Does Open Enrollment Control Premiums? A Case Study from the "Medigap" Market
Thomas Rice Katherine A . Desmond Peter D
F o x
This article analyzes a change in "Medigap" regulations that occurred in Missouri in 1999. It allows Medicare beneficiaries in the state to switch to a different carrier each year so long as they retain the same standardized policy type, without losing their open enrollment privileges. The analysis is based on a comparison of various outcomes in Missouri and those in two comparison states, Kansas and Florida. We found little evidence that the policy change affected premiums charged by insurance carriers in Missouri, but conclude that other desirable aspects of the change make it potentially attractive for other states to follow.
This article provides a case study of a "Medigap" regulation that went into effect in Missouri in July 1999. Prior to that time, Missouri followed the practice of nearly all other states in allowing Medicare beneficiaries to have open enrollment Privileges for all standardized Medigap plans °nly during the first six months after initial en rollment in Medicare at age 65. The purpose of these provisions was to allow less healthy ben eficiaries a window of time in which they could Purchase more affordable coverage, while still Providing some protection to carriers against ad verse selection from those who would seek cov erage after becoming ill. The new Missouri regulation liberalizes open enrollment, by allowing beneficiaries to switch to another carrier on the policy anniversary date—that is, once a year. The primary stipula tion is that they retain the same policy type, al though they may switch to other plan types at 1
2
the discretion of the carrier. Thus, those with Plan C can switch to any other carrier that sells that plan, but they are not guaranteed access to other plan types. This, therefore, precludes someone without prescription drug coverage (covered only in Plans H, I , and J) from having the right to add this coverage if they initially enrolled in a plan without this benefit. The primary purpose of the legislation was to enhance competition in the state's Medigap market. The Missouri regulation appears to have the potential to greatly affect Medigap policy pric ing. Because carriers are at risk of losing their policyholders each year, they may feel compelled to price their policies lower. Whether this occurs depends, in part, on the degree of price competi tion in the Medigap market, something that, to our knowledge, has not been studied carefully. On the one hand, one may expect price competi tion to be high because benefits are standardized 3
4
Thomas Rice, Ph.D., is a professor in the Department of Health Services, School of Public Health, University of California, Los Angeles. Katherine A. Desmond, M.S., is a private consultant. Peter D. Fox, Ph.D., is president of PDF, LLC. Funding for this study came from the Changes in Health Care Financing and Organization initiative of the Robert Wood Johnson Foundation. Address correspondence to Prof. Rice at Department of Health Services, UCLA School of Public Health, $0 S. Young Drive, Los Angeles, CA 90095-1772. Email:
[email protected] 6
Inquiry 41: 291-300 (Fall 2004). © 2004 Excellus Health Plan, Inc. 0046-9580/04/4103-0291
291
Inquiry/Volume 41, Fall 2004
and premium information on competing plans is widely available through state-provided premium comparison guides. On the other hand, price competition is probably reduced because seniors have shown themselves (at least on average) to be least able to understand health insurance informa tion made available to the public (Hibbard et al. 1998; Hibbard et al. 2001). The purpose of this paper is to evaluate the im pact of Missouri's regulatory change. It compares various Medigap market outcomes in Missouri to those of two comparison states over a seven-year period: three years prior to the change, and four years afterward. The analysis should be viewed only as a case study rather than one providing de finitive conclusions, since it compares Missouri's experience to that of just two other states. Methodology
or mandate community rating (National Associa tion of Insurance Commissioners 2000). To determine the effect of Missouri's liber alization of open enrollment, we chose Kansas as a comparison state that does not employ any rating restrictions, and Florida as a state that does. Since 1993, Florida has required that all Medigap policies be sold on an issue-age basis. While Kansas is contiguous to Missouri, Florida is not. Florida is similar to Missouri in that it is a large state in both area and population, with substantial urban and rural areas. The two policy changes in Missouri, both of which would have affected premiums beginning in January 2000, cannot be fully disentangled: the expanded open enrollment provisions (the focus of our study), and the banning of attained-age rating. Our strat egy to doing so is a statistical one, described in the "Findings" section.
Choice of Comparison States
Data Sources
6
To help interpret changes over time in Missouri, it The primary data sources are premium compari is useful to compare its experience to trends in son guides developed and distributed by the three other states. We chose two comparison states study states: Missouri, Florida, and Kansas. Each because effective January 1, 2000, Missouri also of these states publishes a listing of Medigap pre enacted another regulation that precluded what miums, separately by policy type A-J, for all are known as "attained-age" policies. Of the 10 Medigap carriers in their respective states. These Missouri Medigap carriers examined in this study, guides were obtained for seven years: 1997 through 2003 (inclusive). The first three years seven used attained-age rating through 1999. There are three ways in which age is reflected (1997-99) apply to premiums that were set prior in Medigap premiums: attained-age rating, issue- to implementation of the Missouri open enroll age rating, and community rating. Under at ment policy, and the last four, afterward. To keep tained-age rating, beneficiaries pay more in pre the analysis tractable, information was collected miums each year as they age because on on the 10 largest carriers in each state as of the average, medical expenses increase as a person year 2000. For some carriers, premium data were becomes older. With issue-age rating, premiums not available over our entire time period; in those are set based on the age of initial purchase—al cases, we dropped the carrier and chose the next though they can rise in the future if the carrier largest (as defined by number of Medigap policy raises them for all policyholders. Under commu holders). For Missouri, we dropped one such car nity rating, all policyholders in a geographic area rier and included the 1 l largest. For Kansas, we are charged the same amount. As long as the age dropped two such carriers and included the 1 1 distribution of beneficiaries in the area remains and 13 largest. The Florida premium guides rep stable, increases in premiums will depend only resent a sampling rather than a census of carriers; on increases in medical care use and prices. But in order to include carriers with premium data if younger beneficiaries move away, the rising available over the entire time period we had to average age of policyholders will cause premi go as low as the 17 largest firm. The combined market share of the firms we analyzed in Mis ums to rise. Many policymakers dislike attained-age rating souri was 70%. In Kansas and Florida it was because as an individual beneficiary ages, his or 89% and 73%, respectively. For some of the analyses, we also employed her premiums rise—often at the same time the person is facing lower income. Twelve states, in data from the National Association of Insurance cluding Missouri, either ban attained-age rating Commissioners on the number of policyholders 5
I h
th
lh
th
292
Medigap
Market
3,000
• 2,500 •
2,000
-
-
o o
Premium ($)
•
•
•
I • •
o o
•
*
•
•
•
• • *
•
• — • ...
t
•
• •
1 •
#
*
*
1 »
1
—
500
—
-- —
-
—
-
0 19 96
1997
1998
1999
2000
2001
2002
2003
2004
Year
Figure 1.
Missouri Plan F, scatter plot of premiums
for each carrier, by plan type, as well as informa tion on loss ratios. These data, however, were available only with a lag, with 2001 being the most recent at time of analysis.
scriptive statistics on whether policyholders gravitated to cheaper Medigap policies in the year following the Missouri policy change (com paring Missouri and Florida only); and 5) trends in overall loss ratios in the three states.
Analytic Methods Both descriptive and multivariate analyses were conducted. The analyses focus on changes in the level and variability of premiums in each of the three states. Both these measures are impor tant because they each provide an indication of competitiveness. Premium levels indicate wheth er the open enrollment provisions kept insurers from raising premiums as fast in Missouri as in the comparison states, while premium variability shows whether insurers were more likely to price their policies at levels similar to their competi tors, compared to the other states. The study findings are based on five sets of analyses, presented in the following order: 1) scatter plots of premiums charged by each of the 10 carriers in each state; 2) trends in premi ums and the variability in premiums (measured by the standard deviation) over time in each state, averaged among the 10 carriers; 3) ordinary least squares (OLS) regression analyses examining the impact of the open enrollment policy change on Premiums and the variability in premiums; 4) de
Findings Scatter plots show the actual premium charged by each carrier in each state for each plan type. Be cause carriers are required to sell Plan A (along with any others they choose), and because Plans F and C are the most frequently purchased in each state, we focused on these. Plan F is presented here because the plots were similar in nearly all cases for each plan type and it was the most pop ular plan in all three states. Figures 1-3 show this information for Missouri, Florida, and Kansas, re spectively. (Note that Kansas was not able to pro vide premiums for the year 1999.) The patterns for all three states appear fairly similar; most no ticeable, perhaps, is that the dispersion in premi ums for Missouri seems to have increased after the open enrollment regulations were enacted in 2000-the opposite of what was anticipated. 8
To examine these trends more closely, Figures 4 and 5 show unweighted mean premiums and the standard deviation in premiums calcu lated over each state's 10 Medigap carriers. To 9
293
Inquiry/Volume 41, Fall 2004
Figure 2.
Florida Plan F, scatter plot of premiums
facilitate interpretation, values in each state in each year were standardized to the year 2000, that is, divided by the average over all companies in the year 2000. The trends in mean premiums are almost identical (Figure 4) among the three states, and there is no visual evidence that the
Figure 3. 294
trend line fell in Missouri compared to the other two states starting in 2000. There also appear to be no trends in premium variability (Figure 5). To confirm these visual interpretations (which are only shown for Plan F), we carried out regres sion analyses. Unlike the descriptive analyses,
Kansas Plan F , scatter plot of premiums
Medigap Market
which focused on only three plan types, the re gression analyses looked at plan types A - J of fered by the 10 carriers in each state in order to increase sample size and take advantage of all available data. Two dependent variables were spe cified: the natural logarithm of the annual premi
um charged by each carrier in a state for a particular plan type in a particular year and a measure of premium dispersion; the natural log arithm of the absolute value of the actual premi um charged by the carrier for a plan type in a given year minus the average (mean) premium
Inquiry/Volume 41, Fall 2004
Table 1. O L S regression results: predictors of logged Medigap premiums Variable Intercept Kansas Florida Post (20002003) Kansas * Post Florida * Post Plan B Plan C Plan D Plan E Plan F Plan G Plan H Plan I Plan J Attained-age rating Community rating Model R
2
Estimated Standard t coefficient error value Significance 6.43 -.09 .21
.06 .04 .06
102.31 -2.19 3.46
.000 .037 .002
.34 .09 .03 .33 .52 .43 .38 .59 .46 .94 1.06 1.25
.07 .08 .10 .02 .02 .03 .04 .03 .04 .09 .09 .09
4.52 1.11 .34 13.78 21.75 14.83 9.92 23.41 10.84 10.55 11.93 14.20
.000 .276 .735 .000 .000 .000 .000 .000 .000 .000 .000 .000
-.10
.07
-1.52
.138
.18
.05
3.49
.000
.81
Table 2. O L S regression results: predictors of logged variability in Medigap premiums Variable Intercept Kansas Florida Post (20002003) Kansas * Post Florida * Post Plan B Plan C Plan D Plan E Plan F Plan G Plan H Plan I Plan J Attained-age rating Community rating Model R 2
Note: Model corrected for clustering by firm.
Estimated Standard t coefficient error value Significance 3.95 .77 .31
.24 .23 .27
16.58 .34 .11
.000 .739 .910
.77 -.20 .08 .19 .43 .20 .09 .25 .34 1.20 1.02 1.14
.23 .22 .42 .16 .15 .23 .26 .15 .22 .29 .39 .38
3.31 -.91 .18 1.19 2.94 .89 .34 1.65 1.52 4.20 2.64 3.02
.003 .370 .858 .243 .006 .382 .735 .110 .140 .000 .013 .005
.17
.24
.73
.473
.87 .18
.20
4.42
.000
Note: Model corrected for clustering by firm.
charged by our 10 carriers in the state for that plan type and year. The dependent variables were logged to correct for skewed distributions. The estimated equations were specified as: Dependent Variable =/(state dummies, post-period dummy, interactions between state dummies and post-period dummy, plan type dummies, plan age rating dummies), where state dummies represent Florida and Kansas (Missouri being the reference group); and the post-period dummy represents the years 2000-03 (1997-99 being the reference), which correspond to the period after Missouri's regula tory change. We also included two interaction terms (Florida X post-period and Kansas X post-period); nine plan dummies representing plans B-J, respectively (Plan A being the refer ence group); and two dummies representing at tained-age and community-rated policies (with issue-age policies being the reference). As noted earlier, at about the same time that Missouri enacted its open enrollment provisions, it also banned attained-age rated policies. This could have an effect on premiums and thus could con found the effects of the public policy initiative 296
of interest. The ratings dummy variables were de signed to capture the effects of rating practices on premiums (and their variability) over time, so that any remaining impacts could be attributed to the open enrollment provisions. Our analyses included multiple observations for each company in each state (the seven years of data). Because of the resulting lack of indepen dence, we employed a clustering correction. The interaction terms are of primary interest, as they represent the differences between Mis souri and the other two states in the changes from the pre- to post-implementation periods." We hypothesized that, compared to Missouri, Medi-. gap policies in Florida and Kansas were more ex pensive and showed more variation in premiums after Missouri enacted its open enrollment provi sions. The regression results are shown in Tables 1 and 2. Beginning with premium levels in Table 1, there is no evidence that the Missouri change kept premiums in check compared to trends in Florida and Kansas, with p-values for.the two state post-implementation interaction variables not approaching statistical significance. 10
Nearly all but one of the independent vari ables used to predict Medigap premiums were
Medigap
statistically significant at the .01% level, and the signs were as expected. Policies sold in the postimplementation period were more expensive than those in the pre-implementation period—not sur prising given rising health care costs—and all other policy types were considerably more ex pensive than Plan A, which was expected since they provided more benefits. Attained-age poli cies were cheaper (although not quite statistically significant), and community-rated policies were more expensive than issue-age policies. This was expected because the premium comparison guides provide prices for 65-year-olds. If a carrier uses attained-age rating, it can charge 65-yearolds less because it raises prices in later years. The opposite is true of community rating: if a car rier uses community rating, the prices it charges to 65-year-olds are the same as the prices charged to 75- or 80-year-olds since all beneficiaries are required to be charged the same amount, irrespec tive of age and health status. As was the case for premium levels, the Mis souri policy change had no detectable effect on the variation in premiums (Table 2). Neither in teraction term approaches statistical significance. Because we do not have many prior expecta tions about the control variables, they are not dis cussed here. We conducted one sensitivity analysis, where data for the year 2000 were omitted from the analysis. It could be argued that 2000 would be a transition year, since the open enrollment regu lations went into effect only six months prior, potentially after carriers already had decided on the next year's premiums. The results, however, were almost identical, and as before none of the state post-implementation variables approached statistical significance for either the mean pre mium or variability regressions. The analyses up to this point focus only on car rier pricing practices rather than on what consum ers buy. Prices charged do not provide complete information because consumers might gravitate away from more expensive policies—which would show up in the data only if other carriers responded by lowering their prices. This was ex amined in a limited fashion by analyzing whether individuals in Missouri (compared to Florida) who had more expensive policies in 2000 switched to cheaper policies in 2001. Specifically, the top 10 policies in Missouri and Florida were categorized according to whether they were 12
Market
Table 3. Market share in 2000 and 2001, most and least expensive Medigap plans Combined market share (%) Plan Florida Plan C 5 less expensive plans in 2000 5 more expensive plans in 2000 Plan F 5 less expensive plans in 2000 5 more expensive plans in 2000 Missouri Plan C 5 less expensive plans in 2000 5 more expensive plans in 2000 Plan F 5 less expensive plans in 2000 5 more expensive plans in 2000
2000
2001
47.5
46.2
17.8
22.6
41.0
40.1
29.3
30.8
21.4
21.4
42.8
40.6
40.9
35.4
38.9
29.1
among the five more or less expensive in the year 2000, for Plans C and F (the two largest sellers). The market shares of each group of five carriers in 2000 and 2001 were compared, with the ex pectation that the market shares would increase among the five cheaper carriers and fall for the five more expensive ones. The results are shown in Table 3. Florida, the comparison state, shows no pattern of policy holders moving to less expensive plans; if any thing, the opposite occurred. Missouri, however, shows a modest trend in that direction. The market shares of the five least expensive plans compared to the five more expensive plans increased for both Plans C and F, albeit modestly. The effect for Plan F was the larger of the two. In 2000,2% more Missourians purchased one of the cheaper type F plans than expensive ones (40.9% vs. 38.9%); this increased to 6.3% in 2001 (35.4% vs. 29.1%). The final analysis shows trends in overall loss ratios in the three states between 1997 and 2001. One would expect greater price competition— which is the hypothesized result of Missouri's policy change—to cause premiums to be lower 297
Inquiry/Volume 41, Fall 2004
s'
'«
? 8.
7 6
7 4
- • Florida - Kansas
—*—Missouri
I -
• -
1
72
70
64 1997
1998
1999
2000
2001
Yeir
Figure 6.
Loss ratios for carriers of standard Medigap plans
than they would have been otherwise, resulting in a higher proportion of the premium dollar being devoted to benefit payments rather than adminis tration and profits (i.e., loss ratios should rise). For this analysis, all carriers in the three states were examined, not just the 10 largest. If Missou ri's policy were effective in restraining premiums, one would expect an increase in Missouri's loss ratios in 2000 and 2001 (over pre-2000 levels) compared to Florida and Kansas. The results are shown in Figure 6. No obvious trend is apparent. In all three states, loss ratios fell between 1997 and 2001, indicating that carriers were keeping more of the premium dollar for ad ministrative expenses and/or profits. This could be due to the insurance underwriting cycle, how ever. It is true that the overall decline was somewhat smaller in Missouri than in the other states. This was not the case comparing 2001, for which the most recent data were available, with 1999, the last year before Missouri's policy change went into effect. Over that period, loss ratios in both Missouri and Florida fell by 2.5%, but they declined by only .4% in Kansas. 13
Discussion We could detect little evidence that Missouri's open enrollment policy change—which allows 298
seniors to change Medigap carriers on an annual basis without penalty—had an effect on the pre miums charged by carriers in the state. One should be cautious, however, in drawing strong conclusions because of several study limitations. Most important, this is a case study comparing Missouri to only two other states. It is possible that other comparisons would provide different results. Second, while some of the data provide information on experience through 2003, others contain information only through 2001, provid ing just two years of experience after the regula tions were implemented. This is particularly a problem for examining the loss ratios to learn about price competition. Third, all but one of the analyses focus on the behavior of carriers rather than the purchasing behavior of seniors. Finally, although we attempted to control for the fact that Missouri enacted another policy change (the prohibition against the sale of at tained-age rated policies) at the same time it adopted its open enrollment provisions, there is still concern that the multivariate analysis could not fully isolate the two. The results are somewhat ambiguous although certainly leaning in the direction of no impact. There was no evidence that the policy affected premiums in the visual analysis or in the multi variate analysis. Furthermore, of the two other
Medigap
analyses, one showed some very mild effects but the other showed none. We found that consumers in Missouri were slightly more likely to switch to low-cost carriers after the policy change. In con trast, overall loss ratios did not decline relative to those in other states. Nevertheless, the change enacted by Missouri would appear to have other desirable character istics. By allowing seniors to change Medigap policies, they are less likely to feel locked into a policy they perceive to be too expensive, and it does offer the potential of keeping car riers honest in their behavior. Finally, it affords an important protection to those seniors who find themselves in a closed block of business, and who therefore could face rising premiums
Market
due to the lack of replenishment of their risk pool. The state has not collected any data indicating how many Medigap policyholders have taken ad vantage of the regulatory provisions by changing to a less expensive carrier. Insurance department officials, however, perceive that the regulation has been very effective in protecting consumers, both by encouraging insurers to keep their rates af fordable, and by allowing a safety valve for those individuals whose premiums have become too high. The change seems to have done no harm, and one of our analyses indicates that it may have spurred some consumers to choose lower-cost Medigap plans. For those reasons, other states may wish to consider enacting similar provisions. 14
Notes The views expressed in this article are those of the au thors and do not reflect the vim's of the UCLA School of Public Health; the Departments of Insurance in Florida, Kansas, and Missouri and their staffs; or the Robert Wood Johnson Foundation or its staff. The authors would like to thank John Howser and James Casey from the Missouri Department of Insur ance, Don Brown of the Kansas Department of Insur ance, and Linda Ziegler and Rich Rohleto of the Florida Department of Insurance for providing copies of the Medigap premium comparison guides from their respective states, upon which much of the analysis is based. The authors are also grateful to John Howser and James Casey for reviewing a draft of the manu script. Finally, we appreciate guidance provided by the journal's editor, Katherine Swartz. 1 Insurance carriers received official notice of the change on May 11, 1999, from the Missouri De partment of Insurance. Hearings on the proposed regulation occurred about six weeks earlier, on March 25, 1999. The regulation could not have affected 1999 premiums, which became effective on January 1, 1999. However, they likely would have affected 2000 premiums, which did not go into effect until about eight months later. 2 The Omnibus Budget Reconciliation Act of 1990 requires that all Medigap plans sold after July 1992 conform to one of 10 standardized sets of benefits, lettered A-J. (States are permitted to en act regulations that further limit the number of plan choices.) It granted Medicare beneficiaries the right to choose any policy from any carrier without being subject to medical underwriting and without being charged more than others. This open enrollment privilege does not apply after the first six months of Medicare eligibility at age 65, at which point carriers can refuse to provide cov
erage to applicants. Under current law, the only exceptions concern beneficiaries who enroll in a Medicare-rChoice (typically, HMO) plan who disenroll within one year, or whose Medicare+ Choice or Medigap plan go out of business (Fox, Snyder, and Rice 2003). 3 If beneficiaries choose to remain with the same carrier, as most do, they are subject to health re view if they choose another plan type. The Mis souri regulations are limited to those who switch carriers but keep the same plan type. 4 A secondary purpose was to protect beneficiaries in "closed blocks of business" against high pre miums. Carriers that stop selling a particular plan base their premiums on the individuals in that plan. Because the risk pool is not replenished, beneficiaries can find themselves with steeply ris ing premiums as policyholders in this closed block age. The Missouri provision allows benefi ciaries who find themselves in this situation to move to another carrier (personal communication, James Casey, Missouri Department of Insurance, October 2, 2001). 5 This regulation was approved at the same time as the open enrollment provision and also an nounced on May 11,1999. Unlike the latter, how ever, which became effective in July 1999, the banning of attained-age rated policies was effec tive on January 1, 2000. The difference in dates is of little consequence, however, since the impact of both regulations would have occurred effective January 2000. 6 Other states with premium-rating restrictions in clude: Arkansas, Connecticut, Idaho, Massachu setts, Maine, Minnesota, New York, Vermont, and West Virginia (NAIC 2000). Massachusetts and Minnesota are not appropriate comparison states because, along with Wisconsin, they were 299
Inquiry/Volume 41, Fall 2004
exempt from the federal Medigap standardization requirements because they had their own pre-ex isting system when the legislation went into effect. 7 Loss ratios are defined as the proportion of policy premiums that are paid in benefits rather than be ing used for administration and profit. Some of the reasons that loss ratios are below 100% are ad ministrative expenses, advertising, profits, and the need for insurers to retain reserves to cover unan ticipated losses. 8 Plots for the other two plan types are available from the authors. 9 We did not weight each company's contribution to the average premium so as to reflect the price choices available to beneficiaries rather than aver age prices paid. 10 One concern was that the addition of the ratingpractice dummy variables would lead to multicol linearity, because Missouri instituted the two at the same time. But a reason to believe it would not be a problem is that not all carriers in the state were using attained-age rating before it was out lawed. To test for multicollinearity, we calculated variance inflation factors (VIF). Before adding these dummy variables, the maximum V I F was 4.1, and rose to 5.2 after adding the rating varia bles. This is still well below the threshold of 10.0 for "harmful multicollinearity" (Kennedy 1998, p. 190).
11 Our interaction specification can be described as a difference-in-differences model as follows: Y = B + /3,KS+)3 FL + /3 POST + B KS • POST 0
2
3
4
+ B FL • POST + • • 5
Pre Missouri Kansas Florida MO-KS difference MO-FL difference
Post
Pre-Post Difference
Po
Po + Pi
Po + Pi
Po + Pi
Pi Pi
+ P4
Po + Pi
+ P3 + P4 P0 + P2
Pi
+ jB
s
+ Pi + Ps Pi
Pi
Pi
Pl+Ps
+P4
PA Ps
12 Kansas was not included because the largest sell ing carrier in the state in 2000, Blue Cross Blue Shield of Kansas, was not listed in the NAIC data set in 2001. 13 Unfortunately, no data are availability on whether the underwriting cycle in the Medigap market exhibits different trends in different states. 14 Personal communication, James Casey, Missouri Department of Insurance, April 14, 2004.
References Fox, P.D., R.E. Snyder, and T . Rice. 2003. Medigap Reform Legislation of 1990: A 10-Year Review. Health Care Financing Review 24(3): 1-17. Hibbard, J.H., J.J. Jewett, S. Engelmann, et al. 1998. Can Medicare Beneficiaries Make Informed Choices? Health Affairs 17(6): 181-193. Hibbard, J.H., P. Slovic, E . Peters, et al. 2001. Is the Informed-Choice Policy Approach Appropriate for
300
Medicare Beneficiaries? Health Affairs 20(3): 199-203. Kennedy, P. 1998. A Guide to Econometrics. Cam bridge, Mass.: MIT Press. National Association of Insurance Commissioners (NAIC). 2001. Medicare Supplemental Loss Ratios in 2000. Kansas City, Mo.: NAIC.