Working Paper Series
Congressional Budget Office
Washington, D.C.

CBO's Model of Drug Price Negotiations Under the
Elijah E. Cummings Lower Drug Costs Now Act

Christopher Adams
Congressional Budget Office
christopher.adams@cbo.gov

Evan Herrnstadt
Congressional Budget Office
evan. herrnstadt@cbo.gov

Working Paper 2021-01

February 2021

To enhance the transparency of the work of the Congressional Budget Office and to encourage external
review of that work, CBO's working paper series includes papers that provide technical descriptions of
official CBO analyses as well as papers that represent independent research by CBO analysts. Papers in this
series are available at http://go.usa.gov/xUzd7.

This work benefited from the guidance of Joseph Kile and Chapin White. Christine Bogusz was the editor.
Jeffrey Kling reviewed the document. Pranav Bhandarkar and Philippa Haven (formerly of CBO) fact-
checked the document and analysis. Helpful comments were provided by Colin Baker, Michael Falkenheim,
Sebastien Gay, Tamara Hayford, Paul Masi, Sam Papenfuss, Lara Robillard, and Ellen Werble of CBO. Anna
Anderson-Cook of Arnold Ventures, Christopher Garmon of the University of Missouri-Kansas City, and
Benedic Ippolito of the American Enterprise Institute also provided useful comments and suggestions.

www.cbo.gov/publication/56905
Abstract

One of the inputs that the Congressional Budget Office used to estimate the budgetary effects of
the Elijah E. Cummings Lower Drug Costs Now Act (H.R. 3) is a simulation model of price
negotiations. CBO modeled those negotiations using a Nash bargaining framework, which was
based on the gains to each party-the government and the manufacturer-from a successful
negotiation. The gain to the government was estimated to be the avoided cost of purchasing the
next-best alternative treatment, plus the incremental clinical value of using the drug of interest
instead of the alternative (measured in dollars), minus the agreed-upon price of the drug. The
manufacturer's gain was estimated to be the revenue from selling the drug in the United States.
This working paper describes the simulation model in detail, including its data sources and
parameter values, and the sensitivity of the results.

In CBO's analysis, the average resulting drug price would be close to the specified upper bound
of 120 percent of the index of international drug prices. Negotiations would reduce prices by 57
percent to 75 percent, relative to current prices, depending on the data and parameters that were
used in the calculations. H.R. 3 specified upper and lower bounds on the prices resulting from
the negotiations; in CBO's estimation, changes to the upper bound would significantly affect the
prices of the drugs the agency examined.

Keywords: pharmaceutical prices, Nash bargaining, negotiation

JEL Classification: 111, 118
Contents

TrntrOduction........e ce eeesesesecesceeseesoreseceseeeseecaseneeeseceseeeseeseceseeeseeesecsaseseesueeeseeeeseesessseneseeeceseeseseseneneneeaees 1
Overview of Nash Bargaining .............ccscescceteescesceeseesscestesccesaesenseaeesseeenevsnseessnesenevanessessnesenesensons 2
Modeling the Maximum Fair Price Negotiations in H.R. 3.0.0.0... esesseseseesessseressseeeseeneresenereseneees 4
Failed Negotiations ............ccscssccsssesescsscssncsssesessssessesscenesencesensessessscascseneseceseeaseeasenseseaseasesaneneses 4
Determining the Negotiated PLrice.............:ssccssssssssssessecesecsseseesesessnecsteseressesenseaseseneseessnseateseneseees 5
Determining the Maximum Fair Price ............:ccscccssecessscseeceseeseneeesetesseecestecenessneessaeeesenesstessnaeeues 8
Data SOUPCES 00... ee eeeeeeeecceeeeeeeseteeeetecensaeeesnenenecescesesseesaneneceteseeesseceseeetneneseneseeeesseesseeseneneseeeeeeeees 8
Set of Existing Drugs CBO Analyzed .........ccscscssssscessccsescesesscesseessecnnessessseessccsecseesssesasenseeseeses 9
Current Prices of Drugs and Alternative Treatments............ccsccscsesccetccssessscssscesseseessresereeesenes 10
Incremental Clinical Bernefits 0.0... cscscsesescssecscseecesesecescccesceeecesesaessesesenesssensssseeseseneneseuenes 10
International Drug Prices .............:cccssccsssseseecescessseceseceeeeesenecsacecsseesenesseeessessneeeseueesueessnesseseeseess 11
Central Estimate and Sensitivity Analysis ............cccssccsscssscsscesscesssesscssecsnceessesessessacesscesessnessneeenes 12
Key Parameters of the Model and Legislative Details................:cccccsscsssssssseseessesssessesenessnesenenes 12
Sensitivity of the Model to Individual Parameter Choices and Legislative Details ................. 15
Interactions Among the Model's Parameter Values ..............c:cccsscssscessscessecseesesseceseeceessseeesenes 16
Role of the MFP Upper Bound in CBO's Model 0.0... cscccsscsssesssenecsssessecesccenessseseessseeneesnenss 16
Possible Future Changes to the Model.............cscccsccssscsssesscesccsecssessscesresecessecsnevsnessceneesenevarenasenes 17
Allow for Interactions Across Negotiations .............ccsccescssscsscccssesscescscssesscsseescesecesesesasesesseees 17
Model the Manufacturer's Cots..........cccccecccsscccesecsseceseresenccssecssecceeeesanesseeseseeeseuseseeessuessneeeneees 18
Incorporate a Utilization Response to Changes in Prices ............::ccsccssecsssessssessneceeecsneesseeeseees 18
Expand the Secretary's Objective FUNCtION.............cccsssscscssscssesssscescssscsseessscnseessessesenseeneessenes 18
Tables .........ceccccscccssecssecescessneccasecsneeeseessacecsesesacecsussesaesenesesaeesaueceaseceaseceneceneeessessnenesaueeeneeeanesenesesenes 20
FIQ™UIES.........c:cccscccssccesscesencesscecsseccaeecnaeasaeeesauessnsecnsccsaesenseesanenaceesssecsasecaaeceeeesseseseeesaueceaeeessaseneneanes 22
Introduction

The Congressional Budget Office developed a model of price negotiations to estimate the
budgetary effects of the Elijah E. Cummings Lower Drug Costs Now Act.! That legislation-
also known as H.R. 3-would require that the Secretary of Health and Human Services (HHS)
negotiate maximum fair prices (MFPs) for a selected set of drugs, which then would be available
at those prices to insurers participating in Medicare Part D (the prescription drug program),
insurers in the commercial market, and other providers who directly administer prescription
drugs to individuals (such as through infusion or injection).

CBO's model, which is based on the widely used Nash bargaining framework, informed its
estimate of the effect of negotiation on pharmaceutical prices in the presence of international
price data. That effect fed into the remainder of the estimate. This paper describes the model, the
data and parameters used to calibrate it, and the sensitivity of the results to those decisions. In
general, the results were similar using different data sources and parameter values, but the
structure of the negotiations stipulated by H.R. 3 played a major role in constraining the results.
The model described below is for the introduced version of H.R. 3; the version that passed the
House was amended in several ways, including to expand the number of drugs subject to
negotiation. CBO has not analyzed the effects of that proposal.

The legislation would not allow MFPs to exceed 120 percent of the average of the prices-called
the average international market, or AIM, price-for a given drug in six reference countries:
Australia, Canada, France, Germany, Japan, and the United Kingdom. H.R. 3 would also
establish a target price for each drug equal to the lowest price available in any of those countries.
If a manufacturer offered a price at or below the target amount during the negotiation process,
that amount would become the MFP.

The Secretary would choose at least 25 drugs for MFP negotiation each year beginning in 2021,
and those prices would be used in Part D beginning in 2023. The set of drugs would be drawn
from the 125 single-source drugs with the highest federal spending in Part D and with the highest
net spending (excluding any rebates provided by drug manufacturers) in the United States
overall. Single-source drugs have no generic or biosimilar competitors; only the company that
made the drug holds the patent and has the exclusive right to sell the drug. The Secretary must
select the 25 drugs that he or she projects will result in the greatest savings to the federal

 

1 For additional information, see Congressional Budget Office, letter to the Honorable Frank Pallone Jr. on the
effects of drug price negotiations stemming from Title 1 of H.R. 3, the Lower Drug Costs Now Act of 2019, on
spending and revenues in Part D of Medicare (October 11, 2019), www.cbo.gov/publication/55722; and letter to the
Honorable Frank Pallone Jr. regarding the budgetary effects of H.R. 3, the Elijah E. Cummings Lower Drug Costs
Now Act (December 10, 2019), www.cbo.gov/publication/55936.
government or to individuals who are eligible for the maximum fair price." The Secretary also
would negotiate prices for all insulin products in the first year (2021).

Pharmaceutical manufacturers that did not comply with certain requirements of title I would be
subject to an excise tax on all sales of the selected drug. The amount of the tax would be a
percentage of the price of each sale: It would start at 65 percent, would increase by 10 percentage
points for each 90 days of noncompliance, and would be capped at 95 percent. Manufacturers
would be prohibited from deducting the excise tax payments from their income taxes. Thus, the
combination of income taxes and excise taxes on the sales could cause the drug manufacturer to
lose money if the drug was sold in the United States. The legislation also includes civil monetary
penalties for manufacturers that sell drugs at prices higher than the maximum fair price. CBO
and the staff of the Joint Committee on Taxation (JCT) expect that a manufacturer would remove
its drug from the U.S. market rather than pay the excise tax.

Because CBO does not know which drugs the Secretary would select for the set each year and
because the necessary data are available only for certain existing drugs, CBO's bargaining model
estimated the effect of negotiations on the prices for a "proxy cohort" of drugs. For those
estimates, the agency used existing data on Medicare spending, international prices, and relative
therapeutic benefits of those drugs relative to their likely alternatives. The model applies most
directly to drugs that are already on the market when they are selected for negotiations and for
which international price data are available.

CBO's full analysis of the budgetary effects of H.R. 3 used the savings implied by the bargaining
model as only one of many inputs.' The agency translated the output from that model into
reductions in prices for future cohorts of drugs that could be subject to negotiations and then
determined the overall effects on the federal budget. For example, CBO's estimate accounted for
changes in the nature of those cohorts over time relative to the proxy cohort, and it modeled
manufacturers' ability to adjust the availability and price of the drug in foreign markets to affect
the international price data on which the negotiations are based.

Overview of Nash Bargaining

Economists use bargaining models to predict the outcome of a transaction when buyers and
sellers both possess some degree of market power. In such cases, there is typically no single
market price; instead, the price is agreed upon jointly by the two parties. Because the potential
transaction would create value for both parties, an agreement is the mutually preferred outcome.

 

2 Sec. 1192(b) of the Elijah E. Cummings Lower Drug Costs Now Act, H.R. 3, 116th Congress.

3See Congressional Budget Office, letter to the Honorable Frank Pallone Jr. on the effects of drug price negotiations
stemming from Title 1 of H.R. 3, the Lower Drug Costs Now Act of 2019, on spending and revenues in Part D of
Medicare (October 11, 2019), p. 8, www.cbo.gov/publication/55722.
However, even though neither party would want negotiations to fail, it is unclear which party
would dictate the terms of the agreement. Indeed, the point of the bargaining model is to predict
how a price would be set under such circumstances.

CBO used a Nash bargaining framework in its analysis of H.R. 3. Although that framework does
not explicitly model the strategic behavior of the two negotiating parties, it is consistent with
several models of negotiations in which the parties do behave strategically, such as Rubinstein
(1982). As such, the Nash bargaining framework is a standard approach to analyzing
negotiations and is used extensively in antitrust enforcement and regulatory review.>

Multiple recent academic papers have specifically used a Nash bargaining framework to model
pharmaceutical price negotiations.° Ganapati and McKibbin (2019) used it to model the prices of
generic and off-patent drugs across various countries outside the United States; Dubois, Gandhi,
and Vasserman (2019) used it to model drug price setting in Canada and simulate how prices in
the United States and Canada would change if the United States adopted a form of reference
pricing. Finally, Ladkawalla and Yin (2015) used a Nash bargaining framework to model
negotiations among drug manufacturers, insurers, and pharmacies. In its analysis of H.R. 3, CBO
applied the Nash bargaining model to the negotiation of prices for on-patent drugs in the United
States that is described in the proposed legislation.

The prediction of the Nash bargaining framework, known as the Nash bargaining solution, is
determined by two factors: the parties' relative bargaining leverage and the parties' relative
bargaining weights. In the Nash bargaining framework (and as it is defined here), leverage is

 

4 See Ken Binmore, Ariel Rubinstein, and Asher Wolinsky, "The Nash Bargaining Solution in Economic
Modelling," RAND Journal of Economics, vol. 17, no. 2 (Summer 1986), pp. 176-188.
www.jstor.org/stable/2555382; and Ariel Rubinstein, "Perfect Equilibrium in a Bargaining Model," Econometrica,
vol. 50 (1982), pp. 97-110.

5 The Department of Justice, Federal Communications Commission, and Federal Trade Commission have used Nash
bargaining models to litigate and evaluate mergers. For example, see United States v. AT&T, Inc., 916 F.3d 1029
(D.C. Cir. 2019); Jonathan B. Baker and others, "Year in Economics at the FCC, 2010-11: Protecting Competition
Online," Review of Industrial Organization, vol. 39, no. 4 (December 2011), pp. 297-309,
https://doi.org/10.1007/s11151-011-9319-3; and Joseph Farrell and others, "Economics at the FTC: Hospital
Mergers, Authorized Generic Drugs, and Consumer Credit Markets," Review of Industrial Organization, vol. 39, no.
4 (December 2011), pp. 271-296, https:/doi.org/10.1007/s11151-011-9320-x. One recent analysis used a Nash
bargaining model to predict whether a merger is likely to be anticompetitive and finds that the approach outperforms
traditional methods; see Christopher Garmon, "The Accuracy of Hospital Merger Screening Methods," RAND
Journal of Economics, vol. 48, no. 4 (Winter 2017), pp. 1068-1102, https://doi.org/10.1111/1756-2171.12215,

6 See Sharat Ganapati and Rebecca McKibbin, "Non-Tariff Barriers and Bargaining in Generic and Off-Patent
Pharmaceuticals" (draft, Georgetown University, March 2019),
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=33 13630; Pierre Dubois, Ashvin Gandhi, and Shoshana
Vasserman, "Bargaining and International Reference Pricing in the Pharmaceutical Industry" (draft, Harvard
University, May 2019), https://scholar.harvard.edu/files/vasserman/files/dgv.pdf; and Darius Ladkawalla and
Wesley Yin, "Insurers' Negotiating Leverage and the External Effects of Medicare Part D," Review of Economics
and Statistics, vol. 97, no. 2 (May 2015), pp. 314-331, https://doi.org/10.1162/REST_a_ 00463.
determined by the net gain to a party from a successful negotiation relative to a failed
negotiation, not including any payments between the two parties. If that net gain is large, that
party will be able to obtain less favorable terms because a failed negotiation would result in a
large loss for the party. A party with less to gain from a successful negotiation (or, equivalently,
with less to lose from a failed one) has more "leverage" in a colloquial sense. The bargaining
weights encompass any other factors-including the party's patience, the cost of capital, and
bargaining skill-that affect the division of the total gains from an agreement.

The Nash bargaining solution is the price that "splits the surplus" from a successful negotiation
proportional to the parties' relative bargaining weights. If the parties have equal bargaining
weights, once the price is paid the parties will have gained the same amount, on net, from a
successful negotiation relative to a failed negotiation.

The Nash bargaining framework can incorporate constraints on the possible outcomes of the
negotiations, an aspect that is particularly relevant to CBO's analysis of H.R. 3. Although the
constraints do not directly affect CBO's analysis of the bargaining leverage or the bargaining
weights, they ensure that the outcome is restricted to a certain range. If the negotiations would
otherwise fall outside that range, the constraints may effectively "pull" the resulting outcome
back to the boundary of the acceptable range of outcomes.

Modeling the Maximum Fair Price Negotiations in H.R. 3

There are two steps to the model that CBO used to estimate the MFPs that would result from
negotiating the net prices of selected drugs under H.R. 3. In the first step, CBO computes the
negotiated price, which is determined on the basis of the gains from a successful negotiation
relative to a failed negotiation and the relative bargaining weights of the two parties. In the
second step, CBO determines the MFP by comparing the negotiated price to bounds defined in
the legislation. Specifically, the MFP is equal to the negotiated price if it falls between the upper
and lower bounds. If the negotiated price exceeds the upper bound, the MFP is equal to the upper
bound. If the negotiated price is less than the lower bound, the MFP is equal to the lower bound.

Failed Negotiations

The consequences of a failed negotiation are an important determinant of the MFP. If drug
manufacturers faced no meaningful consequences from a failed negotiation, the Secretary would
have very little bargaining leverage.' Under H.R. 3, manufacturers that do not agree to
participate in negotiations or that fail to agree to a negotiated price would be subject to an excise
tax on the drug's sales, in addition to income taxes. CBO and JCT expect that a manufacturer
would take its drug off the U.S. market rather than pay the excise tax, so that tax would have the

 

7 See Congressional Budget Office, letter to the Honorable Chuck Grassley concerning negotiation over drug prices
in Medicare (May 17, 2019), www.cbo.gov/publication/55270.
same effect as if the drug had not been approved for sale or as if the drug was excluded from a
national list of drugs (that is, a formulary) that any insurer could cover. The potential use of the
excise tax and anticipated removal of the drug from the United States would serve as a source of
pressure on both parties to reach an agreement.

In a onetime interaction, both the Secretary and the manufacturer would prefer any agreement to
a failed negotiation. So how can the excise tax act affect the negotiation? It can be understood in
two ways. First, such questions of strategic decisionmaking can be understood through the
Rubinstein model, in which the parties alternate offers until one is accepted; if they persistently
reject one another's offers, the negotiation fails.? However, that never happens in the equilibrium
of the model; instead, the parties immediately agree on the Nash bargaining solution. So even
though the parties always come to an agreement, the spectre of protracted disagreement and
failure disciplines the outcome.

Second, one can appeal to the intuition of a repeated negotiation in which the manufacturer and
Secretary will negotiate over many drugs over time." The manufacturer would not want to reveal
a willingness to accept a very low price regardless of the benefit conferred by the drug, lest it
encourage similarly low offers in the future; thus, the threat of disagreement remains an option as
a response to an unreasonably low offer. Conversely, the Secretary would not want to reveal a
willingness to pay a very high price regardless of the benefit conferred by the drug; thus, the
threat of disagreement remains an option as a response to an unreasonably high offer.

Determining the Negotiated Price

In the absence of bounds on the MFP, the negotiated price is the solution to the bargaining model
that CBO used to estimate H.R. 3. Formally, that solution is the price that maximizes the
weighted product of the two parties' gains from coming to an agreement.

In the model, the manufacturer's benefit from an agreement is equal to its revenue from selling
the drug, which is calculated as the negotiated price multiplied by the quantity sold. The
manufacturer's benefit from a failed negotiation is zero because the manufacturer is expected to
remove the drug from the U.S. market in that case. Therefore, the net gain to the manufacturer
from a successful negotiation is equal to the negotiated price multiplied by the quantity sold in
the U.S. market.

In addition, the Secretary's per-beneficiary, per-year benefit from an agreement is equal to the
average per-beneficiary, per-year value of the drug to a beneficiary taking the drug, minus the

 

8 See the Appendix for a full characterization of the Rubinstein model in the context of MFP negotiations.

° Although CBO does not explicitly model each MFP negotiation as part of a series of bargains, such a framework
provides another intuitive reason that the excise tax could encourage an agreement.
negotiated price.!° That value is the clinical benefit of the drug (as measured in dollars), and it is
multiplied by the quantity sold to Medicare beneficiaries. The benefit from a failed negotiation is
the clinical benefit of the next-best alternative treatment (which might be a different drug or even
a nonpharmaceutical treatment) minus the price of that treatment. Therefore, the gain to the
Secretary from a successful negotiation is the incremental clinical benefit of the drug minus the
incremental price of the drug, relative to the alternative treatment.!!

The model reflects two technical assumptions about the payoffs to the parties. First, the cost of
producing, distributing, and marketing the drug in the United States is assumed to be zero.
Although that may be a reasonable approximation for some drugs, it overstates the gain to the
manufacturer from an agreement. Incorporating nonzero costs to manufacturers would result in
higher negotiated prices. Second, use of the drug does not change with the negotiated price (or
neither party accounts for such a change while negotiating). To the extent that there is a change
in use and it is a factor in the negotiations, a lower price generates more surplus, which creates a
mutual incentive to settle on a lower price than the parties otherwise would. In CBO's
simulation, the negotiated prices for most drugs are predicted to lie above 120 percent of the
AIM price, so small deviations from these assumptions would probably have little effect on the
overall estimate of reduced federal spending.

The Nash bargaining problem is expressed mathematically as:
max [4m(¢ - P) - 4m (Ca - Pa)]*[(4m + Go)p - OF

The gain for each party is its payoff from a completed negotiation minus its payoff from a failed
negotiation. The first term in brackets is the Secretary's gain from a completed negotiation
relative to a failed negotiation:

m= The health benefit to a Medicare beneficiary from receiving that drug (e), minus the amount
of money paid for the drug under the potential negotiated price (p), times the number of
Medicare beneficiaries receiving treatment (q,,), minus

 

10 The model does not account for possible spillover benefits to other beneficiaries, such as benefits associated with
vaccines.

11 When CBO estimates the budgetary effects of a policy that changes the number of prescriptions filled, the agency
typically applies a "medical offset" that captures associated changes in other medical expenditures. The bargaining
model does not account for that medical offset to the drug being negotiated or the relevant alternative treatments.
Instead, CBO applies the medical offset to the changes in future drug spending as computed in the main budgetary
analysis, to which the bargaining model is one input. For more details about the medical offset, see Congressional
Budget Office, Offsetting Effects of Prescription Drug Use on Medicare's Spending for Medical Services
(November 2012), www.cbo.gov/publication/43741.
m The health benefit to a Medicare beneficiary from receiving the alternative treatment (e,),
minus the amount of money paid for the alternative treatment (p,), times the number of
Medicare beneficiaries receiving treatment (q,,).

The second term in brackets is the manufacturer's gain from a successful negotiation relative to a
failed negotiation:

m The revenue from selling the drug to the entire U.S. market, minus the number of Medicare
beneficiaries using the drug (q,,,), plus the number of other fair-price-eligible individuals
taking the drug (q,), times the potential negotiated price (p), minus

m= Zero U.S. revenue from the drug it is taken off the market.

The exponents A and (1 - A) represent the relative bargaining weights of the Secretary and
manufacturer, respectively. In this context, those bargaining weights succinctly capture a number
of factors, including the Secretary's preference to hold out for low prices, potential political
ramifications for the Secretary if negotiations are delayed, the public relations consequences for
the manufacturer if it is perceived as being unwilling to negotiate, and the relative costs of
capital between the parties.

The negotiated price solves the Nash bargaining model above and is equal to

(1 - A)(e - eg + pq). That is, the parties split the incremental clinical benefit of the drug plus
the avoided price of purchasing the alternative treatment. The negotiated price has a sensible
relationship to the opportunity cost of the drug from the Secretary's perspective. If the drug
provides a large therapeutic benefit over the existing alternative, the negotiated price would tend
to be high. If the drug is used in place of an expensive alternative treatment, the negotiated price
would also tend to be high. Those are both cases in which the Secretary would receive large
gains from access to the drug, so his or her leverage would be low. The bargaining weight 2
determines how the negotiated price would divide the net benefit of the drug between the two
parties. If the bargaining weights are 50-50 (that is, if A = 0.5), the split would be even. The
Secretary would receive the net benefit of the drug but pay half of it to the manufacturer through
the negotiated price.'"

 

12 Because current prices are negotiated between manufacturers and insurers, CBO's model assumes a ceiling on
prices that could be negotiated between the Secretary and manufacturers. In the model, the negotiated price is
constrained to a range below the current price and above zero. That parameter makes little difference in CBO's
analysis of H.R. 3 because the MFP is typically constrained by the target price or 120 percent of the AIM price. In
the context of modeling future policies that did not include an AIM-based upper bound or similar constraint,
however, that modeling assumption could be more consequential.
In CBO's model, the cost of the alternative treatment reflects its current-law price. That price
could also be negotiated under the legislation, however, so the two negotiations could interact
with one another because one drug's MFP would feed into the negotiation over the other drug's
MFP. In CBO's full analysis of the budgetary effects of H.R. 3, the agency adjusted its estimate
to account for some competitive pressure among drug prices elsewhere in the analysis rather than
explicitly characterize it in the bargaining model. Nonetheless, the results are similar using a
"Nash-in-Nash" (Nash-equilibrium-in-Nash-bargaining) framework in which the price of an
alternative treatment is that treatment's model-predicted MFP if it is negotiated.

Determining the Maximum Fair Price

H.R. 3 effectively constrains the MFP to lie between the target price and 120 percent of the AIM
price. That constraint does not directly affect the negotiated price, but it can prevent the parties
from agreeing to an MFP equal to the negotiated price.' If the negotiated price was below the
target price, the manufacturer would offer the target price and the Secretary would have to
accept. That outcome could occur if a drug has a small incremental clinical benefit, an
inexpensive alternative, or an especially high target price. Conversely, if the negotiated price
exceeded 120 percent of the AIM price, the parties would agree to an MFP equal to 120 percent
of the AIM price. That outcome could occur if the drug has a large incremental clinical benefit,
an expensive alternative, or an especially low AIM price. Finally, if the negotiated price lies
between the target price and 120 percent of the AIM price, the MFP would equal the negotiated
price. For a graphical summary of the three cases as applied to a hypothetical drug with a target
price of $800 and an AIM price of $1,000, see Figure 1. In that figure, the target price and 120
percent of the AIM price are shown using dotted lines. Between those lines, the MFP is equal to
the negotiated price. Outside those lines, the MFP is equal to one of the bounds.

Data Sources

To model the effect of negotiation on the prices of future cohorts of drugs, CBO analyzed a set
of existing drugs for which the necessary data were available. The model's calculations required
four main data series: the current price of each drug and its alternative treatment(s), the
incremental clinical benefit of the drug in terms of quality-adjusted life years (QALYs) or life

 

13 For more information on the Nash-in-Nash modeling approach, see Allan Collard-Wexler, Gautam
Gowrisankaran, and Robin S. Lee, " 'Nash-in-Nash' Bargaining: A Microfoundation for Applied Work," Journal of
Political Economy, vol. 127, no. 1 (February 2019), pp. 163-195, https://doi.org/10.1086/700729.

14 For an example of a Nash bargaining problem in which the possible outcomes are constrained, see Abhinay
Muthoo, Bargaining Theory With Applications (Cambridge University Press, 1999), pp. 23-24 and Figure 2.5,
https://doi.org/10.1017/CBO97805 1 1607950.

15 For an explanation of how the target price and AIM price upper bound result in this solution (using Rubinstein
(1982) as the strategic model that underpins the Nash bargaining solution), see the Appendix.
years saved, a value that converts that incremental clinical benefit into dollars, and international
prices for each drug.

Set of Existing Drugs CBO Analyzed

Beginning in 2021, H.R. 3 requires the Secretary to choose each year at least 25 drugs for which
prices would be negotiated.'° CBO's analysis of that legislation modeled multiple future cohorts
of negotiated drugs, but it did not project the exact drugs that would be included in those cohorts.
Instead, the analysis focused on a set of existing drugs-the proxy cohort-which formed the
basis for additional modeling of the future cohorts through the 10-year budget period.

The proxy cohort comprises drugs covered by Medicare. Furthermore, because the model
requires information on the incremental clinical benefit of the drugs, CBO restricted the proxy
cohort to include only those drugs that had been evaluated in an Institute for Clinical and
Economic Review (ICER) comparative-effectiveness study at the time of the analysis. That set of
drugs provides a convenience sample, and ICER's studies tend to focus on relatively high-priced,
high-expenditure drugs.'" The proxy cohort includes 28 drugs typically covered by Part D and 10
drugs generally covered by Part B.'® Five of the drugs in the proxy cohort were purchased by
both Part D and Part B in 2017 (the year of analysis).!° The proxy cohort includes all high-
expenditure drugs for which data are available; it is not necessarily the specific set of drugs CBO
expects the Secretary would negotiate.

 

16 Tn the first year of negotiations (2021), all insulin products would also be included.

17 Tn 2017, across the 28 drugs in the sample covered by Medicare Part D, the average total expenditure was $648
million, and the median total expenditure was $248 million. Across the 10 drugs in the sample covered by Medicare
Part B, the average total expenditure was $767 million, and the median total expenditure was $540 million. Of all
the drugs in the sample, Medicare spent the least on cystic fibrosis drug Orkambi ($83 million), which ranked 225th
among all drugs covered by Part D that had a single manufacturer. For comparison, the 125th-ranked single-
manufacturer drug in terms of 2017 Part D spending (Janumet XR, a drug used to control high blood sugar in people
with Type 2 diabetes) cost a total of $223 million. Those total expenditure values and rankings are from the
Medicare Part B and Part D spending "dashboards" and do not account for rebates and discounts. However, they
provide some sense of the ranking of the drugs in the proxy cohort relative to the likely ranking of negotiation-
eligible drugs.

18 Part D is Medicare's outpatient drug benefit, in which beneficiaries usually purchase prescription drugs at retail
pharmacies. Drugs covered under Part B are generally infused or injected, and beneficiaries receive them in a
physician's office or outpatient hospital setting. Some drugs may be covered by both parts of Medicare. H.R. 3 did
not specifically instruct the Secretary to negotiate drugs covered by Part B; rather, the bill directed the Secretary to
choose drugs for negotiation that are commonly used in the market as a whole, which included several Part B drugs.

19 Because alternatives and costs may differ across contexts, CBO separately modeled Part D and Part B
negotiations for those drugs when computing the average MFP.
CBO estimated the MFP for each drug in the proxy cohort, expressed as a percentage of the AIM
price, and used the simple average across drugs (or MFP index) to inform its broader model of
the budgetary effects of negotiations in H.R. 3 on the prices of future cohorts of drugs.

Current Prices of Drugs and Alternative Treatments

For its calculations, CBO started with data from Medicare's aggregate spending "dashboards"
for Part D and Part B."° To compute current drug prices, CBO divided the total annual
expenditure for a drug in 2017 by the number of beneficiaries receiving the drug in that year.
Thus, all prices in CBO's analysis are in terms of dollars per beneficiary-year. The dashboard
prices for Part D do not reflect rebates, which reduce the net price of the drug, so CBO adjusted
those prices using confidential Medicare data on the rebates and discounts that insurers
participating in Part D obtained from manufacturers in 2017.

Incremental Clinical Benefits

CBO used data from a number of ICER's reports to quantify the incremental clinical benefit of a
drug relative to alternative treatments."! ICER produces comparisons of the outcomes that a
patient with a given condition can expect under a variety of treatments. Many of those
comparisons analyze drugs that would probably be eligible for negotiation.

To measure a drug's benefit, CBO used either incremental life years or quality-adjusted life
years as reported by ICER. (One year spent in a hypothetical state of perfect health is equal to
one QALY.) For conditions that are fatal, the incremental change in expected life years may be
an appropriate measure of the drug's effectiveness. For conditions that are rarely fatal but reduce
quality of life, QALYs may be a more appropriate measure." Patients with those kinds of
conditions (such as plaque psoriasis, an autoinflammatory condition that causes skin lesions)
benefit from drugs that improve their quality of life, but they do not necessarily live longer.

Some groups have expressed concerns about the use of QALYs. For example, some argue that
the measure potentially discriminates against older or disabled patients and does not reflect
patient-specific preferences for quality versus length of life. Despite the limitations of using
incremental QALYs and life years to approximate the benefit of a treatment, such an approach

 

20 Centers for Medicare & Medicaid Services, "CMS Drug Spending", www.cms.gov/Research-Statistics-Data-and-
Systems/Statistics-Trends-and-Reports/Information-on-Prescription-Drugs.

21 See Institute for Clinical and Economic Review, "Who We Are," www. icer.org/who-we-are.

22 CBO uses QALYs and life years as proxies to model the value of a drug; however, the model does not require
QALYs or life years to be an explicit criterion in the Secretary's decisionmaking. ICER uses both QALYs and life
years (which it refers to as equal value of life years gained) to communicate the incremental clinical benefit of a
drug; see Institute for Clinical and Economic Review, "Cost-Effectiveness, the QALY, and the evLYG,"
http://icer.org/our-approach/methods-process/cost-effectiveness-the-qaly-and-the-evlyg.

10
was the best available to CBO and is consistent with the approach taken by many countries to
negotiate drug prices."

To model the Secretary's valuation of those clinical benefits, CBO converted the incremental
increase in QALYs or life years to dollar-equivalent units using conversion factors from HHS.
The formula for the negotiated price requires that the clinical benefit be expressed in dollar-
equivalent units so that it can be compared with the drug price being proposed. CBO used the
value of a statistical life year (VSLY) and the willingness to pay fora QALY (WTP-Q) to
monetize life years and QALYs, respectively. (VSLY and WTP-Q are standard concepts used in
regulatory analysis.)** Because the dollar-equivalent measure of the incremental clinical benefit
needed to be in annual per-beneficiary terms, CBO divided the total incremental value by the
number of remaining life years reported by ICER.

International Drug Prices

For its analysis of H.R. 3, CBO consulted two sources of information on the relationship between
U.S. drug prices and international drug prices: an analysis by the staff of the House Committee
on Ways and Means and data underlying an analysis by So-Yeon Kang and others." Since that
time, the Council of Economic Advisers has released a similar study."° Even though the three
analyses differ somewhat in the measures and countries they report, their qualitative findings
largely agree.

CBO's analysis used data only from the six countries that were used to determine H.R. 3's
calculation of the AIM price. When using the results of the studies by Ways and Means and So-
Yeon Kang and others, CBO calculated the ratio of the foreign price to the Medicare Part B or D

 

3 For example, the United Kingdom's National Institute for Health and Care Excellence (NICE) uses QALYs to
help decide whether to approve drugs for sale to the National Health Service; see NICE, "Who We Are and What
We Do", www.nice.org.uk/about/who-we-are/our-charter. For a general summary of the advantages of and concerns
with using QALYs in a cost-effectiveness context, see Peter J. Neumann and Joshua T. Cohen, "QALYs in 2018--
Advantages and Concerns," Journal of the American Medical Association, vol. 319, no. 24 (June 2018), pp. 2473-
2474, https://doi.org/10.1001/jama.2018.6072.

24 HHS recommends using VSLY and WTP-Q in its regulatory impact analyses and has done so in practice. For the
department's recommendations regarding VSLY and WTP-Q, see Department of Health and Human Services,
Office of the Assistant Secretary for Planning and Evaluation, Guidelines for Regulatory Impact Analyses (January
2017), pp. 15-22, https://aspe.hhs.gov/pdf-report/guidelines-regulatory-impact-analysis. For an example of the
department's use of VSLY in a regulatory benefit calculation (a final rule issued by the Food and Drug
Administration), see Use of Materials Derived From Cattle in Food and Cosmetics, 81 Fed. Reg. 14718, 14730
(March 18, 2016), https://go.usa.gov/xAaEd.

2° See staff of the House Committee on Ways and Means, A Painful Pill to Swallow: U.S. vs. International
Prescription Drug Prices (September 2019), https://go.usa.gov/xVeff (PDF, 1.2 MB); and So-Yeon Kang and
others, "Using External Reference Pricing in Medicare Part D to Reduce Drug Price Differentials With Other
Countries," Health Affairs, vol. 38, no. 5 (May 2019), pp. 804-811, http://tinyurl.com/y4eejm6m.

26 See Council of Economic Advisers, Funding the Global Benefits to Biopharmaceutical Innovation (February
2020), https://go.usa.gov/xAaP8.

11
price, net of rebates and discounts, for each individual drug common to the studies and the proxy
cohort. CBO then applied the average of those ratios across drugs and countries to the current
U.S. price of the drug to calculate the AIM price. The agency used the average ratio of the lowest
foreign price to the U.S. price across drugs to calculate the target price.

Because the report from the Council of Economic Advisers presented only the average price
ratios between the United States and various foreign countries across all drugs in its study, CBO
applied the average of the price ratios between the United States and the six countries specified
by H.R. 3 to compute the AIM price. The agency used the smallest of the six price ratios (from
Japan for all drugs and from the United Kingdom for retail drugs only) to compute the target
price.

The average price ratios across drugs in the proxy cohort inform the full budgetary analysis of
H.R. 3 only through the MFP index. For that full budgetary analysis, CBO used a combination of
data underlying the studies by Ways and Means and So-Yeon Kang and others to model the
relationship between the AIM price and the U.S. price separately for each future cohort. CBO
then applied the MFP index (based on the proxy cohort) to the AIM price of future cohorts of
drugs as a starting point for estimating the overall budgetary effects of the legislation.

Central Estimate and Sensitivity Analysis

To test the validity of its analysis, CBO varied some parameters of its model under a range of
sensitivity cases. For each drug in the proxy cohort, CBO computed the MFP and the percentage
reduction in price. The MFP index and the average price reduction are the simple averages of the
MFP relative to the AIM price and of the percentage reduction in price relative to current U.S.
prices, respectively, across drugs. Solving the model under those various cases shows that the
results were generally similar when CBO used alternative data sources or parameter values.
Varying the upper bound on the MFP (which is specified in the legislation) could affect the
outcomes substantially, however.

Key Parameters of the Model and Legislative Details
CBO considered a range of values for each of several parameters and legislative details.

Source of International Prices. CBO's central estimate uses the analysis by the House Ways
and Means Committee. One sensitivity case uses the analysis by So-Yeon Kang and others. The
other two sensitivity cases use measures from the analysis by the President's Council of
Economic Advisers; one uses a full sample of drugs, and the other limits the analysis to drugs
bought in a retail setting.

12
Bargaining Weights. The central estimate assigns equal bargaining weights (0.5 for the
Secretary and 0.5 for the manufacturer)."' The sensitivity cases assign the Secretary a bargaining
weight as low as 0.25 and as high as 0.75.78

Secretary's Valuation of Therapeutic Benefits. In its central estimate, CBO used willingness-
to-pay values suggested by HHS, converted to 2017 dollars."? The VSLY was $388,000 per life
year, and the WTP-Q was $507,000 per QALY.*° In accordance with guidance from HHS, the
sensitivity cases used values that were 50 percent higher and 50 percent lower.*!

Identification of the Next-Best Alternative Treatment. The next-best alternative treatment
determines the incremental benefit of a given drug and the price the government would have paid
for the alternative if that drug was not available. Predicting exactly which alternative negotiators
would have in mind is difficult, however. In CBO's central estimate, when negotiations fail, the
treatment of beneficiaries depends on the alternative. If the proxy cohort has no other treatments,
then beneficiaries would be treated using the standard of care identified by ICER. The standard
of care might involve a nonpharmaceutical treatment (for example, the best supportive care for
multiple sclerosis) or an older pharmaceutical treatment (for example, methotrexate for
rheumatoid arthritis). In the central estimate, if there are other treatments for the same condition

 

27 CBO often assigns probabilities of 50 percent to each of two outcomes when it does not have enough information
to identify either outcome as more likely than the other. See Congressional Budget Office, Estimating the Cost of
One-Sided Bets: How CBO Analyzes the Effects of Spending Triggers (October 2020), Box 1,
www.cbo.gov/publication/56698. Because the United States does not have experience negotiating drug prices upon
which CBO can draw, the agency's central estimate is that the surplus would be split evenly between the Secretary
and the manufacturer.

28 The manufacturer's bargaining weight is always one minus the Secretary's bargaining weight.

29 CBO previously reported that it used $400,000 and $520,000 for the VSLY and the WTP-Q, respectively; those
values were expressed in 2018 dollars rather than 2017 dollars. CBO expects that the difference would have had a
small effect on the agency's overall estimate of changes in federal spending under H.R. 3. See Congressional Budget
Office, letter to the Honorable Frank Pallone Jr. on the effects of drug price negotiations stemming from Title 1 of
H.R. 3, the Lower Drug Costs Now Act of 2019, on spending and revenues in Part D of Medicare (October 11,
2019), www.cbo.gov/publication/55722.

30 The WTP-Q value ($507,000) is based on Department of Health and Human Services, Office of the Assistant
Secretary for Planning and Evaluation, Guidelines for Regulatory Impact Analysis (January 2017), p. 21,
https://aspe.hhs.gov/pdf-report/guidelines-regulatory-impact-analysis (PDF, 1.7 MB). The VSLY ($388,000) is
based on the value used by the Food and Drug Administration in Use of Materials Derived From Cattle in Food and
Cosmetics, 81 Fed. Reg. 14718, 14730 (March 18, 2016), https://go.usa.gov/xAaEd. The general guidance from
HHS is to use a central estimate of about $10 million for the value of a statistical life in 2017. To confirm that the
VSLY that CBO used is consistent with the HHS guidance, CBO spread the HHS value of a statistical life over 50
remaining years of life and used a discount rate of 3 percent, resulting in an equivalent VSLY of about $382,000.
Other federal agencies use similar values of a statistical life. (All values have been inflated to 2017 dollars.)

31 See Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation,
Guidelines for Regulatory Impact Analyses (January 2017), pp. 16-17, https://aspe.hhs.gov/pdf-report/guidelines-
regulatory-impact-analysis (PDF, 1.7 MB).

13
in the proxy cohort, then half of the beneficiaries would receive the treatment with the highest
total amount of Medicare spending and half would receive the standard of care. Two sensitivity
cases vary the share of beneficiaries that would receive an alternative treatment if it was
available in the proxy cohort.

u In the first sensitivity case, all beneficiaries would receive the alternative treatment that
accounted for the highest Medicare spending.

m In the second sensitivity case, all beneficiaries would receive the standard of care, even if an
alternative treatment existed in the proxy cohort.

In some cases, there are multiple alternative treatments in the proxy cohort. In CBO's central
estimate, anyone not receiving the standard of care would receive the alternative treatment with
the highest Medicare spending. The following sensitivity cases are instead based on a standard
model of consumer demand.*"

mu In the first sensitivity case, beneficiaries would receive the alternative treatment in the proxy
cohort in proportion to the share of Medicare beneficiaries that received each alternative
treatment in 2017-and no beneficiaries would receive the standard of care when an
alternative treatment is available.

m In the second sensitivity case, half of the beneficiaries would receive an alternative treatment
in the proxy cohort in proportion to the share of Medicare beneficiaries that received each
alternative treatment in 2017, and half of the beneficiaries would receive the standard of care.

Legislative Details. H.R. 3 established bounds on the MFP. The sensitivity cases allow for three
possibilities: The MFP could be as high as the current price, as low as zero, or both. For a

summary of the wide range of sensitivity cases that CBO considered for those parameters, see
Table 1.

 

32 Share-based substitution is implied by a logit demand model of differentiated products. See Gregory J. Werden
and Luke Froeb, "The Effects of Mergers in Differentiated Products Industries: Logit Demand and Merger Policy,"
Journal of Law, Economics, and Organization, vol. 10, no. 2 (October 1994), pp. 407-426,
https://doi.org/10.1093/oxfordjournals.jleo.a036857.

14
Sensitivity of the Model to Individual Parameter Choices and Legislative Details

In CBO's central estimate, the MFP is 116 percent of the AIM price, on average, a result that is
largely robust to the choice of single model parameters and data sources (see Table 2).** The
MFP index remains fairly close to the central estimate in most sensitivity cases because the
negotiated prices for most of the drugs in the proxy cohort lie well above 120 percent of the AIM
price. Thus, even when the model's parameters are varied substantially, most drugs still have an
MFP equal to 120 percent.

For example, suppose that 120 percent of the AIM price for a drug is $1,000, and the central
estimate of the negotiated price is $1,500 (and thus the MFP is $1,000). If the value used for
WTP-Q was the parameter being varied, the MFP would remain at $1,000 unless the WTP-Q
was more than one-third lower. Similarly, if the Secretary's bargaining weight was the parameter
being varied, the MFP would remain at $1,000 unless that bargaining weight increased from 0.5
to 0.67 or more.

CBO's central estimate of the average price reduction from negotiations in H.R. 3 is 68 percent
of current prices. But that result changes somewhat depending on the source of the international
price data. If the source of that data differs, the same MFP index for a given drug can imply
different savings rates. Thus, the model can produce different average savings even if the
negotiations are constrained by the upper bound (120 percent of the AIM price), resulting in
more variation in the savings rate. Even in that case, though, the range is quite small except when
the model uses the CEA's international price ratio that includes nonretail drugs.

As part of its sensitivity analysis, CBO modeled the impact of the legislation without a statutory
upper bound on the MFP; in that case, CBO estimates that price negotiations would still result in
some reduction in the average drug's price (see the bottom section of Table 2). Specifically,
prices would be 22 percent lower than under current law, on average, under the legislation.*4
That finding implies that even if firms managed to distort the AIM price by changing their
behavior in foreign markets, the MFP negotiations would result in some price reductions
nevertheless. In contrast, the presence of the target price (which acts as a lower bound) is much

 

33 CBO previously reported that the MFP was 114 percent of the AIM price, on average. The difference was due to
an error in data preparation that has since been corrected, but CBO expects that the difference would have had a
small effect on the agency's estimate of changes in federal spending under H.R. 3. See Congressional Budget Office,
letter to the Honorable Frank Pallone Jr. on the effects of drug price negotiations stemming from Title 1 of H.R. 3,
the Lower Drug Costs Now Act of 2019, on spending and revenues in Part D of Medicare (October 11, 2019),
www.cbo.gov/publication/55722.

34 Estimates of other policies to implement a mechanism that would involve the Secretary negotiating drug prices
would depend on the specifics of those proposed policies. Those specifics could affect the structure of the model and
the appropriate parameter values.

15
less important. There are very few drugs in the proxy cohort for which the negotiated price falls
below the target price, so it rarely plays a role in determining the MFP.

Interactions Among the Model's Parameter Values

CBO also performed a sensitivity analysis that accounted for interactions among all possible
combinations of the various data sources and parameter values (presented in Table 1), holding
the upper and lower bounds on the MFP fixed at the legislatively specified values.

The outcomes are similar across all parameter values that CBO examined, although there is some
variation (see Figure 2 and Figure 3). The MFP index varies more when the Secretary has more
bargaining power. That is the case because the negotiated price is more likely to fall below 120
percent of the AIM price and, therefore, the model's exact parameters directly affect the value of
the MFP index. Even so, considering the full range of possible outcomes, the average price
reduction is contained in a reasonably tight range-roughly 57 percent to 75 percent. As was the
case for the results in Table 2, the spread in the average price reduction is largely driven by
differences in the international price data. When the sensitivity cases do not include those using
the CEA's international price comparison that includes retail and nonretail drugs (the blue series
in Figure 2 and Figure 3), the spread in savings rates narrows to 67 percent to 75 percent.

The range of outcomes grows as the bargaining weight assigned to the Secretary rises. A higher
bargaining weight could indicate that the Secretary drives a harder bargain or is more skilled at
negotiating. All else being equal, a given drug's negotiated price is more likely to fall below 120
percent of the AIM price as the Secretary's bargaining weight rises. Because the negotiated price
is still rarely below the target price, the MFP would then equal the negotiated price. Therefore,
changing the exact values of the various parameters is more likely to change the MFP as more
drugs move into that range.

Role of the MFP Upper Bound in CBO's Model

The MFP upper bound had a large effect on CBO's estimate of the average MFP from
negotiations under H.R. 3. The proposed legislation set the MFP upper bound at 120 percent of
the AIM price, resulting in MFPs that were 32 percent of current prices, on average. If H.R. 3
had instead specified that the MFPs could be no larger than 100 percent of the AIM prices, the
MFPs would have been 27 percent of current prices, on average (see Figure 4). The MFPs
increase as the upper bound rises, eventually settling at 78 percent of the current price, on
average, if the upper bound is at least 360 percent of the AIM price.

Changing the upper bound only affects the MFP of drugs in the proxy cohort with negotiated
prices above that bound.*> Most drugs in the proxy cohort fall into that category when the upper

 

35 For an illustration of how the upper bound and negotiated price interact to determine the MFP, see Figure 1. The
negotiated price is also constrained to be less than or equal to the current price.

16
bound is 120 percent of the AIM price. For those drugs, when the upper bound increases slightly,
the MFP also rises. Thus, the savings on those drugs decrease. For drugs with negotiated prices
already below the AIM price, the MFP and savings are unchanged.

As the upper bound increases (moving from left to right in Figure 4), it surpasses the negotiated
price of drugs one by one, and the MFPs of those drugs are no longer affected by the upper
bound. As the MFPs of fewer drugs are affected, an additional increase in the upper bound has a
smaller effect on average savings across all drugs in the proxy cohort, and the line becomes
flatter and flatter. At about 360 percent of the AIM price the line becomes horizontal; because
there are no drugs in the proxy cohort with a negotiated rate above that value, increasing the
upper bound further has no effect on the average savings.

The sensitivity analysis shows that CBO's estimate of the average savings from negotiations in
the proxy cohort would have been very similar under a wide range of alternative model
parameters. In contrast, changing the legislatively defined bounds on the MFP could have a
larger effect on the estimated savings. Specifically, CBO's estimated average price reduction of
68 percent would have been slightly larger if the upper bound was set equal to the AIM price and
about two-thirds smaller if the legislation had not specified an upper bound on the MFP.

Possible Future Changes to the Model

CBO continually reviews its models. For future analyses, the agency would consider a number of
changes to its framework for modeling drug price negotiations. Four such changes are examined
below. It is unclear how incorporating the below changes would affect CBO's estimate of federal
savings under the legislation.

Allow for Interactions Across Negotiations

The model used in the cost estimate for H.R. 3 assumed that negotiations are completely
independent. CBO's model of a drug price's negotiation uses the alternative treatment's current
price as part of the surplus from an agreement, even if the alternative is also likely to be
negotiated. That approach could be reasonable if it was unlikely that the prices of multiple drugs
targeting the same condition would be negotiated.

To the extent that the assumption of independence is not reasonable, however, the agency could
model negotiations as occurring simultanousely in a Nash-in-Nash framework.*© That framework

 

3© Other researchers use frameworks similar to Nash-in-Nash to model negotiations between governments and
pharmaceutical manufacturers. See Sharat Ganapati and Rebecca McKibbin, "Non-Tariff Barriers and Bargaining in
Generic and Off-Patent Pharmaceuticals" (draft, Georgetown University, March 2019),
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=33 13630; and Pierre Dubois, Ashvin Gandhi, and Shoshana
Vasserman, "Bargaining and International Reference Pricing in the Pharmaceutical Industry" (draft, Harvard
University, May 2019), https://scholar.harvard.edu/files/vasserman/files/dgv.pdf.

17
finds a full set of MFPs that are self-reinforcing, in the sense that if an alternative treatment is
also likely to be negotiated, its MFP is part of the benefit from an agreement, rather than its
current price.

Model the Manufacturer's Costs

CBO's model simplifies the manufacturer's objective function by setting marginal costs of
production and distribution and the average cost of marketing in the United States at zero.
Although that approximation may be reasonable in some cases, future analyses could incorporate
central estimates of those cost parameters if the appropriate data were available. For example,
manufacturing costs of some specialty drugs could be large enough to affect negotiations. In
addition, a drug's marketing costs could be an important component of negotiations. Those costs
are typically high, and some share of the costs would not be recoverable by the time a drug's
price was negotiated. The manufacturer would avoid incurring any future marketing costs if no
agreement was reached, though, and those avoided costs would affect the negotiated price.
Estimating incremental future marketing costs for an individual drug that is already on the
market in the United States could be very difficult, however.

Incorporate a Utilization Response to Changes in Prices

CBO's model uses a technical assumption that the quantity of a drug sold does not respond to its
price. That assumption is reasonable given that beneficiaries (especially Medicare beneficiaries)
tend to be insulated from the full cost of a drug through the (usually small) cost-sharing amounts
they pay. That said, there is probably some quantity response; in its budgetary analysis of H.R. 3,
CBO estimated that beneficiaries would increase their utilization of drugs in response to
reductions in drug prices.

Future modeling of price negotiations might incorporate a similar utilization response. Given the
potentially large increase in modeling complexity and the likely small magnitude of the response,
though, such an effort might not be worthwhile. If future modeling did allow for a utilization
response in Medicare Part D, the model also could account for offsetting changes in medical
expenditures in keeping with CBO's typical approach.

Expand the Secretary's Objective Function

In CBO's model, the Secretary's objective function includes only Medicare beneficiaries.
Because utilization of a drug does not vary with its price in that model, the negotiation can be
expressed by accounting only for the per-beneficiary net benefit of the drug. In future analyses,
the Secretary might also put weight on the well-being associated with purchases in the
commercial market.

Incorporating commercial purchases would raise a number of analytical and empirical questions.

m First, how would the Secretary weight health benefits and savings to enrollees in commercial
insurance plans?

18
m Second, how would the Secretary weight savings to commercial insurers and beneficiaries
that would benefit from access to the MFP?

m= Third, what prices do beneficiaries actually pay in the commercial market?

If the benefits to commercial beneficiaries were similar to the benefits to Medicare beneficiaries,
putting weight on the commercial beneficiaries would have a limited effect on the negotiated
price. All told, the analytical complexity and data required might outweigh the benefits of a more
detailed model.

19
Tables

Table 1: Full Range of Sensitivity Cases CBO Considered

 

Parameter

Possible Values
(Value used for central estimate in bold)

 

Source of international price data

Secretary's bargaining weight

Monetary value of incremental clinical benefit

Share of beneficiaries receiving standard of
care given a failed negotiation (if there is an
alternative drug in the proxy cohort, otherwise
100%)

Alternative drug received by beneficiaries
given a failed negotiation (if multiple
alternatives are present in the proxy cohort)

 

Committee on Ways and Means, So-Yeon
Kang and others, CEA, CEA (retail only)

0.25, 0.375, 0.5, 0.625, 0.75

50%, 100%, 150% of central VSL/WTP-Q
case

0, 50%, 100%

Alternative with the highest expenditure,
weighted average based on beneficiary shares

 

Data source: Congressional Budget Office.

CEA = Council of Economic Advisers; VSL = value of statistical life; WIP-Q = willingness to pay for an additional

quality-adjusted life year.

20
Table 2: CBO's Central Estimate and Its Sensitivity to Parameters

 

 

MFP Index Average Price Reduction
Case (Percentage of (Percentage of
AIM price) current price)
Central Estimate 116 68
International Price Data:
So-Yeon Kang and others' international prices 117 71
CEA's international prices 116 58
CEA's international prices (retail only) 117 71
Bargaining Weights:
HHS Secretary's bargaining weight = 0.25 118 67
HHS Secretary's bargaining weight = 0.75 109 70
Monetary Value of Incremental Clinical Benefit:
VSL/WTP-Q 50 percent lower 114 68
VSL/WTP-Q 50 percent higher 117 67
Alternative Treatment If Negotiation Fails:
All beneficiaries receive highest-spending alternative
if one is available M7 67
All beneficiaries receive SOC, even if an alternative
is available 116 68
Beneficiaries receive an alternative proportional to
Medicare shares M7 67
Half of beneficiaries receive an alternative 116 68
proportional to Medicare shares, half recetve SOC
Legislative Details:
MFP lower bound is zero 115 68
MFP upper bound is the current price 281 22
MFP lower bound is zero and upper bound is the 280 2

current price

 

Data source: Congressional Budget Office.

AIM = average international market; CEA = Council of Economic Advisers; HHS = Department of Health and
Human Services; MFP = maximum fair price; SOC = standard of care; VSL = value of statistical life; WTP-Q =

willingness to pay for an additional quality-adjusted life year.

21
Figures

Figure 1: The Maximium Fair Price As Determined by the Negotiated Price, the Target Price, and
the AIM Price

1,400

1,200

 

1,000

800

 

600

A400 Target price 120 percent of AIM price

Maximum Fair Price (Dollars)

200

0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000
Negotiated Price (Dollars)

Data source: Congressional Budget Office.

The maximum fair price is plotted as a function of the negotiated price under a hypothetical scenario in which the
target price is $800 and the AIM price is $1,000.

AIM = average international market.

22
Figure 2: Sensitivity Range of the MFP Index As a Percentage of the AIM Price

Estimate

120 _-¥¥ PP EF sarees tt seen ~r-------| CBO's Central f

 

-
-
oO

ok
oOo oO
Oo oOo

80

" Set of Cases

60 * All Sensitivity Cases

50 ® Without CEA "All Drugs" Case

40
30
20

MFP Index (Percentage of AIM price)

10

0.250 0.375 0.500 0.625 0.750
Secretary's Bargaining Weight

Data source: Congressional Budget Office.

There are a range of MFPs associated with a given value of the Secretary's bargaining weight. The dot indicates the
MFP holding all parameters at their central values, except for the bargaining weight. The orange series includes the
full set of sensitivity cases. The blue series omits any sensitivity case for which the international prices were

determined using CEA's price ratio that was calculated using both retail and nonretail drugs.

AIM = average international market; CEA = Council of Economic Advisers; MFP = maximum fair price.

23
Figure 3: Sensitivity Range of the Average Price Reduction Relative to Current Prices

100
90
80

a t es f* Simm a So F i sence J Acme re CBO's Central Estimate

60
Set of Cases

50 * All Sensitivity Cases

~® Without CEA "All Drugs" Case
40

30
20

10

Average Price Reduction (Percentage of current price)

0.250 0.375 0.500 0.625 0.750
Secretary's Bargaining Weight

Data source: Congressional Budget Office.

There are a range of prices associated with a given value of the Secretary's bargaining weight. The dot indicates the
price holding all parameters at their central values, except for the bargaining weight. The orange series includes the
full set of sensitivity cases. The blue series omits any sensitivity case for which the international prices were

determined using CEA's price ratio that was calculated using both retail and nonretail drugs.

CEA = Council of Economic Advisers.

24
Figure 4: Effect of the MFP Upper Bound on Average MFPs Relative to Current Prices

100

oo
oO

Oo
oO

hb
oO

 

NO
oO

120 percent of AIM price (as specified in H.R. 3)

Average MFP (Percentage of current price)

100 | 150 200 250 300 350 400
MFP Upper Bound (Percentage of AIM price)

Data source: Congressional Budget Office.

AIM = average international market; MFP = maximum fair price.

25
Appendix: The Maximum Fair Price in a Rubinstein
Alternating-Offers Model

This appendix contains a derivation of the maximum fair price using the strategic bargaining
model from Rubinstein (1982).>" The parties take turns making offers and counteroffers until an
agreement is reached. As the length of time between offers gets smaller, the subgame perfect
equilibrium (SPE) of the Rubinstein model converges to the solution of an appropriately defined
Nash bargaining model.

Suppose, without loss of generality for the final result, that the Secretary of the Department of
Health and Human Services makes the first price offer. The drug manufacturer then decides
whether to accept that price or make a counteroffer in the next period. That back-and-forth
continues for an infinite number of periods, which can be interpreted as the idea that there is
always time for another offer. If the parties perpetually reject each other's offers, the drug is
taken off the market, and both parties get zero payoff.**®

The notation for the model is described in Table A-1.*' If an offer p is accepted in period t, the
Secretary gets 6£(V - p) and the manufacturer gets df,p.

Table A-1: Notation Used in the Rubinstein Model

 

 

 

 

 

 

 

Parameter Description
Ps» Pm Potential offers made by the Secretary (s) and the manufacturer (m).

1; The discount rate of party i.
A The amount of time between offers.
6; The discount factor of party i, equal to exp(-7;A).

?,p The target price and 120 percent of the average international market price,

~ respectively.

V =e--e, +p, | The surplus from an agreement (the incremental clinical benefit plus the
cost of the next-best alternative treatment).

 

 

 

 

Data source: Congressional Budget Office.

 

37 See Ariel Rubinstein, "Perfect Equilibrium in a Bargaining Model," Econometrica, vol. 50 (1982), pp. 9-110.
38 That outcome corresponds to the payoff from a failed negotiation in the Nash bargaining model.

39 To save on notation, the model is expressed in per-beneficiary, per-year terms. As in the Nash bargaining
framework in the main body of the paper, the number of beneficiaries does not affect the MFP if it does not vary
with the price.

26
The SPE is the following: The Secretary always offers p; and accepts any offer from the
manufacturer Pm < Pm, Whereas the manufacturer always offers p,, and accepts any offer from
the Secretary p, = p;, where:

 

p -wW<p
Pm = HV, BV € [p,p]
Pp wW>p
and
Smp, pV <p
ps = 45muV, pV € [p,p]
1-85
and p = 15.8,"

Confirming that a set of strategies constitutes an SPE requires checking that both parties are
making optimal decisions in each possible period, assuming the other party does not otherwise
deviate from its potential equilibrium strategy. The confirmation of the SPE when pV € [p, P|

closely follows Proposition 3.1 in Muthoo (1999).

If uV < p, the target price constrains the offers. If it is the Secretary's turn to make an offer, he
or she knows that if that offer is rejected, the manufacturer will offer p and the Secretary will

accept it. However, the manufacturer incurs a cost from waiting for a period. From the
perspective of this round of offers, the manufacturer's payoff from rejecting the Secretary's offer
and offering p is 6,,p. Thus, if the Secretary offers a price of 6,,p, the manufacturer will accept

it. If the Secretary tries to offer a slightly lower price, the manufacturer will prefer to reject the
offer, so the Secretary is making the best possible offer that will be accepted. On the other side,
when it is the manufacturer's turn to make an offer and it offers p (the target price), the Secretary

must accept the offer as per H.R. 3, which completes the proof.

When uV > p, the upper bound (120 percent of the AIM price) constrains the offers. If it is the
Secretary's turn to make an offer, he or she knows that if that offer is rejected, the manufacturer
will offer p,and the Secretary will accept it. From the perspective of this round of offers, the

 

40 See Abhinay Muthoo, Bargaining Theory With Applications (Cambridge University Press, 1999), pp. 45-46,
https://doi.org/10.1017/CBO97805 11607950.

27
manufacturer's payoff from rejecting the Secretary's offer and offering p is 6,,p. Thus, if the
Secretary offers a price of 6,,p, the manufacturer will accept it. If the Secretary tries to offer a
slightly lower price, the manufacturer will prefer to reject the offer, so the Secretary is making
the best possible offer that will be accepted. On the other side, when it is the manufacturer's turn
to make an offer and it offers p (120 percent of the AIM price), the Secretary accepts the offer if
the payoff from accepting, which is V - 7, is at least as large as the discounted payoff from

rejecting the offer and having his or her next offer accepted, which is 6,(V - 5,,p). That

(1-6s)
(1-6m5s)
manufacturer cannot have an offer higher than p accepted (per H.R. 3), the manufacturer cannot
do better than offering p, which completes the proof.

 

condition is equivalent to whether V = pV 2 p, which it is, since nV > p. Because the

The SPE in the standard Rubinstein model corresponds to the standard Nash bargaining solution
when the time between offers (A) becomes arbitrarily small. In that case, 6, > 1,5, > 1, and

ho = . In the case of the MFP negotiation, the outcome of the SPE converges to:

 

 

 

 

 

Ts+
if ( Ts )v<
P, " Ts + % Pp
( Is )v if ( Is )ve
PVN + Tn yt % + 1% [P,P
D if ( is \v>p
DP; 1 1, + Tm p

That is the solution to CBO's Nash bargaining model if the manufacturer and the Secretary have

Ys Tm

 

and ,
Tst!m Tst1m

 

bargaining weights equal to respectively, and the MFP is bounded by p and p.

28