private-plan benefit while reserving a government right to override the plans’ central function, so they wrote the override out of the statute.

The sponsors’ case rested on competition as a cost-control strategy. Senator Chuck Grassley of Iowa, the principal Senate author of the 2003 law and then chairman of the Senate Finance Committee, framed the benefit around choice: “The basis of such legislation is the right to choose for seniors. No one is forced to do anything.” He and his allies described three conditions the benefit had to meet: it had to be voluntary, beneficiaries had to share costs so that spending stayed disciplined, and competition rather than government mandates had to drive prices down. On that last point Grassley was blunt about the logic behind the clause: the government could not be allowed to interfere with the delivery of medicine or to dictate which drugs would and would not be covered, and that was why the law contained a noninterference provision. During the floor debate on the conference report, Senators Max Baucus of Montana and Dianne Feinstein of California joined Grassley in defending the language against the charge that it enriched industry. Baucus said the provision was “not intended to pad the pockets of drug manufactures” and “not intended to pad the pockets of the insurance companies.” Grassley stated the purpose directly: “The purpose of this bill is to ensure that Medicare beneficiaries get the benefit of negotiated discounts that the private sector is able to achieve.” Feinstein added that the bill would ensure seniors paid less for prescription drugs than they paid without it. In 2007, Grassley cited a Congressional Budget Office letter concluding that government negotiation of drug prices would have negligible effect on federal spending and that the government would not obtain significant discounts from manufacturers across a broad range of drugs, a finding his allies used to argue that the clause cost the Treasury nothing. Grassley also pointed out that versions of noninterference language had appeared in several Democratic prescription drug bills between 1999 and 2003, including measures associated with Senators Moynihan, Daschle, and Kennedy and with House Democrats led by Dick Gephardt, which complicated the claim that the clause was an industry invention.

The opponents’ case rested on the opposite reading of the same text. On the Senate floor, Senator Jack Reed of Rhode Island quoted the Des Moines Register editorial board’s description of the conference report as “a big, sloppy kiss to the pharmaceutical and insurance industries,” and added his own gloss: “It is a huge payoff to pharmaceutical companies and to the insurance industry. It is not really about giving seniors what they deserve.” Senator Edward Kennedy of Massachusetts led the filibuster against the conference report, which failed on a 70-to-29 cloture vote before the Senate approved the agreement 54 to 44 on November 25, 2003, three days after the House approved it 220 to 215. Opponents argued that barring the Secretary from the negotiating table surrendered the government’s largest source of leverage and guaranteed that manufacturers would charge Medicare whatever they chose. The charge stuck in public memory even as the benefit took effect, and it shaped every later attempt to revisit the clause.

Viewed as an amendment, Part D changed what title XVIII is. Before 2003, the title described a government that insured and paid. After 2003, it also described a government that subsidized and supervised. The amendment did not repeal the old model for hospitals and doctors; it layered a new model beside it for drugs, and the two models ran side by side through the rest of the decade. That layering is why the statute’s history matters for this series’ framework of amendments and evolution: every later argument about the government’s role in drug prices starts from the sentence Congress wrote in 2003, the sentence that told the Secretary to stay out of the negotiation.

Design choice	Statutory provision	Alternative Congress rejected	Consequence that followed
Delivery through competing private plans	SSA 1860D-11, 42 U.S.C. 1395w-111 (PDP regions, bids, plan approval)	A government-run drug benefit that pays pharmacies directly	CMS contracts with PDP sponsors and MA organizations; beneficiaries choose among plans in 34 regions
Annual competitive bidding to set payments	SSA 1860D-11(b) (submission of bids; plan approval)	Administratively set payments to plans	The national average bid sets the government subsidy share and the beneficiary premium
Noninterference with price negotiations	SSA 1860D-11(i), 42 U.S.C. 1395w-111(i)	Secretary-led negotiation or government-set prices	Plans and manufacturers negotiate; the Secretary is barred from interfering, requiring a formulary, or setting a price structure
Voluntary enrollment with a late-enrollment penalty	SSA 1860D-1 (eligibility, enrollment); 1860D-13 (premiums; late enrollment penalty)	Mandatory enrollment with automatic premium deduction	Anyone entitled to Part A or enrolled in Part B may opt in; the penalty polices adverse selection
Government fallback plans	SSA 1860D-11(f) and (g) (limited-risk and fallback plans)	Leaving thin markets uncovered	Fallback plans authorized where private plans offer too few choices; annual reports to Congress required
Coverage gap between initial coverage and catastrophic protection	SSA 1860D-2 (standard benefit design)	First-dollar coverage with no gap	Beneficiaries paid full cost in the gap until later legislation phased it down
Medicare+Choice renamed Medicare Advantage with restructured payment	MMA Medicare Advantage payment provisions	Keeping the Medicare+Choice name and payment rules	Renamed program with bid-based benchmarks and regional payment floors
Health savings accounts created	MMA provisions creating Health Savings Accounts	No tax-advantaged savings vehicle in the bill	HSAs paired with high-deductible health plans as a new coverage option
Forty-five percent general-revenue funding trigger	MMA Section 801	No fiscal tripwire in the statute	First determination in 2006, first funding warning in 2007, presidential response required
Discount-card bridge for 2004 and 2005	MMA transitional provisions	No coverage before the 2006 launch	Discount cards and transitional assistance operated until Part D took effect
Extended House roll-call vote	House Rule XX, clause 2(a), as applied	A fifteen-minute vote	The longest recorded House vote in modern practice through 2010
Pre-vote cost estimates withheld from Congress	Administrative instruction, not statute	Congress voting with the higher estimate in hand	GAO and inspector general examinations; the debated figure understated costs

Strip away the bidding formulas and the subsidy schedules, and Part D is a delegation twice over. Congress created a federal entitlement and then handed both its administration and its price negotiation to private plans, keeping for the Secretary only the powers of a market steward: approving bids, policing formularies, and paying the bills. The noninterference clause is the sentence where Congress said the design aloud, barring the government from the bargaining table it had built. Everything else in the benefit, the regions, the benchmarks, the subsidies, the fallback plans, was built to make that delegation work in markets the government does not control.

The strangest feature of the Medicare prescription drug benefit created in 2003 was the stretch in the middle where the insurance stopped. The statute defined a standard benefit for the launch year of 2006 with four phases. First came a $250 deductible. Then the plan paid 75 percent of drug costs between $251 and $2,250, with the enrollee covering the other 25 percent. Then came the span in which the plan paid nothing at all, until the enrollee’s own out-of-pocket spending reached $3,600. Only then did catastrophic protection begin, with the plan covering 95 percent of further costs and the enrollee paying nominal amounts, $2 for a generic or preferred multisource drug and $5 for anything else. That middle span, where the enrollee paid the full price of every prescription, became known across the country as the donut hole.

The arithmetic behind the hole mattered as much as the nickname. The $3,600 figure was “true” out-of-pocket spending, meaning only money the enrollee actually paid counted toward it. Payments made on the enrollee’s behalf by an employer plan or another third-party arrangement did not count, a rule that prevented wraparound coverage from quietly erasing the gap. In the standard design, reaching $3,600 in true out-of-pocket costs corresponded to $5,100 in total drug spending. Both thresholds, along with the deductible and the initial limit, were fixed for the launch years and subject to annual indexing afterward under the statute’s update rules. Beneficiaries with low incomes received separate premium and cost-sharing subsidies that softened or erased these phases, but for the standard enrollee the hole was concrete: several thousand dollars of annual drug spending with no plan contribution at all.

A worked example shows how the phases stacked. Consider an enrollee with $6,000 in covered drug spending during 2006. She paid the $250 deductible, then 25 percent of the next $2,000 in costs, or $500, bringing her true out-of-pocket total to $750. She then paid full price through the gap until her own spending reached $3,600, which required another $2,850 of her money and brought total drug spending to $5,100. Only the final $900 of her annual drug bill fell under catastrophic protection, where her share dropped to a few dollars per prescription. Her total out-of-pocket cost for $6,000 in drugs came to roughly $4,450 plus premiums, and most of it was incurred inside the hole.

Participation was voluntary, and the Congressional Budget Office estimated the average monthly enrollee premium at $35 for 2006. Beneficiaries could obtain the benefit through stand-alone prescription drug plans or through Medicare Advantage prescription drug plans that bundled drug coverage with medical coverage. The law did not force every plan to reproduce the standard four-phase shape; it permitted alternative designs of actuarially equivalent value, and some plans chose to cover part of the gap in exchange for higher premiums. But the standard design set the template, and the hole defined the public argument about the benefit from the day it launched on January 1, 2006.

The standard benefit also functioned as the benchmark against which every competing plan was measured. Actuarial equivalence meant a plan could reshape the cost sharing, lower the deductible, or cover some generic drugs inside the gap, so long as the total value of the package matched the standard design. Enhanced plans could go further, offering benefits beyond the standard package, usually for a higher premium. This structure turned the donut hole into a competitive variable: plans that filled part of the gap could advertise that fact, and beneficiaries willing to pay more could buy their way around the hole. The gap was therefore not only a cost-control device but a product-differentiation instrument, a feature of the market the statute had deliberately created rather than a mere omission.

Because the full benefit did not begin until 2006, the statute built a bridge for the two intervening years. From mid-2004 through 2006, beneficiaries could purchase a Medicare-endorsed prescription drug discount card that secured negotiated discounts at the pharmacy counter, and low-income enrollees received transitional financial assistance toward their drug purchases along with a subsidized enrollment fee for the cards. The cards were a stopgap, not insurance, and they phased out as the permanent benefit phased in, but they reveal how the designers thought about the interim problem: some relief now, delivered through private vendors, without committing the Treasury to the full entitlement before its start date.

The low-income protections sat outside the standard four-phase design and softened it substantially. The statute provided premium and cost-sharing subsidies for beneficiaries with incomes below 150 percent of the poverty line, reducing or eliminating the deductible, the coinsurance, and the gap for the poorest enrollees. This two-tier structure explains part of the political durability of the design: the hole fell hardest on middle-income beneficiaries with substantial drug needs but no subsidy, while the lowest-income enrollees were largely shielded from it.

The standard design distributed its value unevenly across the spending distribution, and that unevenness explains much of the political heat the hole generated. Enrollees with very low drug spending never reached the gap and received a straightforward subsidy on their modest costs. Enrollees with very high spending passed through the gap into catastrophic protection, where the plan absorbed nearly everything beyond nominal cost sharing. The worst value fell on enrollees in the middle: beneficiaries whose annual drug costs landed them squarely inside the hole paid premiums all year and then paid full price for the very prescriptions that had pushed them there. The design was most generous at the two ends of the spending distribution and least generous in the middle, which is precisely why the nickname stuck and why the middle became the focus of the later legislation.

Why did Congress build a coverage gap into Part D?

Congress built the gap to hold down the ten-year budget cost of the new benefit while protecting enrollees against catastrophic prescription spending. By concentrating federal dollars at the top of the spending distribution rather than on routine prescriptions, lawmakers kept the official ten-year price near $400 billion and followed an insurance principle: cover the rare disaster, not the predictable routine.

The budget math came first. The 2004 budget resolution had designated $400 billion over ten years for a Medicare reform package that included a prescription drug benefit, and the conference agreement had to fit inside that number to survive. Every dollar of plan payment inside the gap would have raised the Congressional Budget Office score. Leaving the middle uncovered let the authors promise catastrophic protection for every enrollee while keeping the estimate inside the figure the White House and congressional leadership had agreed to defend. A benefit that paid from the first prescription onward would have cost substantially more and would have forced higher premiums, deeper cuts elsewhere in the bill, or an admission that the price tag was larger than advertised. The politics were tight enough already: the conference agreement passed the House 220 to 215 on November 22, 2003, in a vote held open for hours while the leadership gathered the final ayes, and the Senate approved it 54 to 44 three days later before President George W. Bush signed it on December 8.

The second reason was the insurance logic embedded in the design. An insurance benefit, as distinct from an assistance program, exists to shield people against costs that are large relative to their resources. Moderate, predictable drug spending is something most households can budget for; the financial disaster is the beneficiary whose annual drug bill runs into five figures because of a cancer regimen or a transplant protocol. The standard design aimed federal money at that upper tail of the spending distribution. The deductible and the gap asked enrollees to finance their own routine costs, while the catastrophic phase capped the worst outcomes. One could argue the design asked too much of the middle, and many did, but the structure reflected a deliberate choice about what insurance is for.

The design also embodied a deliberate choice about how drug prices would be set. The statute did not permit the government to negotiate prices directly with manufacturers. Instead it relied on the competing private plans to negotiate prices and determine payments, a wager on market bargaining rather than administered pricing. Critics attacked that choice from the start, arguing that the largest purchaser of prescription drugs in the country was being denied the bargaining power its scale would imply, but the prohibition on government negotiation was integral to the private-delivery model: the plans, not Washington, were to be the buyers. Understanding the hole therefore requires understanding the delivery system around it. The gap was not simply a missing layer of coverage; it was the cost-sharing core of a benefit that the government financed but private insurers designed, priced, and administered.

The third reason was the memory of 1988, which haunted every negotiation. Congress had tried a Medicare drug benefit once before, and the episode ended in repeal. The Medicare Catastrophic Coverage Act of 1988 phased in an outpatient prescription drug benefit that would have paid half the cost of drugs above $600 starting in 1991 and 80 percent above a deductible once fully implemented in 1993 the earlier repealed drug benefit. That benefit never took effect. Seniors revolted against the income-related premiums financing the whole law, angry that they were being asked to pay up front for benefits phased in years later, and the Congressional Budget Office nearly doubled its estimate of the drug provision’s five-year cost, from $5.7 billion to $11.8 billion, within a single year, which fed the sense that beneficiaries were buying into an open-ended obligation. Congress repealed the entire act in November 1989, only seventeen months after passage. Two lessons traveled forward into the 2003 talks. The first was about financing: do not fund a new benefit with visible new premiums levied on all beneficiaries. The 2003 designers answered with general-revenue financing and enrollee premiums tied only to the drug benefit itself. The second was about timing: do not collect money now for coverage that starts later. The new benefit began paying on its start date, even if its middle phase asked enrollees to pay their own way for a stretch.

That 1988 benefit deserves to be kept distinct from Part D, because the two are sometimes blurred in retellings. The earlier provision was a component of a catastrophic-care law, administered inside traditional Medicare, aimed at high drug spenders through a deductible that only a fraction of beneficiaries would ever reach, and it died before a single prescription was ever covered under it. Part D was a separate, voluntary, privately delivered insurance benefit with its own financing, its own enrollment rules, and its own four-phase design. The donut hole belonged to the second episode, not the first, and conflating them collapses two decades of policy history into one.

The gap, in turn, was not a permanent fixture of the statute, and a second recurring error treats it as if it were. In 2010 Congress enacted the major health law of that year, which began closing the hole on a schedule the later legislation that closed the gap. Enrollees who reached the gap in 2010 and did not receive low-income subsidies got a one-time $250 rebate. Starting in 2011, manufacturers of brand-name drugs were required to fund a 50 percent discount on their products purchased inside the gap, with the value of that discount counting toward the enrollee’s out-of-pocket threshold, and the law laid out a year-by-year phase-down under which the enrollee’s share of costs in the gap would fall to 25 percent. The statute also moderated the annual growth of the out-of-pocket threshold between 2014 and 2019, holding down the price of admission to catastrophic protection during the phase-out years. The original design’s middle phase thus entered a planned dismantling only four years after the benefit launched, which is worth stating plainly: the donut hole was written into the 2003 act, and later legislation wrote it back out.

The donut hole drew the most public attention, but the 2003 act made two other structural changes to American health financing that reshaped the systems they touched. The first was the remaking of Medicare’s private-plan program. The Balanced Budget Act of 1997 had created Medicare+Choice, opening the program to a wider set of private plan types, but by 2003 the program was shrinking. Enrollment in private plans had fallen from 6.2 million in 1998 to 4.6 million in November 2003, and the number of participating risk plans had dropped from 346 to 155. Plans complained that payment rates were too low and too unpredictable to sustain service, and beneficiaries in many counties found themselves with few private options or none at all.

The 2003 act answered by renaming Medicare+Choice as Medicare Advantage and restructuring what plans were paid. Payment rates rose beginning in March 2004, with plans guaranteed at least 100 percent of what traditional Medicare spent in the counties they served, and the law imposed a new minimum annual update, 6.3 percent for 2004 against the 2 percent updates of the prior years. The statute replaced the old administered pricing with a system of bids and benchmarks, created a stabilization fund to give plans an incentive to enter and stay in the program, and established regional preferred-provider plans to widen the program’s geographic reach beyond the county-based plans of the old regime. It also authorized a six-year comparative cost adjustment demonstration in which traditional fee-for-service Medicare would compete directly against private plans in a limited set of metropolitan areas, a cautious experiment with the premium-support idea that had divided the conference. Medicare Advantage prescription drug plans, which bundled medical coverage with the new drug benefit, became one of the two main channels through which beneficiaries received Part D, the other being stand-alone prescription drug plans, so the private-plan overhaul and the drug benefit were built to operate as a pair.

The rename itself carried a message. Medicare+Choice had become associated with plan withdrawals and shrinking options, and Medicare Advantage promised the opposite by name. The regional preferred-provider organizations, new in 2006, were meant to solve the geographic problem the county-based system had never cracked: plans that could operate across multi-state regions rather than assembling county-by-county networks.

The same title carried several smaller structural provisions that showed the direction of travel. Beginning in 2007, the Part B premium would rise for high-income beneficiaries, phased in over five years, introducing an income-related element into a premium that had always been uniform. The Part B deductible rose to $110 in 2005 and was indexed from 2006 onward. And the act created a fiscal tripwire: if general-revenue funding for the entire Medicare program ever exceeded 45 percent of total outlays, the President would be required to submit corrective legislation and Congress to consider it. None of these provisions drew the headlines that the drug benefit did, but together they marked a turn toward means-testing at the top and budget discipline across the program.

The second structural change had nothing to do with Medicare, and it may prove the longer-lived of the two. The act created health savings accounts, tax-preferred accounts available to individuals enrolled in qualifying high-deductible health plans. Contributions went in on a tax-advantaged basis, earnings accrued tax free, and withdrawals for qualified medical expenses were tax free as well, a triple tax advantage that neither ordinary savings nor the older flexible spending accounts could match. Unused balances rolled over from year to year instead of expiring, and the accounts belonged to the worker and traveled from job to job. The design grew out of the earlier medical savings account pilot, which Congress had hemmed in with enrollment caps and a sunset date; the new accounts removed those restraints and opened the model to far more of the under-65 insurance market. The theory was consumer direction: pair a low-premium, high-deductible policy for major medical bills with a personal account for routine expenses, and let households weigh price against care with their own money at stake. Enrollment grew quickly once the accounts became available in 2004, surpassing three million account holders within a few years. Both the individual and the employer could contribute to an account, though the money belonged to the worker regardless of who funded it, and eligibility required enrollment in a qualifying high-deductible plan meeting Treasury standards, with deductibles of at least $1,000 for individual coverage or $2,000 for family coverage. The accounts could be obtained through either the individual or the employer-group insurance markets, which is why their footprint extended well beyond any single employer or insurer. For households that stayed healthy, the unused balance compounded into a growing reserve for future medical costs; for households with heavy routine costs, the high deductible meant the account could be exhausted long before the insurance began to pay. That asymmetry was the point of contention in every debate about the model, and it was built into the statute from the beginning.

Taken together, the three changes trace a single philosophy. The drug benefit asked beneficiaries to insure against the catastrophic tail of drug spending while managing the middle themselves; the private-plan overhaul asked the market to deliver Medicare benefits under restructured public payments; the savings accounts asked workers to manage routine medical spending with their own tax-sheltered dollars. The donut hole drew the most attention because it was the most visible and the most felt, and its scheduled phase-out under the 2010 law showed that even the most deliberate design features are provisional. But the mechanics underneath, the four-phase benefit, the renamed and repriced private-plan program, and the new savings vehicle, were the act’s real architecture, and each of them altered the financing of American health care in ways that outlasted the controversies of the enrollment years. The donut hole is the feature the public remembers, yet it was only one element of a statute that reorganized how drug costs were shared, how private plans were paid, and how working households saved for medical care. The hole could be legislated away because it was a parameter. The architecture proved harder to dislodge.

The conference report on H.R. 1 came back to the House late on the night of November 21, 2003, after months of bargaining between the chambers. The House had passed its version of the Medicare prescription drug legislation in June by a single vote, and the Senate had passed its own version that same month. Reconciling the two measures consumed the summer and the fall, and when the conference committee at last produced its report, Republican leaders scheduled floor action for the middle of the night. Debate began around midnight, and the roll call on the conference report was ordered at about 3:00 a.m. on November 22. Because the House had not adjourned, the Congressional Record dates the action to the legislative day of November 21, even though the roll call closed in the early hours of November 22. Opponents said the hour was chosen to keep the proceedings out of public view. The stakes were plain: opponents described the conference report as the most sweeping change to Medicare in the program’s thirty-eight year history. What followed made the hour a secondary matter. The electronic roll call that began at about 3:00 a.m. did not close until about 5:51 a.m., a span of roughly two hours and fifty-one minutes that news accounts described as the longest vote in House history: the longest recorded House vote in modern practice through 2010.

To understand the night, it helps to start with what House rules actually require. Clause 2(a) of Rule XX provides that the minimum time for a record vote or quorum call by electronic device shall be fifteen minutes. The Congressional Research Service, in its account of House voting procedure, stresses that fifteen minutes is the minimum time rather than a fixed or maximum time. In practice the period allowed is often extended so that as many members as possible can reach the floor and record themselves, and electronic votes frequently consume twenty minutes or longer. The Speaker, or the chair of the Committee of the Whole, may close a vote at any time after the fifteen minutes have elapsed, and the chair sometimes allows several more minutes for members on their way to the chamber. While a roll call is open, members may change their recorded positions at any time before the chair announces the result, and no position may be recorded afterward. The CRS account notes the other edge of this discretion as well: the chair’s power to decide how long to leave a roll call open after fifteen minutes could be used to the majority party’s advantage, since a chair whose side is losing a close vote could keep the roll call open far longer than fifteen minutes to reverse the outcome by persuading members to change their positions or by waiting for more members to arrive. Speakers have announced, on the other hand, that they would not close electronic votes while members in the chamber were seeking to be recorded.

House members had seen this playbook before. When the House first passed its version of the bill in June, the initial electronic tally stood at 214 in favor and 218 against, according to contemporaneous accounts. Leadership persuasion then moved three Republicans: Ernest Istook changed his vote to present after learning that Representative C. W. Bill Young of Florida, absent because of a death in the family, would have voted in favor had he been present, and Butch Otter and Jo Ann Emerson switched to the affirmative under pressure from the leadership. The bill passed that night 216 to 215 with one member voting present. The June proceedings had already established the pattern of keeping the roll call open while the leadership worked wavering members, a tactic the November effort repeated. In November, with the conference report on the floor, Hastert and DeLay set out to repeat the June result, and the conference itself had given them a fragile product to sell: the Congressional Research Service later noted that one major difference fought out in conference was the premium support concept, under which the original fee-for-service program would have been required to compete against the new private-plan program, an idea the final agreement confined to a six-year comparative cost experiment in a limited number of metropolitan areas.

What broke the fifteen-minute rule on the November 2003 House vote?

Nothing broke the rule as written, because clause 2(a) of Rule XX sets fifteen minutes as the minimum time for a record vote, not a maximum, and gives the chair discretion to keep the roll call open longer. The controversy was over whether the leadership used that discretion to reverse a losing outcome.

The question matters because the two sides read the same rule and saw different obligations in it. Defenders read a grant of discretion with no upper bound, designed to make sure no member is shut out of a decision. Opponents read a convention built around fifteen minutes precisely to prevent a chair from manufacturing a majority after the House has spoken. The night of November 21 and 22 put both readings to the test.

The roll call began at about 3:00 a.m. on Saturday, November 22, 2003. About forty-five minutes into the voting, news accounts reported, the conference report was losing, 219 to 215, with Representative David Wu of Oregon not yet recorded. With Wu unrecorded, just three changed votes would have been enough to reverse the outcome. At that point the leadership could have gaveled the proceedings closed and accepted defeat. Instead, Speaker Dennis Hastert of Illinois, Majority Leader Tom DeLay of Texas, and Majority Whip Roy Blunt of Missouri spent the early morning hours working the floor, urging wavering Republicans to switch. A Democratic account later entered in the Congressional Record described the three leaders moving among members in the middle of the night to assemble the switches that would produce the result they wanted. Representative Ernest Istook of Oklahoma, described in news accounts as a wavering vote, changed his position, and the margin narrowed. Still short of a majority, the leadership kept the roll call open.

A Democratic member’s later floor account described the tally as frozen through the small hours: 216 in favor and 218 against at 5:00, still 216 to 218 at 5:30, and unchanged at 5:45, nearly three hours after the roll call had been ordered. Then, the account continued, two Republicans, one from Idaho and one from Arizona, emerged from the cloakroom and marked paper ballots by hand, because the plastic electronic cards could no longer be used that late in the proceedings, and handed the green cards to the clerk before the Speaker gaveled the proceedings closed.

The most replayed sequence of the night came near the end. At about 5:50 a.m., Representatives Butch Otter of Idaho and Trent Franks of Arizona changed their votes to the affirmative. With passage in reach, Wu recorded his vote in favor, and Democratic Representatives Calvin Dooley of California, Jim Marshall of Georgia, and David Scott of Georgia moved into the affirmative column. Two members moved the other way in the closing minutes: Representative Brad Miller of North Carolina and Representative John Culberson of Texas changed from the affirmative to the negative. At about 5:51 a.m. the chair announced the result, and the House agreed to the conference report, 220 to 215. Contemporaneous reporting put the affirmative side at 204 Republicans and 16 Democrats. The roll call had run about two hours and fifty-one minutes from the time it was ordered, roughly three hours in the common telling.

Defenders of the leadership built their case on the text of the rule. Fifteen minutes, they noted, is the floor of House practice, not its ceiling, and the chair’s discretion to extend a roll call is explicit in the rules and confirmed by precedent. Votes routinely run past fifteen minutes so that members delayed off the floor can record themselves, and members keep the right to change their recorded positions until the chair announces the result. On this view nothing about the procedure departed from the rules: the roll call stayed open, members continued to deliberate and to revise their positions, and the final tally reflected the positions members held when the chair closed the proceedings. Defenders added that closing the roll call at the forty-five minute mark, with members still seeking to make themselves heard, would itself have been the questionable act, since Speakers had announced a policy of not closing votes while members in the chamber sought to be recorded. The leadership’s position was that it was doing what majorities are expected to do, working within the chair’s recognized discretion to assemble support for a measure it considered vital. The conference report passed, on this account, because a majority of the House ultimately recorded itself in favor, 220 to 215, and the length of the roll call was the price of giving every member the chance to be counted.

Defenders also pointed out that the chair’s prerogative had been exercised by both parties. In one earlier episode described on the House floor, a vote was held open past seventeen minutes at the request of the Democratic cloakroom while members were meeting with the President at the White House, and the chair’s authority to lengthen a roll call was described on the floor as clear in the rules. On this view the November 2003 proceedings differed in degree rather than in kind: the discretion was the same one chairs of both parties had used, and the rules had never set an upper limit on it. Defenders also noted that the closing minutes showed movement in both directions, with two members switching from the affirmative to the negative even as others moved the other way, which they offered as evidence that members were deliberating rather than simply being herded.

The opponents’ case, pressed most forcefully by Minority Leader Nancy Pelosi of California and other Democrats, started from the same facts and drew the opposite conclusion. After forty-five minutes the measure was losing, 219 to 215, and the only business conducted during the next two hours, opponents argued, was pressure. The roll call was kept open, in their telling, for the sole purpose of reversing the outcome by persuading members to change recorded positions, which is exactly the use of the chair’s discretion that the CRS account flags as a potential partisan advantage. Democratic members said on the House floor that the public had not seen leaders working through the night to produce the desired result, and one Democratic account in the Congressional Record called the episode “the subversion of democracy.” Critics also contrasted the night with the policy announced by the Speaker in January 1995, which had called for closing electronic votes as soon as possible after the guaranteed fifteen minutes, citing an October 1991 session in which the chair strictly enforced that discipline with the members’ cooperation. The 2003 roll call, opponents argued, was the opposite of that discipline. Representative Frank Pallone of New Jersey, speaking in the House after the Thanksgiving recess, said the board had been held open for three hours even though most members had voted, and described his constituents as outraged by the bill and by the reported arm-twisting behind its passage. On December 8, 2003, a House resolution alleging intentional misuse of House practices in holding a vote open for approximately three hours for the sole purpose of circumventing the will of the House was held to constitute a question of the privileges of the House, a parliamentary acknowledgment that the charge was serious enough to be heard. A Democratic resolution condemning the extended roll call and seeking a formal complaint against the leadership was tabled on a party-line vote of 214 to 189. The House Manual records that a question of privileges over the approximately three-hour vote was raised again on December 8, 2005, a sign of the episode’s staying power as a parliamentary reference. The argument had a legislative afterlife. When Democrats took the majority in the 110th Congress, the 2007 House rules package added a sentence to clause 2 of Rule XX providing that a record vote by electronic device shall not be held open for the sole purpose of reversing the outcome of such vote. During the debate on that package, a Democratic leader recalled staying in the chamber into the small hours of the morning for three hours while the 2003 vote was held open so that Republican leaders could, in his words, twist enough arms to win it, and said the new language would keep it from happening again. The provision was repealed in the rules package of the 111th Congress in 2009, but its adoption showed how the November 2003 roll call had become the reference point for the claim that the fifteen-minute convention needed a written guard against exactly what happened that night.

The night’s conduct produced a formal reckoning the following year. Representative Nick Smith, a Michigan Republican retiring at the end of the Congress, posted a statement on his official congressional website the day after the roll call saying he had been offered campaign support for his son Brad, who was running for Smith’s seat, in exchange for changing his vote to the affirmative. Smith later clarified that no explicit offer of campaign money had been made, but that he had been offered substantial and aggressive campaign support which he had assumed included financial backing. The House Committee on Standards of Official Conduct established an investigative subcommittee, and the inquiry reached deep into the leadership: the subcommittee deposed the Speaker, the Majority Leader, the chairman of the Ways and Means Committee, and numerous other senior members. On September 30, 2004, the full committee adopted the subcommittee’s report as H. Rept. 108-722.

The report’s findings cut in several directions. The subcommittee found that no group, organization, business interest, or corporation of any kind had offered one hundred thousand dollars or any other specific sum to support Brad Smith’s candidacy in order to induce the congressman’s vote, and that the National Republican Congressional Committee had not offered endorsement or financial support for the son’s candidacy in exchange for the father’s vote. Smith’s public statements to that effect, the report concluded, appeared to have been the result of speculation or exaggeration, and Smith had failed to cooperate fully with the committee’s chairman and ranking member as they tried to develop the facts. The report said Smith had failed to exercise reasonable judgment and restraint, and held him accountable for public statements that risked impugning the reputation of the House. The subcommittee also found, however, that Majority Leader DeLay had personally offered his endorsement of Brad Smith in exchange for Representative Smith’s vote in favor of the Medicare bill. That offer, the subcommittee concluded, went beyond the bounds of maintaining party discipline. The promise of political support for a member’s relative, the report said, is not related to the functioning of government and should not be made or accepted in exchange for a vote, and linking official action to personal benefit as a quid pro quo could support a finding that DeLay had violated House rules.

A second thread of the inquiry concerned Representative Candice Miller, also a Michigan Republican. The subcommittee found that Miller had made a statement to Smith on the House floor during the roll call that referenced his son’s candidacy, that Smith had fairly interpreted the statement as a threat of retaliation for voting against the bill, and that Miller’s statements were improper and had contributed to Smith’s decision to make his public allegations. The subcommittee concluded that Miller shared a portion of the responsibility for a course of events that risked impugning the reputation of the House. Two other members, Representatives Randall “Duke” Cunningham of California and James Walsh of New York, had also made statements to Smith referencing his son’s candidacy, but the subcommittee concluded that neither had violated House rules. The committee’s final action was a public admonishment of all three principal figures, Smith, Miller, and DeLay, through publication of the report, with no recommendation of further proceedings against any of them under House and committee rules. Beyond the three admonishments, the report set out to clarify the governing standards, advising members, employees, and officials of the House that linking official actions with personal considerations in the manner described in the report was out of bounds. Brad Smith, the son at the center of the allegations, went on to lose the primary for his father’s seat.

Three days after the House acted, the Senate took up the conference report under the threat of a filibuster and agreed to it on November 25, 2003, by 54 to 44, and President George W. Bush signed the measure into law on December 8, 2003, as Public Law 108-173. The prescription drug benefit at the heart of the bill would not take effect until 2006, but the procedure that carried the bill through the House left its own mark well before then. The roll call of November 22, 2003, ordered at about 3:00 a.m. and closed at about 5:51 a.m. at 220 to 215, became the case study for every later argument about how long a House vote may run and to what end: the longest recorded House vote in modern practice through 2010, and the reason the fifteen-minute rule continued to be debated through 2010 as a question of what the rule permits and what it was meant to prevent. In Congressional Research Service accounts of House voting procedure written afterward, the chair’s discretion to keep a roll call open past fifteen minutes is explained alongside an explicit warning that it could be used to the majority party’s advantage when its side is losing, the November 2003 night having supplied the example.

The prescription drug benefit that the Medicare Modernization Act created in 2003, known as Medicare Part D, carries two financial histories, and they are usually told as if they were one. The first is the fight over what the benefit would cost before Congress voted. The second is what the benefit actually cost after it started paying claims. This section keeps them apart because the documents keep them apart. One is the story of an estimate that lawmakers never saw. The other is the story of spending that came in lower than nearly every forecast. Neither record erases the other, and each one rests on findings that named institutions put in writing.

The first history begins inside the Centers for Medicare and Medicaid Services in the summer of 2003. Richard S. Foster, the agency’s chief actuary, had spent months modeling the drug proposals moving through Congress, and his figures kept landing far above the number the White House was citing in public. Foster was not a partisan combatant. The New York Times had described him as having a reputation for careful assessments, and his office had a long tradition of dealing directly with the legislative branch, a tradition Foster understood himself to be bound by. Around the middle of June, Foster told Thomas A. Scully, the CMS administrator, that the administration’s numbers were off by many billions of dollars. Foster did not keep the analysis inside the agency. He sent his calculations to Doug Badger, a White House health policy adviser, and to staff at the Office of Management and Budget and the Department of Health and Human Services. Inside the executive branch, the higher number was known. On Capitol Hill, where members were about to vote, it was not.

Scully ordered Foster to keep the higher figures from Congress. When Foster objected, Scully threatened to fire him. The confrontation left a paper trail almost immediately. In an email to colleagues dated June 26, 2003, sent just before the first House vote on the drug legislation, Foster called the episode nightmarish and wrote that he was perhaps no longer in grave danger of being fired but that a strong likelihood remained that he would have to resign in protest over the withholding of important technical information from key policymakers for political reasons. Knight Ridder obtained a copy of the message. Cybele Bjorklund, the Democratic staff director of the House Ways and Means health subcommittee, told Knight Ridder that Scully had told her himself that he had ordered Foster to withhold the information and that Foster would be fired for insubordination if he disobeyed. Scully later offered a narrower account. He said he had curbed Foster on only one specific request, made by Democrats on the eve of the first House vote in June, because he believed they wanted the numbers to disrupt the floor debate. He denied that he had threatened to fire Foster, while acknowledging that he had told Foster to withhold the figures from Congress. Foster’s account was not narrow. He said Scully told him, “We can’t let that get out.”

The House and Senate cast their final votes in November 2003 with the lower figure framing every argument. The Congressional Budget Office had scored the legislation at $395 billion over ten years, and the number debated on Capitol Hill was a $400 billion ceiling that the bill was not supposed to breach. Fiscal conservatives in the House supported the measure only because they believed that ceiling would hold. President George W. Bush signed the Medicare Prescription Drug, Improvement, and Modernization Act on December 8, 2003. Scully resigned the same month. Then, in January 2004, the administration released its own ten year estimate for the drug benefit: $534 billion, a figure $134 billion above the $400 billion number discussed in the debate. The higher number had existed before the votes. It reached the public only after them.

The margin made the missing number matter. The House approved the conference report in the early hours of November 22, 2003, by 220 votes to 215, after Republican leaders held the roll call open for nearly three hours, from about 3:00 a.m. to 5:51 a.m., the longest electronic vote in the chamber’s history. The Senate followed on November 25 with a 54 to 44 vote. With majorities that thin, the votes of fiscal conservatives who had been promised a $400 billion ceiling carried decisive weight. Rangel would later say at the March hearing that disclosure of the higher figure would have killed the bill. The remark was partisan, and no one could prove how members would have voted with different information. But the closeness of the 220 to 215 tally is a matter of record, and it is why the suppressed figure has remained the central fact of the episode rather than a footnote.

The disclosure came in stages through March 2004. Reporters obtained Foster’s June 2003 email, and on March 13 the Philadelphia Inquirer and the Washington Post reported that the chief actuary had been ordered to withhold unfavorable estimates that exceeded what Congress seemed willing to accept by more than $100 billion. Foster confirmed the reporting. HHS Secretary Tommy Thompson asked the department’s Office of Inspector General to investigate the allegations. On March 18, 2004, eighteen Senate Democrats, including Minority Leader Tom Daschle, Edward Kennedy, Hillary Rodham Clinton, and Frank Lautenberg, wrote to the Government Accountability Office requesting a legal opinion on whether federal law had been violated. Four of the senators also wrote to Attorney General John Ashcroft asking the Justice Department to examine whether two federal criminal statutes on withholding information from Congress applied. The requests put the episode into formal inquiries on two tracks at once, one inside the department and one in the investigative arm of Congress itself.

Foster gave his account under oath on March 24, 2004, before the House Ways and Means Committee. He told the committee that he had shared his higher figures with White House, HHS, and OMB officials, and he described for the first time in public how Scully had threatened to fire him if he answered lawmakers’ requests for estimates of the pending bills. His number had moved as the legislation changed, he said, but it had stayed between $500 billion and $600 billion over ten years. He reminded the committee that conference report language in the 1997 Budget Act had described the Medicare actuary as an independent office charged with giving Congress prompt and impartial information. Scully, Foster testified, had dismissed that language in unprintable terms. After the confrontation, Foster consulted a lawyer inside HHS, who advised him that Scully did have the authority to block the disclosures. Foster stayed at the agency rather than resign, describing Scully’s demands as inappropriate and unethical. Charles Rangel, the committee’s ranking Democrat, told Foster that the bill would have died if the higher figure had been public before the vote. Democrats on the committee did not get the conclusive proof of wider White House direction that some of them had hoped for, but the hearing put Foster’s version of the instruction and the threat into the congressional record under his own name.

Why did the Part D cost estimates change after the vote?

The pre-vote number came from the Congressional Budget Office, which scored the bills at $395 billion over ten years. The post-vote number came from the administration, whose own actuaries put the same legislation at $534 billion. Different models, different assumptions about enrollment and drug prices, and different institutional roles produced the gap.

The two numbers were built for different purposes. The CBO score was the official cost of the legislation under House and Senate budget rules, which is why it framed the debate and why the $400 billion ceiling carried political force. The mechanics of that scoring, and why Congress treats the CBO number as binding, are explained in our guide to how CBO scores legislation. The $534 billion figure was the administration’s own projection, built on modeling by Foster’s office and released by OMB in January 2004, after enactment had made the question retrospective. The institutions were answering different questions on different timelines.

The assumptions behind the numbers mattered more than the arithmetic. CBO and the CMS actuaries diverged on how many beneficiaries would enroll in the new benefit, how quickly prescription prices would rise, how aggressively private plans would bid for enrollees, and how much the benefit’s coverage gap would cost. Foster’s figures also moved with the legislation itself. His testimony put his own range between $500 billion and $600 billion as the bills changed shape through 2003, which meant there was never a single Foster number so much as a band of them, all of them above the figure Congress was using. None of this modeling was secret inside the executive branch. Foster had circulated his analysis to the White House, HHS, and OMB months before the final votes. What changed after the vote was not the underlying work. What changed was which number the public was allowed to see.

The inspector general’s inquiry reported first. On July 6, 2004, Dara Corrigan, the department’s acting principal deputy inspector general, issued findings that confirmed the core of Foster’s account. Her report stated that CMS had not provided information requested by members of Congress and their staffs and that Scully had threatened to sanction Foster if he disclosed unauthorized information. On the legal question, though, the report broke in the administration’s favor. Corrigan concluded that neither the threat nor the withholding had violated any criminal law, accepting the Justice Department’s view that the CMS administrator possessed final authority to determine the flow of information to Congress and that the chief actuary had no authority to disclose information to Congress independently. The report added that Scully, had he still worked for the government, might have faced disciplinary action for a possible violation of the department’s standards of ethical conduct. He had resigned seven months earlier, so the finding led nowhere. The report also recorded Scully’s own version of events. He denied that he had threatened Foster’s job, while acknowledging that he had instructed the actuary to withhold the figures. On the central question the investigators sided with Foster’s account, finding that the threat had been made. Reaction split along party lines. Pete Stark, the ranking Democrat on the Ways and Means health subcommittee, said the episode amounted to the administration investigating itself and finding it had done nothing wrong. Nancy Johnson, the Connecticut Republican who chaired that subcommittee, called the report a major embarrassment to Democrats because it rejected their partisan motives.

The Government Accountability Office answered the senators two months later, and its legal conclusion pointed the other way. In an opinion dated September 7, 2004, numbered B-302911 and signed by General Counsel Anthony H. Gamboa, GAO held that Scully’s order fell squarely within section 618 of the Consolidated Appropriations Act of 2004. That provision, a government-wide rider, barred the use of appropriated funds to pay the salary of any federal official who prevented another federal employee from communicating with Congress. GAO wrote that the legislative history showed Congress had aimed the provision at exactly this kind of bar on employee communication, and it concluded that the department’s appropriation had therefore been unavailable to pay Scully’s salary. GAO also rejected constitutional objections raised in the record by HHS associate general counsel Katherine M. Drews and by Jack L. Goldsmith III of the Justice Department’s Office of Legal Counsel, holding that the rider did not violate the separation of powers. GAO’s reading of the legislative history was blunt. It called Scully’s bar on Foster’s communications a prime example of what Congress had meant to prohibit when it wrote the rider, language aimed at keeping frontline federal employees free to give lawmakers programmatic information. The opinion carried no enforcement mechanism of its own, and Scully was long gone from the department. Its significance was declaratory: the investigative arm of Congress had put in writing that the episode was not merely sharp politics but a use of appropriated funds that the appropriations law did not allow. The opinion had been requested by letter dated March 18, 2004, from eighteen senators, and it named Scully and Foster directly in framing the question presented. The two formal reviews thus documented the same events and split on the law. The inspector general found no criminal violation. GAO found a violation of an appropriations rider. On the facts, both agreed: the instruction was given, the threat was made, and Congress voted without the higher estimate.

The two principals went in opposite directions after 2004. Scully left the department in December 2003 and moved into lobbying for health care companies, a transition the Times noted when the inspector general’s report appeared. Foster stayed. He had told his colleagues in June 2003 that he might have to resign in protest, but after the HHS lawyer’s advice and his own deliberation he remained in the chief actuary’s chair, the same post from which he had modeled the drug benefit’s cost in the first place. The asymmetry is part of the episode’s meaning. The official who gave the instruction faced no sanction, since the inspector general found no criminal violation and the GAO opinion, though pointed, carried no enforcement power. The official who received it kept his job and his reputation for careful, nonpartisan arithmetic.

The second financial history begins where the first one ends, and it runs in the opposite direction from what the pre-vote fight would lead a reader to expect. Once Medicare Part D began operating in January 2006, the program’s costs kept undershooting the forecasts. The undershoot showed up in premiums first, in enrollment next, and in the ten year budget projections after that.

Premiums told the story earliest. In March 2005, administration officials projected that the average monthly premium for standard drug coverage would be $37 in 2006 and $41 in 2007. Competitive bidding among the private plans offering the benefit undercut those projections almost immediately. In August 2006, CMS Administrator Mark McClellan announced that the actual average premium for 2006 had come in at about $24, roughly 40 percent below the original projection, because insurers had submitted competitive bids and beneficiaries had favored lower priced plans. The pattern held for the rest of the decade. CMS’s published weighted averages were $24 in 2007, $25 in 2008, $28 in 2009, and $30 in 2010. On August 18, 2010, the agency announced that the average premium would hold at $30 for 2011. The Kaiser Family Foundation reported in November 2009 that nearly 27 million beneficiaries were enrolled in Medicare Part D plans, two thirds of them in standalone prescription drug plans, with 1,576 such plans offered nationwide for 2010, down from a peak of 1,875 in 2007 but well above the 1,429 offered in the first year.

The ten year spending projections fell in the same direction. After the program’s third open enrollment season, federal officials announced that the new ten year estimate for the drug benefit had dropped to $243.7 billion. That figure was $117 billion below the estimate made the previous summer and 38.5 percent below the original ten year projection for the 2004 to 2013 window that the 2003 debate had used. The officials credited three developments: a slowdown in the growth of drug spending, lower estimates of plan spending, and higher rebates from manufacturers. By that point 25.4 million beneficiaries were enrolled in the program’s plans, and about 90 percent of the nation’s 44 million Medicare beneficiaries had drug coverage from Medicare or another source, including employer plans receiving the retiree drug subsidy and programs such as TRICARE and the Federal Employees Health Benefits Program.

The 2010 Medicare trustees report extended the trend into the official long range outlook. CMS reported that projected costs for the drug benefit were slightly lower than in the previous year’s report, reflecting lower than expected costs in 2008 and 2009, partly offset by the added cost of phasing out the benefit’s coverage gap under the 2010 health reform law. The trustees were not attributing the decline to any single policy lever. The pharmaceutical market itself was doing much of the work: major brand name medicines were losing patent protection, fewer expensive new blockbusters were reaching the market, and generic substitution was rising across the classes of drugs that Medicare beneficiaries used most. The cost record, in other words, did not vindicate anyone’s pre-vote arithmetic. It reflected a drug market that turned out to be cheaper than the models had assumed and a bidding system that pushed plan prices below the forecasts.

Each side of the later debate has tried to use one ledger to close the other, and the record resists both attempts. Defenders of the process have pointed to the low cost record as proof that the pre-vote fight changed nothing of substance. But accountability does not work that way. The question the controversy raises is whether lawmakers voted with the information the executive branch possessed, and on that question the documented answer is no, regardless of what the program later cost. Critics have pointed to the suppression as proof that the benefit was a fiscal catastrophe in the making. The spending record answers that claim just as firmly. Premiums landed far below the projections, the ten year estimates were marked down by more than a third, and the trustees reported the decline in their own annual accounts. The episode and the expenditures are two separate findings, and collapsing them into a single verdict distorts both.

Set side by side, the two histories answer different questions, and this section refuses to let one answer stand in for the other. The pre-vote controversy asks whether Congress voted with full information. The documented answer is no. Richard Foster’s higher figures were kept from lawmakers at Thomas Scully’s direction, a fact the inspector general confirmed and the Government Accountability Office gave legal weight in a published opinion. The post-enactment record asks whether the benefit then broke the forecasts. The documented answer is also no. Premiums, enrollment-adjusted spending, and the ten year projections all landed below the early estimates, and the trustees said so in their own reports. The low cost record does not retroactively justify the suppression, and the suppression does not erase the savings. An honest account of this accountability episode keeps both ledgers open and declines to let either one balance the other.

From Passage to Launch: The Senate, the Discount Cards, and the Forty-Five Percent Trigger

The conference report that survived the House still had to survive the Senate, and the Senate nearly stopped it. Senator Edward Kennedy led a filibuster against the conference report, arguing that the drug benefit was needlessly complex and that the noninterference clause surrendered the government’s bargaining power to the pharmaceutical industry. Supporters, led in the chamber by the bill’s Senate authors, answered that a flawed drug benefit that could pass was worth more than a perfect one that could not, and that competition among private plans would hold down costs better than administered pricing. The Senate invoked cloture by 70 votes to 29, ending the filibuster, and then adopted the conference report by 54 votes to 44 on November 25, 2003. President George W. Bush signed the measure on December 8, 2003, as Public Law 108-173. The closeness of both votes, and the procedural force required to close debate in the Senate, marked the act as one of the most bitterly contested domestic statutes of its era even before any beneficiary enrolled.

Enactment did not mean immediate coverage. The Part D benefit took effect on January 1, 2006, leaving two full years during which the elderly had a new statute and no new drug coverage. Congress built a bridge for the interval. The act authorized Medicare-approved prescription drug discount cards, available through private sponsors in 2004 and 2005, which offered negotiated discounts at pharmacies to enrolled beneficiaries. Alongside the cards, the act provided transitional assistance for low-income beneficiaries to help with drug costs during the waiting period. The bridge was modest by design. Discount cards carried no government subsidy of the drug price itself, and the transitional assistance reached only a defined low-income population. But the bridge served a political purpose beyond its dollars. It gave the administration and the plans a live enrollment and card-issuance system to operate, test, and debug before the full benefit launched, and it gave beneficiaries a first, limited experience of choosing among private drug plan sponsors, which was the core consumer skill the 2006 benefit would demand.

The act also built in a fiscal tripwire that had nothing to do with drugs. Section 801 of the Modernization Act requires the Medicare trustees, beginning with their 2005 report, to determine each year whether general revenue funding is expected to exceed 45 percent of total Medicare outlays in the current fiscal year or any of the following six. General revenue funding means total Medicare outlays minus dedicated financing sources, so the test measures how much of the program the payroll tax and premiums do not cover. Two consecutive affirmative determinations constitute a Medicare funding warning, and once a warning issues, sections 802 through 804 require the President to submit legislation responding to the warning within fifteen days of the next annual budget request, with expedited congressional procedures for considering it. The statute requires no enactment and creates no automatic spending restraint, and either chamber may alter the procedures by majority vote. The trustees made their first determination of excess general revenue funding in the 2006 report, which produced the first funding warning in 2007, and the warnings repeated each year through the end of the decade. Supporters of the trigger described it as an early alarm that would force recurring attention to the program’s financing. Analysts skeptical of the device, including Joseph Antos of the American Enterprise Institute, noted that the 45 percent mark had no actuarial basis and could as easily have been set higher or lower. The trigger thus joined the act’s other design choices as a mechanism whose consequences would play out long after the 2003 debate ended, through procedures the enacting Congress wrote but could not control.

The administrative build-out filled the rest of the interval. During 2005 the Centers for Medicare and Medicaid Services approved plan bids, organized the country into the 34 prescription drug plan regions the statute contemplated, and published the standard benefit parameters for the 2006 launch year. Plans that wanted to offer the benefit had to demonstrate adequate pharmacy networks, compliant formularies, and the capacity to administer enrollment and appeals, all under CMS review. The agency also ran the first open enrollment season in late 2005, during which beneficiaries chose among the competing plans for coverage beginning January 1, 2006. The two-year delay that critics had attacked as a political convenience thus functioned as an implementation runway, and the program that launched on the first day of 2006 arrived through the private-plan machinery the statute had specified rather than through any government-run fallback.

Congress gave itself a long runway, and then spent it building a marketplace. The Medicare Modernization Act was signed in December 2003, but the prescription drug benefit it created did not take effect until January 1, 2006. The two year interval was not idle time. It was the period in which the federal government wrote the rules for a benefit it would not itself administer, private insurers built the plans through which the benefit would flow, and tens of millions of beneficiaries learned that their drug coverage would henceforth depend on a choice among competing private offerings. Implementation, in other words, was where the act’s central design choice, delegation, stopped being a legislative compromise and became an operating reality.

The Centers for Medicare and Medicaid Services spent 2004 drafting proposed regulations and issued the final Part D rule in January 2005. The regulatory philosophy was stated plainly in the preamble to those rules. A Kaiser Family Foundation report on the federal role, written by Toby Edelman of the Center for Medicare Advocacy, quoted CMS directly: “The goal of the MMA was to encourage private sector organizations who meet the law’s requirements to offer a range of Part D plan options for Medicare beneficiaries by providing flexibility in plan design and management.” Plan flexibility was not a side effect of the rollout. It was the stated method of the rollout. For a fuller map of that rulemaking process, see our CMS Medicare rulemaking guide.

Before the permanent benefit arrived, Congress built a bridge. The act created a Medicare approved prescription drug discount card program with transitional assistance for low income beneficiaries, operating through 2004 and 2005, so that seniors would see some help with drug costs while the permanent machinery was assembled. By statute, the discount card program stopped applying to drugs dispensed after December 31, 2005, and on the next morning the permanent benefit began. The handoff was designed to be seamless. Whether it felt seamless to beneficiaries is a separate question, and January 2006 gave it a hard test.

The delivery architecture Congress chose had two doors and no public option behind either one. Beneficiaries in traditional fee for service Medicare could enroll in a stand alone prescription drug plan, a PDP, offered by a private insurer bearing part of the financial risk for drug costs. Beneficiaries in the renamed Medicare Advantage program could get drug coverage integrated with their medical benefits through a Medicare Advantage prescription drug plan. Every plan was a private contractor. CMS did not sell insurance, did not set a formulary, and did not negotiate prices with manufacturers; the statute’s noninterference clause barred the Secretary from doing so. Plans submitted annual bids to CMS, and CMS used those bids to compute a national average premium that anchored the government’s subsidy and the benchmark beneath which low income subsidy beneficiaries were automatically enrolled. Congress authorized government fallback plans for any region where private bids failed to materialize. The fallback was never the point. The point was that private bids would materialize, and the law was written to make that outcome overwhelmingly likely.

The money followed the bids. Each plan sponsor submitted an annual bid representing its expected cost of delivering the standard benefit, and CMS blended those bids into a national average premium that set the government’s per enrollee subsidy and the benchmark for low income auto enrollment. Sponsors that bid below the benchmark could be assigned low income subsidy beneficiaries automatically, which made the benchmark a competitive prize worth chasing and gave the largest national insurers a strong incentive to bid aggressively for market share. The government therefore did not set prices for the benefit; it set the formula by which private bids became public subsidies. That distinction is the delegated model in miniature. Washington wrote the equation, and the market filled in the numbers.

They did materialize, in numbers that startled even the program’s defenders. In 2006, the first year of the benefit, 1,429 stand alone prescription drug plans entered the market, offered by 65 different firms, though 1,222 of those plans came from just 14 national or near national organizations, the major insurers and pharmacy benefit managers that dominated the new business. Choice was abundant and uneven: beneficiaries in Alaska could pick among 27 PDPs, while those in Pennsylvania and West Virginia faced 52. Medicare Advantage plans, required to include a drug benefit in their offerings, added an average of 2.4 regional managed care plans to the mix in each area. The average PDP monthly premium came in at 32 dollars, within a range that stretched from 2 dollars to 100 dollars, and the majority of plans discarded the standard benefit’s structure in favor of tiered copayments. Few plans paid for drugs once members reached the coverage gap, the doughnut hole in which beneficiaries owed the full cost of their medications until catastrophic coverage began. The market that Congress had summoned behaved like a market: wide variation, aggressive pricing at the low end, and benefit designs tuned to attract healthy enrollees.

Enrollment followed the script Congress wrote for it, which was a script about persuasion rather than compulsion. Part D was voluntary, with a late enrollment penalty for those who delayed without creditable coverage from another source, and the initial enrollment window ran through May 15, 2006. The Department of Health and Human Services reported that by June 11, 2006, 22.5 million Medicare beneficiaries had prescription drug coverage from a Medicare Part D plan: 10.4 million had enrolled voluntarily, 6.6 million were dual eligibles auto enrolled into PDPs, and 5.5 million had enrolled through Medicare Advantage plans. Another 10.4 million retirees held creditable drug coverage from employer plans, and an estimated 5.4 million more had creditable coverage from other sources such as the Department of Veterans Affairs. The administration’s headline figure was that about 90 percent of Medicare’s 43 million beneficiaries had creditable prescription drug coverage within six months of launch. Testimony before the Senate Special Committee on Aging later put the residual in starker terms: roughly 4.5 million beneficiaries, about one in ten, had no source of drug coverage at all. For a brand new voluntary entitlement built from scratch in two years, the enrollment numbers were a genuine administrative achievement. They were also the achievement of the delegated model working exactly as designed, with private plans doing the selling and the government doing the counting.

The voluntariness was engineered, not casual. Beneficiaries who delayed enrollment without creditable drug coverage from another source faced a permanent late enrollment penalty that raised their Part D premium by one percent for each month they went without coverage, a surcharge that followed them for as long as they held the benefit. The penalty was the enrollment engine of a voluntary program: without a mandate, Congress needed a financial reason to pull healthy beneficiaries in early, because a drug benefit that attracted only the sick would see its premiums spiral. Combined with the auto enrollment of dual eligibles and the employer subsidy that paid sponsors of retiree plans to keep their drug coverage creditable, the penalty turned the 2006 enrollment drive into a multi channel capture system. The design did not merely offer a benefit; it built the incentives that would populate it.

The launch itself was rougher than the enrollment totals suggest, and the roughness also traced back to the design. On January 1, 2006, about six million dual eligibles, seniors poor enough to qualify for both Medicare and Medicaid, were moved off Medicaid drug coverage and into private Part D plans in a single day. Medicaid had paid for their prescriptions for years; now a PDP chosen by an algorithm would do it. Pharmacy databases did not always know which plan a beneficiary belonged to. Formularies did not always cover the drugs a beneficiary had been taking. Copayments that were supposed to be nominal sometimes printed out at 80 or 90 dollars. About half the states plus the District of Columbia took emergency action to keep paying for prescriptions under state financing while the federal program’s errors were fixed. Health and Human Services Secretary Michael Leavitt ordered plans to cover any drug, formulary or not, for 30 days during the transition, capped cost sharing at 5 dollars for the affected seniors, and authorized enrollment in a default plan at the pharmacy counter as a last resort. Joseph Antos, the health economist at the American Enterprise Institute, called the chaos one of the “fully expected transition problems” of the rollout and predicted it would take weeks or months rather than days to resolve. He was right about the cause and roughly right about the timeline. The transition problems were fully expected because the design made them predictable: shifting six million of the poorest and sickest beneficiaries from a government run benefit to private plans overnight was always going to produce database mismatches and coverage gaps, and the emergency measures were patches on a design choice, not accidents of execution.

The friction was not confined to January. Federal and state officials spent 2006 fixing the Plan Finder tool, staffing up call centers that had left beneficiaries and pharmacists on hold, and counseling confused enrollees through a benefit whose complexity critics had warned about from the start. Reports that some sponsors violated marketing rules and did too little to prevent fraud gave the program’s opponents fresh material, and the coverage gap hit more enrollees than early projections had suggested, with about a quarter of Part D enrollees reaching the doughnut hole in 2007 according to Kaiser Family Foundation analysis. CMS responded the way a contract manager responds: with enrollment fixes, marketing enforcement, and rule clarifications, not with a redesign of the benefit. The problems were managed inside the delegated model, which is another way of saying the model was never on trial. Only its execution was.

That design choice, delegation, is the through line of this article and of the benefit’s early history. Congress created the entitlement, defined its standard benefit, funded the subsidies, and then contracted out nearly everything else: plan administration to private insurers, price negotiation to those insurers and their pharmacy benefit managers, and even the enrollment of the poorest beneficiaries to an auto assignment algorithm. CMS retained genuine powers, approving plan bids, policing marketing abuses, running the enrollment systems, and paying the subsidies, but the government never touched a prescription. The beneficiary’s relationship was with a private plan. The plan’s relationship was with manufacturers and pharmacies. The government’s relationship was with the plan. Every layer of that chain was a deliberate answer to the political problem of 2003: a drug benefit that the federal government administered directly could not assemble a majority, so Congress built one that the private sector would administer under federal contract and called the result a modernization.

How did the 2003 act pay the private plans that run Part D?

It renamed Medicare+Choice as Medicare Advantage, required those plans to carry prescription drug coverage, and created stand alone prescription drug plans as a new plan type. Above all, it made private plans the only door into the new drug benefit, turning them from one alternative to traditional Medicare into the program’s delivery system.

The renaming mattered less than the rewiring. Before 2003, private plans under Medicare+Choice were an alternative that beneficiaries could select instead of traditional Medicare, and the government’s posture toward them was ambivalent, expanding and contracting payments as political winds shifted. The 2003 act made private plans structurally indispensable. A beneficiary who wanted the new drug entitlement could not get it from the government directly; the only path ran through a PDP or a Medicare Advantage drug plan. The act also deepened the financial integration: plans that bid below the national average premium benchmark could enroll low income subsidy beneficiaries automatically, which turned the benchmark into a competitive prize and gave large national firms a powerful incentive to bid aggressively for market share. The fallback plan provision, authorizing government run coverage where private bids were absent, functioned less as a safety net than as a signal. Its existence proved Congress understood that a purely private market might leave gaps; its irrelevance in practice proved how thoroughly the subsidies drew private bidders in. What changed, in short, was the role of the private plan from option to infrastructure. Medicare’s private plans stopped being a choice within the program and became the program, at least for prescription drugs.

The financial plumbing reinforced that transformation. Because the noninterference clause barred the Secretary from negotiating prices or establishing a formulary, price discipline had to come from the plans themselves, each negotiating with manufacturers through pharmacy benefit managers and each assembling a formulary to steer enrollees toward cheaper drugs. CMS reviewed bids for actuarial soundness but did not second guess the prices plans had extracted. This was delegation at its purest: the government set the subsidy formula and let private negotiators determine what the subsidized dollars would buy. Defenders of the model argued that dozens of competing negotiators would outperform a single government buyer, since each plan had to win enrollees with some combination of low premiums and broad formularies. Critics argued that fragmenting purchasing power across scores of plans surrendered exactly the leverage that made government drug purchasing cheap elsewhere, and that the plans’ real customers were their shareholders, not their enrollees.

Verdict

The delegated benefit model deserves to be judged on the terms its architects chose, and on those terms its strongest case is formidable. Joseph Antos and his American Enterprise Institute colleagues, writing in a 2007 assessment of the program, declared that “Medicare Part D is succeeding beyond expectations in terms of beneficiary satisfaction and costs,” and argued that the competitive model brought private competition into play “to offer seniors lower prices and greater choice,” which they called “far better than a government-controlled system.” The numbers available through the end of the decade supported the cost half of that claim in the terms that matter to budget scorekeepers: Part D premiums and federal outlays came in below the Congressional Budget Office projections that had accompanied the legislation, enrollment exceeded expectations, and beneficiary satisfaction surveys showed majorities content with their coverage. Antos’s case is that competition among private plans, each bidding for enrollees and negotiating with manufacturers, delivered a benefit more cheaply than the government actuaries predicted, and that the prediction error is the truest measure of the design’s success. If the test of an entitlement is whether it covers the people it promised to cover without breaking the budget projections made for it, Part D passed, and delegation was the mechanism of the pass.

The strongest criticism of the model is equally rooted in the design, and it has a named author with a number attached. Representative Henry Waxman of California, as chairman of the House Oversight and Government Reform Committee, charged that the same delegation that produced low premiums produced a windfall for drug manufacturers. His committee’s investigation found that because Part D plans paid higher prices than Medicaid had paid for the drugs used by dual eligibles, manufacturers collected an estimated 3.7 billion dollars in additional revenue in 2006 and 2007 alone on those beneficiaries. “In effect,” Waxman said at a July 2008 hearing, “Medicare Part D has given the major drug companies a taxpayers-funded windfall worth billions of dollars.” The mechanism of the windfall was the delegation itself: dual eligibles had previously received drugs under Medicaid, where federal law required manufacturers to pay substantial rebates, and their transfer to private Part D plans moved them into a system where no such rebates were required and where the noninterference clause barred the government from demanding them. Waxman’s criticism does not dispute that beneficiaries got coverage. It disputes that the price of that coverage was fair, and it locates the overpayment in the precise design feature the architects celebrated, the decision to let private plans rather than the government do the buying.

That selection happened through procedure as much as persuasion. The legislative path available in 2003 admitted only designs that preserved private insurers, pharmacy benefit managers, and manufacturers as the benefit’s operating core, because any design that displaced them could not hold its coalition together. The noninterference clause, the voluntary structure, the bid based subsidy, and the fallback provision that nobody needed were all artifacts of that constraint: each one answered an objection that a directly administered benefit could not survive. This is what the series thesis means when it says procedure determines which design survives. The enrollment success and the manufacturer windfall were both downstream of a choice that was itself downstream of what the legislative process would tolerate.

The verdict this article reaches follows from the series thesis that design choices determine a program’s behavior and procedure determines which design survives. Part D behaves the way it does because Congress designed it that way, not because private markets are inherently efficient or inherently corrupt. The low premiums and the manufacturer windfall are not contradictory findings; they are two outputs of the same design. Competitive bidding among plans pushed premiums down, because plans compete on the premium beneficiaries see. Fragmented private negotiation let manufacturer prices for dual eligibles rise, because no single negotiator held the leverage the government had held under Medicaid, and the noninterference clause made sure no government negotiator would appear. The confusion of January 2006, the abundance of plan choice, the tiered formularies, the doughnut hole that few plans filled: each was the delegated model expressing its own logic. None of it was accidental, and none of it was the product of poor execution alone. A program built on private plans behaves like private plans.

And the delegated design survived because it was the design the procedure of 2003 could produce. A directly administered federal drug benefit, with government negotiation and a government formulary, could not assemble a majority in the Congress that wrote this law; the private plan model could, because it preserved a role for insurers, pharmacy benefit managers, and manufacturers, and because it let the benefit’s supporters describe the result as competition rather than bureaucracy. That is not a defense of the design and not an indictment of it. It is the mechanism the thesis describes: procedure selects among designs, and then the surviving design determines behavior for decades. Judging Part D therefore requires judging delegation on its own terms, weighing Antos’s lower than projected costs against Waxman’s documented windfall, and recognizing both as the predictable behavior of the model Congress chose. The benefit that took effect on January 1, 2006, was never a government program with private contractors attached. It was a private market with a government subsidy attached, and everything that followed, the enrollment triumphs and the transition chaos and the arguments about who profited, followed from that single design choice.

Frequently Asked Questions

Q: What is Medicare Part D?

Medicare Part D is the voluntary outpatient prescription drug benefit that the Medicare Prescription Drug, Improvement, and Modernization Act of 2003 added to Medicare, with coverage beginning in January 2006. Unlike Parts A and B, which pay health care providers directly, Part D operates through private insurers: beneficiaries enroll either in stand-alone prescription drug plans or in Medicare Advantage plans that include drug coverage. Enrollees pay a monthly premium, an annual deductible, and cost sharing for prescriptions, while the federal government pays the plans a direct subsidy, reinsurance payments for catastrophic costs, and extra payments covering low-income enrollees. Enrollment is voluntary, but beneficiaries who put off signing up without other qualifying drug coverage face a permanent late-enrollment penalty added to their premiums. When the initial sign-up period closed in May 2006, 22.5 million beneficiaries had enrolled in Part D plans.

Q: How did the Medicare Modernization Act pass the House in 2003?

The House approved the conference agreement for H.R. 1 on November 22, 2003, by a vote of 220 to 215. The roll call began around 3 a.m. on a Saturday morning and ran for nearly three hours, far beyond the usual 15-minute voting period, ending at about 5:55 a.m. House Republican leaders, including Speaker Dennis Hastert, Majority Leader Tom DeLay, and Majority Whip Roy Blunt, spent the extended session working the floor to persuade members to change their votes and assemble a majority. Democratic critics described the marathon roll call as a violation of House norms. Representative Nick Smith of Michigan later said unnamed colleagues had offered campaign support for his son, who was running to succeed him, in exchange for his vote; he voted against the bill, and the House ethics committee examined the episode. President George W. Bush signed the measure into law on December 8, 2003, as Public Law 108-173.

Q: Why was the 2003 Medicare vote held open for three hours?

The roll call was held open because the bill lacked the votes to pass when voting began. House Republican leaders kept the electronic voting board open for nearly three hours, from about 3 a.m. to 5:55 a.m. on November 22, 2003, while they worked to flip enough members to reach 218 votes. The final tally was 220 to 215 in favor. Democratic critics called the extended vote a breach of House practice without modern precedent, arguing that the normal 15-minute rule existed precisely to prevent leaders from pressuring members after the outcome was known. Supporters of the bill, including the Republican leadership, treated the long roll call as a legitimate exercise of the majority’s power to complete a close and consequential vote. The episode also drew attention because Representative Nick Smith said colleagues had linked campaign help for his son to a vote switch, an allegation that prompted an ethics inquiry.

Q: What is the Medicare donut hole?

The donut hole is the coverage gap built into the standard Medicare Part D benefit design. Under the standard design, a beneficiary paid a deductible and then shared drug costs with the plan up to an initial coverage limit. Once total drug spending by the beneficiary and the plan reached that limit, coverage stopped and the beneficiary paid the full cost of prescriptions until personal out-of-pocket spending reached a higher threshold. At that point catastrophic coverage began and the plan paid nearly all further costs for the year. The gap was designed to hold down the program’s cost and encourage price sensitivity, but it meant that beneficiaries with moderate drug needs faced the full price of their medicines for part of the year. In the first year of the program, few plans covered drugs inside the gap, and only about 4 percent of enrollees had plans covering both brand-name and generic drugs there.

Q: Why can Medicare not negotiate drug prices?

The 2003 law barred the government from negotiating prices directly and assigned that job to the private prescription drug plans and the pharmacy benefit managers they hire. Supporters of this design, including the Bush administration, argued that competition among many plans for enrollees would hold drug prices down more effectively than a single government negotiator, and that federal price setting would interfere with the private insurance market the law created. Critics, including Democratic lawmakers such as Representative Charles Rangel, argued that the world’s largest drug purchaser should use its buying power directly and that the ban protected pharmaceutical industry profits at beneficiaries’ expense. Defenders of the provision replied that plans already extracted rebates and discounts from manufacturers. Congress therefore left price negotiation to market participants, and the question of whether the government should negotiate remained one of the most contested features of the program.

Q: What is the Medicare noninterference clause?

The noninterference clause is section 1860D-11(i) of the Social Security Act, added by the 2003 Medicare law, which provides that the Secretary of Health and Human Services “may not interfere with the negotiations between drug manufacturers and pharmacies and PDP sponsors, and may not require a particular formulary or institute a price structure for the reimbursement of covered part D drugs.” In practice the clause gives the federal government no direct role in setting or negotiating the prices Medicare pays for outpatient prescription drugs. Instead, each private prescription drug plan negotiates its own discounts and rebates with manufacturers and pharmacies. Supporters of the clause, including the Bush administration, said it preserved a competitive private market for drug coverage. Critics, including Democratic lawmakers, said it needlessly tied the government’s hands and inflated program costs. The clause drew a sharp line between Part D and programs such as Medicaid, which uses mandatory manufacturer rebates.

Q: How much did Medicare Part D actually cost?

Part D cost substantially less than the estimates that drove the 2003 debate. The Congressional Budget Office had estimated the benefit would cost $395 billion over its first ten years, while the administration’s own actuary projected $534 billion. After the program began, actual spending came in well below both figures. The Medicare trustees reported that the federal cost of the drug benefit in 2009 was $60.8 billion, far below the roughly $111 billion annual cost projected at enactment. The Congressional Budget Office later estimated that Part D would cost about $136 billion less over the 2007-2013 period than originally expected. Analysts at the Congressional Budget Office and the Medicare trustees attributed the gap mainly to lower-than-expected enrollment, which explained roughly half the difference, and to a system-wide slowdown in prescription drug spending growth driven by greater use of generic drugs. Supporters of the law’s private-plan design pointed to the lower costs as evidence that competition worked; skeptics called the difference a forecasting error rather than a policy victory.

Q: What happened to the Medicare actuary who questioned the Part D estimate?

Richard S. Foster, the chief actuary for the Centers for Medicare and Medicaid Services, calculated that the drug benefit would cost about $534 billion over ten years, well above the $395 billion estimate from the Congressional Budget Office that the Bush administration was citing. While Congress was still considering the bill, CMS administrator Thomas A. Scully told Foster not to share his higher estimate with lawmakers and, according to Foster, threatened to fire him if he disclosed it. Foster consulted a department lawyer, who advised that Scully had the authority to block the disclosure, so Foster stayed in his post without releasing the numbers. The higher estimate became public only after the bill was signed. A 2004 investigation by the department’s acting principal deputy inspector general confirmed that Scully had threatened sanctions against Foster, though it found no criminal violation. Scully resigned in December 2003 and became a health care lobbyist; Foster kept his job as chief actuary.

Q: Why did the 2003 act replace Medicare+Choice with Medicare Advantage?

The 2003 law renamed and restructured the private-plan part of Medicare because the existing Medicare+Choice program, created in 1997, was losing participating plans and enrollees as payment rates fell short of plan costs. The act replaced it with Medicare Advantage and changed plan payments to a system of competitive bids measured against regional benchmarks, so that plans bidding below the benchmark could offer extra benefits or lower premiums to attract enrollees. It also created a stabilization fund to encourage plans to enter and stay in the program. Supporters, including the Bush administration, argued that managed competition would deliver Medicare benefits more efficiently than administered prices. Critics argued that the new payment system subsidized private plans at costs above traditional Medicare. The law further authorized a six-year comparative cost adjustment pilot in a limited number of metropolitan areas to test having traditional Medicare compete directly against private plans.

Q: How do prescription drug plans bid for Medicare Part D contracts?

Each year, every organization offering a Part D plan submits a bid to the Centers for Medicare and Medicaid Services stating what it will cost to provide the standard drug benefit to an average enrollee. The agency computes a national average from those bids and sets a benchmark, then pays each plan a direct subsidy equal to a fixed share of that benchmark. The beneficiary’s monthly premium is roughly the difference between the plan’s own bid and the government subsidy: plans that bid below the average can charge lower premiums and attract more enrollees, while plans bidding above the average must charge higher premiums. The design was meant to make plans compete on efficiency rather than on benefit cuts, since all plans must cover at least the standard benefit. The government also pays reinsurance covering most costs above the catastrophic threshold and extra payments for low-income enrollees.

Q: What help did low-income beneficiaries get under Part D?

The law created a low-income subsidy, often called Extra Help, that paid most or all of the premiums, deductibles, and cost sharing for beneficiaries with limited income and assets. The subsidy had full and partial levels depending on income relative to the poverty line and on countable resources. About six million people who were eligible for both Medicare and Medicaid were automatically enrolled in prescription drug plans and received the full subsidy without applying, since their Medicaid drug coverage ended when Part D began. Other qualifying beneficiaries applied through the Social Security Administration. In 2006 about 8.3 million beneficiaries received the subsidy, most of them dual eligibles. The subsidy was one of the largest federal commitments in the law, because low-income beneficiaries use more prescriptions on average and would otherwise have faced the full cost sharing of the standard benefit.

Q: What covered Medicare drug costs between 2004 and 2006?

Because the full Part D benefit did not begin until January 2006, the 2003 law created a temporary Medicare-approved drug discount card program for 2004 and 2005. Beneficiaries could buy a card from a private sponsor and use it to obtain discounted prices at participating pharmacies. Low-income beneficiaries who enrolled also received transitional assistance, a federal credit applied toward their drug purchases during the two interim years. The card program was voluntary and separate from the permanent Part D structure, and it ended when Part D coverage started. Enrollment in the discount cards fell short of expectations, and the program functioned mainly as a bridge that familiarized beneficiaries and pharmacists with the idea of Medicare drug coverage before the permanent benefit launched. The experience informed the outreach campaign that accompanied the Part D enrollment period in late 2005 and early 2006.

Q: How did the 2003 act change Part B premiums for higher-income beneficiaries?

Beginning in 2007, the law required higher-income beneficiaries to pay larger Part B premiums than the standard amount, a policy known as income-related premium adjustment. The higher premiums were phased in over five years, so that by the end of the phase-in period the highest-income beneficiaries paid substantially more than the standard premium while the standard premium continued to cover about one quarter of Part B costs for everyone else. The act also raised the Part B annual deductible to $110 in 2005 and provided for annual indexing starting in 2006. Supporters of the change argued that asking affluent beneficiaries to pay more was a fair way to slow the growth of general-revenue financing for Medicare. Critics argued that the thresholds penalized middle-class retirees in high-cost areas and moved Medicare away from its tradition of uniform premiums for a uniform benefit.

Q: What is the 45 percent general revenue trigger in the 2003 law?

Section 801 of the 2003 act created a fiscal tripwire known as the Medicare funding warning. If the Medicare trustees projected in two consecutive annual reports that general revenue funding would exceed 45 percent of total Medicare outlays at any point in the coming seven years, a funding warning was triggered. General revenue funding meant total Medicare spending minus dedicated financing sources such as payroll taxes and beneficiary premiums. Once triggered, the law required the President to submit legislation responding to the warning and required Congress to consider it under expedited procedures. Supporters of the provision said it forced elected officials to confront Medicare’s long-term financing rather than ignore it. Critics said the 45 percent figure was arbitrary and that the trigger was designed to build political pressure for benefit cuts or privatization rather than to improve the program’s finances.

Q: Why did the 2003 law create Health Savings Accounts?

Alongside its Medicare provisions, the 2003 act created Health Savings Accounts, tax-advantaged accounts that individuals could pair with high-deductible health plans to pay for medical expenses. Contributions were deductible, account balances grew tax free, and withdrawals for qualified medical expenses were not taxed. Supporters of the accounts, including the Bush administration, argued that giving consumers control over a pool of their own health care dollars would make them more cost-conscious shoppers and slow the growth of health spending. Critics argued that the tax benefits flowed mainly to higher-income households that could afford to fund the accounts, while people with low incomes or chronic conditions gained little. The accounts applied to the under-65 insurance market rather than to Medicare itself, but their inclusion showed that the 2003 law aimed at reshaping health care financing well beyond the prescription drug benefit.

Q: Is there a penalty for enrolling in Medicare Part D late?

Beneficiaries who go without creditable prescription drug coverage for a continuous period after first becoming eligible for Part D face a permanent late-enrollment penalty. The penalty equals one percent of the national base beneficiary premium for each full month without coverage, and it is added to the monthly premium for as long as the person stays enrolled. For example, someone who waited two years without other drug coverage would pay a premium roughly 24 percent higher than the base amount. The law defined creditable coverage to include employer retiree plans and certain other sources at least as generous as the standard Part D benefit, so beneficiaries with such coverage could delay enrollment without penalty. The penalty was designed to discourage healthy beneficiaries from waiting until they needed expensive drugs, which would have left the risk pool sicker and driven premiums higher for everyone.

Q: How did dual eligibles get drug coverage under Part D?

About six million beneficiaries who qualified for both Medicare and Medicaid, known as dual eligibles, were automatically enrolled in stand-alone prescription drug plans just before Part D began. Their Medicaid prescription drug coverage ended on December 31, 2005, and Medicare coverage started the next day, so the auto-enrollment was designed to prevent any gap in their access to medicines. The Centers for Medicare and Medicaid Services assigned each dual eligible at random to a qualifying plan in their region and notified them of the assignment, with the right to switch plans. The federal government paid the full low-income subsidy for these enrollees, covering their premiums and nearly all cost sharing. States, which had previously paid for these beneficiaries’ drugs through Medicaid, were required to pay the federal government a phased-down share of those costs, a payment known as the clawback.

Q: How far apart were the official cost estimates for Part D?

The two official estimates differed by well over $100 billion. The Congressional Budget Office estimated that the drug benefit would cost $395 billion over its first ten years, a figure the Bush administration cited while selling the bill to Congress. The administration’s own chief actuary, Richard Foster, estimated the cost at about $534 billion over the same period, roughly $139 billion more, and his internal estimates had ranged between $500 billion and $600 billion while the bill was being written. The gap mattered politically because President Bush had pledged to spend no more than $400 billion, and a group of about thirteen conservative House Republicans had promised to vote against any bill exceeding that ceiling. The administration released Foster’s higher estimate only after the bill was signed, prompting Democratic charges that Congress had voted without knowing the true price tag.

Q: How many drug plans competed in the first year of Part D?

In 2006, the first year of the benefit, 65 organizations offered 1,429 stand-alone prescription drug plans nationwide. The number of choices varied sharply by region, from 27 plans in Alaska to 52 in Pennsylvania and West Virginia, and most regions also had Medicare Advantage plans with drug coverage. Average monthly premiums were about $32, though premiums ranged from $2 to $100 depending on the plan and region. Enrollment concentrated quickly: ten companies accounted for about 72 percent of all Part D enrollees, with the two largest sponsors drawing members through brand recognition and low premiums. Most plans used tiered copayments rather than the standard plan’s deductible and coinsurance structure, and very few covered drugs inside the coverage gap. Witnesses before the Senate Special Committee on Aging noted that many beneficiaries found the array of plans confusing, but supporters said the competition held premiums below expectations.

Q: What was the comparative cost adjustment pilot in the 2003 act?

The comparative cost adjustment program was a six-year pilot authorized by the 2003 law to test premium support inside Medicare. In a limited number of metropolitan statistical areas, traditional fee-for-service Medicare would have competed directly against Medicare Advantage plans: each side’s payment would be set by comparing its costs, and beneficiaries would have paid more or less depending on which option they chose. Supporters of the pilot, including the Bush administration, argued that direct competition between the government-run program and private plans would reveal the true cost of each and reward efficiency. Critics argued that the experiment stacked the deck against traditional Medicare and was a first step toward replacing the guaranteed benefit with a voucher. The pilot was one of the most ideologically charged provisions in the conference agreement and illustrated how far the 2003 law went toward introducing market mechanisms into Medicare.