![U.S. flag](https://www.dol.gov/themes/opa_theme/img/flag-favicon-57.png)
An official website of the United States government
The pricing of health insurance in the Consumer Price Index (CPI) is difficult. Health insurance can be viewed as a composite good. The total price of health insurance is reflected by the insurance premium, but much of that premium is used to indirectly purchase medical goods and services on behalf of the beneficiary. The total insurance premium includes these indirectly purchased medical goods and services, as well as services provided by the insurance company (such as claims processing, risk protection, and negotiating with providers). The U.S. Bureau of Labor Statistics (BLS) prices these two components of the total premium separately in the CPI, and the CPI for health insurance measures only the price of services provided by the insurance company. The current method for pricing health insurance in the CPI is referred to as the “indirect method.” This method contrasts the “direct method” for pricing health insurance, which prices the total insurance premium and is used in the BLS Producer Price Index (PPI).
One limitation of the indirect method is that it results in a health insurance index that can be difficult to interpret and cannot be directly compared with measures that capture total-premium inflation. However, an implied total-premium measure can be constructed from CPI data. In this article, I discuss the issue of measuring health insurance inflation, describe the current method for pricing health insurance in the CPI, calculate a measure of implied total-premium inflation in the CPI, and compare that implied measure with total-premium inflation as measured by the PPI.
Implied total-premium inflation in the CPI generally tracks the year-over-year change in the overall CPI for medical care (hereafter also referred to as “medical CPI”). Over longer periods, however, the implied measure increases more than the medical CPI, because medical services that exhibit above-average inflation (such as hospital services) have a greater weight in the implied measure than in the medical CPI. From December 2005 to December 2022, the implied total-premium index increased by 77.9 percent, an average annual increase of 3.4 percent. Over the same period, the CPI for medical care increased by 67.8 percent, an average annual increase of 3.1 percent. The CPI total-premium index increased more than the industry PPI for health insurance between 2005 and 2022, but the two measures of total-premium inflation did not track closely with each other over shorter periods, a divergence that can partly be explained by a data lag in the CPI for health insurance.
In a direct method for pricing health insurance, one would track movements in insurance premiums, holding constant the quality of insurance, and use these price relatives to construct a health insurance index. Historically, however, the CPI program has been unable to consistently control for changes in insurance quality, such as those related to policy benefits and health-related risk factors. As a result, BLS developed an indirect method for pricing health insurance in the CPI—a method called the retained-earnings method. This section discusses the methods of direct and indirect pricing, as well as the strengths and limitations of each.
The retained-earnings (indirect) method for pricing health insurance decomposes insurance premiums into two components reflecting how the premiums are used by insurance companies: earnings retained by the companies and benefits paid out on behalf of customers. Retained earnings are defined as the leftover premium income retained by insurance companies after the payout of benefits and are used to cover administrative costs or kept as profit. The CPI program views this component as the cost of administering insurance services (such as the payout of claims), and the CPI for health insurance implicitly estimates this cost.
The price of health insurance services is defined as the ratio of retained earnings to benefits paid out. For example, if retained earnings rise because of increased premiums and benefits paid out remain constant, the price of health insurance services would increase. To control for changes in medical care utilization, the CPI program defines health insurance benefits in real terms. The index for health insurance services from period t − 1 to period t is given by the following equation:
where REt and REt−1 are retained earnings in, respectively, periods t and t − 1; Bt and Bt−1 are health insurance benefits paid out in, respectively, periods t and t − 1; and and
are price indexes for health insurance benefits in, respectively, periods t and t − 1 (these indexes are based on the price indexes for noninsurance medical goods and services).
The preceding equation has to be adjusted to reflect that the data for retained earnings and benefits are annual whereas the price indexes for medical goods and services are monthly.1 The following equation shows how the monthly price relative for the index for health insurance services is calculated in practice:
where REy and REy−1 are retained earnings in, respectively, years y and y – 1; By and By−1 are health insurance benefits paid out in, respectively, years y and y – 1; wi is a weight; and and
are the medical CPIs in, respectively, months t and t – 1. The monthly relative for health insurance services between months t and t − 1 is calculated by using the 12th root of the annual relative for the retained-earnings-to-benefits ratio. This value is multiplied by the monthly relative of the price indexes used to deflate the benefits in the retained-earnings-to-benefits ratio in order to generate the real value of the benefits. The index for health insurance benefits is a weighted average of the MCPIs for different categories of medical goods and services (the weights used in the calculation represent the shares of benefit expenditures on each category).
CPI methodology reallocates the health insurance expenditure weights associated with benefits paid out because the index for health insurance captures only the price of health insurance services. The weight representing the share of total-premium expenditures that go toward paying benefits is reallocated to the MCPIs on the basis of the share of benefits that go toward each service type, for each type of insurance. After the reallocation of benefit expenditures, the noninsurance medical care indexes carry the weight of both patient out-of-pocket payments and insurance reimbursements for medical care. Therefore, to be consistent with the scope of the weights, each of these noninsurance indexes must track the price relatives of the total reimbursement (the sum of insurance reimbursements and patient out-of-pocket payments).
The PPI program uses direct pricing for its commodity and industry health insurance indexes.2 Unlike the indirect pricing used in the CPI, this alternative method directly prices total health insurance premiums. The PPI program collects premium prices directly from insurance companies. The insurance plans that are priced are sampled randomly by plan type (group managed care, group fee for service, individual comprehensive, dental, Medicare supplemental, other medical service, and indemnity and other health insurance).
To capture pure price changes, the PPI program must hold plan quality constant over time. The program attempts to control for all risk factors that might affect price in order to produce constant-quality price measures. The price-determining factors may vary by policy type and insurer. In general, for individual and group policies, the most common factors that are held constant during repricing include actuarial value, type of coverage, amount of coverage, group composition, occupation, earnings, smoking status, and geographic location.
Each premium price collected by the PPI program is the total premium received by an insurance company for the sale of a policy, including an employer’s contribution to employer-provided health insurance. The PPI for health insurance is not a pure measure of premium inflation because it is adjusted for earned returns on the invested portion of a premium. BLS publishes a commodity PPI for health insurance and an industry PPI for direct health and medical insurance carriers. These indexes are almost identical measures of premium inflation. (See chart 1.)
The direct and indirect methods for pricing health insurance both have advantages and disadvantages. BLS has been considering whether to continue using the indirect method for the CPI, asking the Committee on National Statistics (CNSTAT) to review the method and make a recommendation on its future use. In a final report issued in 2022, CNSTAT reviewed the strengths and weaknesses of different pricing methods and recommended that the CPI program continue with indirect pricing, for now.3
Two limitations of the indirect method are most relevant for the present study: (1) the CPI for health insurance based on indirect pricing is difficult for data users to interpret, and (2) the index is not directly comparable to other measures of health insurance prices. These limitations can be addressed by creating a total-premium inflation measure for the CPI, and in its 2022 report, CNSTAT recommended that BLS produce such a measure.
The indirect method is also limited by the fact that it yields an ex post inflation measure based on actual medical care prices and utilization, whereas insurance premiums are priced ex ante on the basis of expected prices and utilization. One implication of having an ex post measure is that the retained-earnings and price data used in the measure’s calculation are lagged. Another implication is that the retained-earnings-to-benefits ratio would fluctuate with unexpected changes in utilization, making the indirect measure more volatile. In its 2022 report, CNSTAT recommended several improvements to the indirect method that could address these limitations. Starting in October 2023, two of these improvements (smoothing the retained-earnings data and increasing the frequency of retained-earnings data updates) were implemented in the CPI.
Direct pricing also has limitations. The method’s most significant limitation is the difficulty of obtaining a constant-quality measure of total premiums. As noted earlier, plan features and risk factors must be held constant, and the data collection required for providing constant-quality prices can increase respondent burden and lead to lower response rates. Another limitation of direct pricing is its implication for the pricing of noninsurance medical goods and services. Under direct pricing, the relevant price is the out-of-pocket price instead of the total reimbursed price, which data users tend to view as the more relevant price for these categories.
Comparing movements in the CPI for health insurance (which covers only the insurance services component of the total premium) with movements in directly priced health insurance indexes (which measure total-premium inflation) is complicated because these indexes measure fundamentally different things. Chart 2 shows 12-month percent changes in the CPI for health insurance and the industry PPI for health insurance. As seen in the chart, insurance services inflation as measured by the CPI is much more volatile than total-premium inflation as measured by the PPI. The CPI series reflects the impact of using annual retained-earnings data, with the changeover to a new year of retained-earnings data generating inflection points in the series. The CPI series also shows periods of very high inflation.
Because insurance companies spend most premium income on medical goods and services, the prices of these goods and services have a large impact on total-premium inflation. Chart 3 shows the year-over-year changes in the indexes for noninsurance medical goods and services. Inflation for these categories tends to be less volatile than that for insurance services, and this lower volatility contributes to the relative stability of total-premium inflation.
In its 2022 report, CNSTAT noted the challenges of comparing the results of indirect and direct pricing, recommending that BLS produce a total-premium inflation measure for the CPI.4 Using the CPI for health insurance and the indexes for noninsurance medical goods and services, one can create an index for implied total-premium inflation in the CPI. This implied index is a weighted average of the index for health insurance services, IS, and the MCPIs for noninsurance medical goods and services:
where wS is a weight representing the share of premiums kept as retained earnings, and wB is a weight representing the share of premiums spent on benefits. The weights wi represent the shares of benefit expenditures on each of nine categories of noninsurance medical goods and services. Because the weights are fixed, the resulting total-premium index is a Lowe index (it is only an approximate measure of true premium inflation because the weights could change). Given that the CPI for health insurance is composed of four subindexes and the weights vary by type of insurance, the calculation of the total-premium index should be performed at the subindex level. Also, to allow a comparison with the PPI, one must exclude the subindex covering Medicare Part B, which is not included in the PPI for health insurance.
Using historical national-level CPIs for health insurance (IS) for the different insurance subindexes, along with national-level indexes for medical goods and services, I construct a CPI total-premium index. The health insurance information is at the national level. Although the prices of medical goods and services vary at the local level, this variation is unlikely to have a large impact on the overall total-premium measure. Therefore, the measure’s aggregation is based on national-level indexes and weights.
The weights wi are based on the factors used to reallocate total-premium expenditures to the medical goods and services categories and are the same as the weights used in the calculation of the CPI for health insurance. The weights for the share of premiums that go toward retained earnings and benefits are calculated as part of the biennial Consumer Expenditure Surveys updates to expenditure weights. The weight ws is the weight that remains associated with the CPI for health insurance after the reallocation of benefit expenditures, and the weight wB reflects the premium expenditures that are reallocated. A total-premium relative is calculated for each of the four subindexes of the CPI for health insurance, and the subindexes are then aggregated into a single index for total-premium inflation. The weights used to aggregate the subindexes are based on total out-of-pocket premium expenditures. Because each subindex captures multiple types of health insurance (with different weights for reallocating total premiums), the weights used to construct the total-premium index represent averages across insurance types.
In this section, I present results for the CPI total-premium index. Chart 4 compares this index with the CPI for medical care, the industry PPI for health insurance, and the CPI for health insurance. From December 2005 to December 2022, the CPI total-premium index increased 77.9 percent, less than the CPI for health insurance (89.1 percent) but more than the CPI for medical care (67.8 percent) and the industry PPI for health insurance (64.1 percent).
In chart 4, the impact of the COVID-19 pandemic on the CPI for health insurance is first reflected in the October 2021 index value, which is the first to capture 2020 data. In 2020, during the pandemic, the utilization of nonurgent medical care fell substantially, leading to a large increase in retained earnings. In the following year, retained earnings fell sharply as utilization increased. Because the series ends in December 2022, only 3 months of this large decrease in retained earnings are reflected in the CPI for health insurance. However, large declines in the index would continue through September 2023 (the index value was 133.6 as of September 2023).
Total-premium inflation can be broken down into two components: inflation in insurance services (measured by the CPI for health insurance) and inflation in medical benefits. Chart 5 shows 12-month percent changes in the CPI total-premium index and its components. As seen in the chart, the CPI for health insurance is much more volatile than the index for medical benefits, but it has less weight in the CPI total-premium index. As a result, the total-premium index tracks more closely with the benefits index. However, the gap between the two indexes (i.e., between total-premium inflation and benefits inflation) increases when the CPI for health insurance is at extreme values.
Chart 6 compares 12-month percent changes in the CPI total-premium index and the medical CPI. Although the two indexes track fairly closely with each other, total-premium inflation tends to be higher than medical care inflation. The close movement of the indexes is not surprising because both of them are aggregated from the same underlying indexes (CPIs for medical goods and services). However, the two indexes differ in the weights of their underlying indexes. Because the CPI total-premium index does not include out-of-pocket expenditures on medical goods and services, the CPI for health insurance has a larger weight in it than in the CPI for medical care.
Chart 7 removes the CPI health insurance component from the CPI total-premium index and the CPI for medical care, revealing how the two indexes, and the gap between them, are affected by the different weights of the indexes for noninsurance medical goods and services. As seen in the chart, the benefits index trends higher than the medical CPI excluding health insurance. The reason for this difference is that hospital services, which have exhibited persistently higher inflation than other medical care categories, are more heavily weighted in the total-premium index.
Table 1 shows the weights of the subindexes in both the CPI for medical care and the CPI total-premium index for December 2019. These weights represent 2017–18 expenditure shares and are the most recent weights available prior to the COVID-19 pandemic distorting medical care utilization. The table shows that the subindexes for health insurance services, hospital services, and physician services have higher weights in the CPI total-premium index than in the CPI for medical care. The reverse is true for all other index categories. The subindex for nonprescription drugs has a 4.36-percent weight in the CPI for medical care but a 0.00-percent weight in the CPI total-premium index, a difference due to the fact that expenditures on nonprescription drugs are entirely out of pocket (none of the reallocated premium weights is assigned to this category). Because of these weighting differences, the CPI total-premium index increases faster than the CPI for medical care. The last column of table 1 shows average annualized index changes from December 2005 to December 2022. For this period, the subindex for hospital services, which is more heavily weighted in the CPI total-premium index, exhibits the highest average inflation rate. By contrast, the subindex for nonprescription drugs, which has no weight in the CPI total-premium index, exhibits the lowest inflation rate.
Kategorie | Share in CPI total-premium index | Share in CPI for medical care | Average annualized index change, December 2005–December 2022[1] |
---|---|---|---|
Medical goods | |||
Prescription drugs | 8.99 | 13.40 | 2.43 |
Nonprescription drugs | 0.00 | 4.36 | 0.15 |
Medical equipment and supplies | 0.32 | 0.84 | 0.50 |
Noninsurance services | |||
Physician services | 28.19 | 20.50 | 2.15 |
Dental services | 3.41 | 11.21 | 3.17 |
Eyeglasses and eye-care services | 2.09 | 4.18 | 1.05 |
Other professional services | 3.23 | 5.37 | 1.81 |
Hospital services | 33.90 | 24.75 | 5.07 |
Nursing homes | 0.83 | 1.38 | 3.50 |
Care of elderly at home | 0.03 | 0.78 | 2.26 |
Health insurance | 19.01 | 13.24 | 3.82 |
[1] The average annualized changes for nonprescription drugs and medical equipment and supplies are for the December 2009–December 2022 period because the published versions of the subindexes for these categories start in December 2009. Note: CPI = Consumer Price Index. Source: U.S. Bureau of Labor Statistics. |
For the entire 2005–22 period covered in table 1, the CPI total-premium index increases only slightly more than the industry PPI for health insurance, but the inflation rates of the two indexes do not track closely with each other. This divergence may be due to the indexes having different scope. To close the difference in scope, I remove the health insurance category that includes Medicare Part B from the CPI total-premium index, because Medicare Part B is not included in the industry PPI for health insurance.
Chart 8 plots 12-month percent changes in the CPI total-premium index, the CPI total-premium index excluding Medicare Part B, and the industry PPI for health insurance. As seen in the chart, the CPI total-premium index excluding Medicare Part B trends similarly to the overall CPI total-premium index, which suggests that differences in scope are not primarily responsible for the divergence in CPI and PPI inflation patterns. An alternative explanation for this divergence may be the lagged data used in the construction of the CPI for health insurance. Peaks and troughs in the industry PPI for health insurance tend to be followed by peaks and troughs in the CPI total-premium index, and the lag between these index movements (1–2 years) is consistent with the lag in the data. However, some movements in the CPI total-premium index have no corresponding movements in the industry PPI for health insurance, a discrepancy that may partly be explained by the fact that the PPI program adjusts total-premium inflation for investment returns.
An indirect measure of health insurance inflation can reveal premium changes before they appear in a premium measure based on direct pricing. Unexpected shocks in medical care utilization show up in the CPI for health insurance despite the absence of changes in total premiums. After such shocks, insurance companies adjust premiums to restore their reserves. As noted earlier, the CPI for health insurance is an ex post measure based on realized prices and utilization, whereas total premiums are set by insurance companies ex ante on the basis of expected prices and utilization. In its 2022 report, CNSTAT argued that an ex ante measure of health insurance is the appropriate measurement target, suggesting that smoothing multiple years of retained-earnings data would allow an ex post measure to more accurately reflect an ex ante measure. In October 2023, the CPI program shifted its health insurance methodology to a 2-year moving average.5 The program also adopted the use of half-year retained-earnings data, a methodological change that would shorten the lag with which the data first enter an index. Combined, these changes will likely lead the CPI total-premium measure to more closely track total premiums as measured by the PPI.
The CPI for health insurance measures the price of health insurance services other than indirectly purchased medical goods and services. This index is difficult for the public to interpret because the price most directly associated with health insurance is that reflected by the total insurance premium. However, a measure of the total premium can be constructed from CPI data. In this article, I have constructed an implied CPI total-premium index. From December 2005 to December 2022, the average annual growth in this implied measure was slightly higher (3.45 percent) than growth in the industry PPI for health insurance (2.96 percent) and the CPI for medical care (3.09 percent).
The CPI total-premium index tracks closely with the CPI for medical care because it is a weighted average of the same components (albeit with different weights) as those included in the medical index. Total-premium inflation tends to be higher than medical care inflation because the hospital services index, which increased the most among all medical CPIs from December 2005 to December 2022, is more heavily weighted in the CPI total-premium index than in the CPI for medical care. Data users can use the CPI for medical care as an approximation of total-premium inflation.
The total-premium measure presented in this article is preliminary and can be improved. First, the aggregation used to construct the measure is at the national level, but it can be implemented at the regional level by using weights that reflect regional differences in total-premium spending. Second, the indexes for medical goods and services include noninsurance payers, but when one considers inflation in health insurance benefits, these payers should be excluded from the CPIs for noninsurance medical care.
Brett Matsumoto, "Measuring total-premium inflation for health insurance in the Consumer Price Index," Monthly Labor Review, U.S. Bureau of Labor Statistics, April 2024, https://doi.org/10.21916/mlr.2024.6
1 Starting in October 2023, the retained-earnings component was switched from an annual relative to a relative based on a 2-year moving average of retained-earnings data.
2 For a description of the Producer Price Index methodology, see “Producer Price Index for the direct health and medical insurance carriers industry—NAIC 524114” (U.S. Bureau of Labor Statistics), https://www.bls.gov/ppi/factsheets/producer-price-index-for-the-direct-health-and-medical-insurace-carriers-industry-naics-524114.htm.
3 Modernizing the Consumer Price Index for the 21st century (Washington, DC: The National Academies Press, 2022), https://doi.org/10.17226/26485.
4 Recommendation 5.2 (see ibid.) states that the U.S. Bureau of Labor Statistics (BLS) should produce a total-premium deflator for the Consumer Price Index and compare the historical movements of the index with direct measures of premium inflation. Recommendation 5.3 states that BLS should publish the total-premium measure.
5 “Improvements to the CPI health insurance index” (U.S. Bureau of Labor Statistics, 2023), https://www.bls.gov/cpi/additional-resources/improvements-cpi-health-insurance-index.htm.