The following section describes the data analysis processes of this report, including weighting and statistical testing.

Weighting

Given the method used to obtain the sample, probabilities of selection are not available. However, it is possible to derive survey weights by appropriately representing each filing’s relative size from the 2011 NAIC (number of membermonths by carrier), 2010 MEPSIC (estimated enrollment distribution by state by product for the small group market), and, in some cases, carrier filings (number of reported members). We combined the individual and conversion markets for weight calculation and refer to them in this discussion as the “individual market”. The final weights represent the contribution to the estimates for each filing.
A multistep process was used to calculate weights. The first six steps implemented at the “statemarketyear” level – (that is, each observation with a specific combination of those three variables gets its own weight). The last three steps are calculations at the level of “marketyear”. The list below, as well as Tables 4 and 5, summarize each step used for weighting.
 Initial carrier weights were generated (to reflect the carrier’s relative contribution to the estimates);
 Applied withincarrier filing adjustments to the initial carrier weight (to adjust for multiple filings by a carrier);
 Applied statelevel product adjustments to the initial carrier weights (small group market only) (to adjust for product enrollment distributions);
 Applied withincarrier enrollment adjustments to the initial carrier weights (to adjust for the relative size of each filing);
 Applied weight control adjustments to the initial carrier weights (to control the weights to sum to one within a stateyear);
 Calculated final statelevel weights (for use in deriving statelevel estimates);
 Applied nationallevel adjustments (to adjust for the relative size of each state);
 Applied nationallevel singlefiler adjustments (to control the influence of singlefilers within a state);
 Calculated final nationallevel weights (for use in deriving statelevel estimates).
Initial Carrier Weights: The source for the initial carrier weights is the 2011 NAIC supplemental data file. We used information on membermonths from NAIC to assign initial carrier weights to reflect the relative contribution to the estimates by carrier within strata defined by market (individual, small group), year, and state. Initial carrier weights are defined as:
where
M_{TYSi }= number of membermonths reported from the 2011 NAIC for sample carrier i reporting in year Y from market type T in state S
WithinCarrier Filing Adjustments: As each carrier may have multiple filings within a market type/year/state, the initial carrier weight must be adjusted to reflect the number of filings within each carrier so as not to overrepresent carriers with multiple filings.^{29 }The withincarrier filing adjustment is defined as:
where
n_{TYSi} = number of filings for carrier i from market type T for year Y in state S
StateLevel Product Adjustments (small group market only): For the small group market, an adjustment to the survey weights is made so that distributions of the resulting survey weights reflect estimated enrollment distributions from MEPSIC by state and product. (This information is not available for the individual market.) The sum of the withincarrier filing adjusted weights by product type is adjusted to reflect the MEPSIC distributions. The product adjustment is defined as:
where the sum in the first term of the denominator is across all filings for product type P for which small group filings G^{30} were obtained for year Y from state S, and the sum in the second term is across all filings for which small group filings were obtained for year Y from state S. This latter term is used to scale the first term to sum to 1.0.
estimated enrollment distribution of the small group market for product type P (relative to product types for which filings were obtained in year Y) within state S from the 2010 MEPSIC data; f refers to a filing obtained from sample carrier i reporting in year Y from market type G (see footnote 28, infra) in state S
For example, from Table 4 ID=234, the numerator=0.258, the first part of the denominator is the sum of filing adjusted weights for P=HMO (0.803314), and the second part of the denominator is the sum of all filing adjusted weights (1.0). The formula then is 0.258 / (0.803314 / 1) = 0.320880, which is Column I in Table 4.
Not all small group filings had their product type identified. As a result, no adjustment is applied to those specific filings.
WithinCarrier Enrollment Adjustments: The survey weights are further adjusted to reflect the relative size (if known) of each filing for a carrier. Each filing contained information on either the number of covered members, the number of contracts, both, or neither. Using filings with both the number of covered members and the number of contracts, an estimate of the number of covered members was imputed for those filings with only the number of contracts. The withincarrier product adjustment is defined as:
where
E_{TYSPi} = number of members reported on filing f from individual carrier i for product P from market type T for year Y in state S
n_{TYSPi} = number of filings for carrier i for product P from market type T for year Y in state S
StateLevel Weight Adjustments
As the sum of the preliminary survey weights are not constrained to equal 1.0, the weights must by adjusted so as to control the sum of the survey weights to be equal to 1.0.
For the individual market, the statelevel weight adjustment is defined as:
For the small group market, the statelevel weight adjustment is defined as:
Final StateLevel Weights
The final statelevel survey weight for the individual market can thus be defined as the product of the initial carrier weight and the adjustments made for the individual market:
SW_{IYSPif} = IW_{IYSi} * CFA_{IYSi }* CEA_{IYSPif} * SWA_{IYSPif}
The final statelevel survey weight for the small group market can thus be defined as the product of the initial carrier weight and the adjustments made for the individual market:
SW_{GYSPif }= IW_{GYSi} * CFA_{GYSP }* SPA_{GYSP }* CEA_{GYSPif }* SWA_{GYSPif}
These final statelevel survey weights sum to one with a state/market/year.
NationalLevel Adjustments
We applied a national adjustment to the final statelevel survey weights to reflect the relative sizes across states within a market type (Table 5). The national adjustment is defined as:
where
S’ = set of sample states for which more than one filing was obtained for year Y
M_{TS' }= number of membermonths reported from the 2011 NAIC for all carriers from market type T in state S’
NationalLevel SingleFiler Adjustments
Given the uncertainty associated with estimates from states with only one filer a final adjustment is applied so as to have those single filers included in the estimates but representing only themselves. This is accomplished through separate adjustments being applied to the single filers and all other filers. Note that this adjustment factor applies to only four filings in the individual market in 2011 and two in the individual market in 2012.
The nationallevel single filer adjustment is defined as:
S” = set of sample states for which only one filing was obtained for year Y
M_{TSY }= number of membermonths reported from the 2011 NAIC for all carriers from market type T in state S (the set of sample states for which filings were obtained for year Y)
M_{TS'Yi }= number of membermonths reported from the 2011 NAIC for sample carrier i reporting in year Y from market type T in state S”
Final NationalLevel Weights
The final nationallevel survey weight is then defined as:
NW_{TYSPif} = SW_{TYSPif} * NA_{TSY }* NSA_{TSYi}
Table 4: StateLevel Weight Calculation for 2012 Connecticut Small Group Market
ID (A) Carrier (B) Product (C) Carrier MemberMonths from NAIC (D) Initial Carrier Weight (E) Number of Filings within Carrier (F) WithinCarrier Filing Adjustment Factor (G) Filing Adjusted Weight MEPSIC Distribution (H) Adjusted MEPSIC Distribution State Product Adjustment (I) Product Adjusted Weight Enrollment from Filing (J) WithinCarrier Enrollment Adjustment (K) Preliminary Weight StateLevel Weight Adjustment (L) Final StateLevel Weight (M) Total 2,968,503 1.525 0.968 0.931 0.788 1.000 213 11209 HMO 95,747 0.032254 2 0.500 0.016127 0.2226 0.2578 0.3173 0.005117 13237 1.9735 0.010098 1.268 0.012618 214 11209 PPO 95,747 0.032254 2 0.500 0.016127 0.6409 0.7422 3.9567 0.063811 178 0.0265 0.001693 1.268 0.002116 222 60217 HMO 989,174 0.333223 1 1.000 0.333223 0.2226 0.2578 0.3173 0.105727 55582 1.0000 0.105727 1.268 0.132115 231 78026 HMO 554,159 0.186680 1 1.000 0.186680 0.2226 0.2578 0.3173 0.059231 45093 1.0000 0.059231 1.268 0.074014 236 95675 HMO 700,295 0.235908 2 0.500 0.117954 0.2226 0.2578 0.3173 0.037425 42498 1.6292 0.060975 1.268 0.076194 237 95675 PPO 700,295 0.235908 2 0.500 0.117954 0.6409 0.7422 3.9567 0.466714 9671 0.3708 0.173037 1.268 0.216226 240 95935 HMO 317,663 0.107011 4 0.250 0.026753 0.2226 0.2578 0.3173 0.008488 5961 0.5258 0.004463 1.268 0.005577 241 95935 PPO 317,663 0.107011 4 0.250 0.026753 0.6409 0.7422 3.9567 0.105854 26262 2.3165 0.245214 1.268 0.306417 242 95935 HMO 317,663 0.107011 4 0.250 0.026753 0.2226 0.2578 0.3173 0.008488 1864 0.1644 0.001396 1.268 0.001744 243 95935 PPO 317,663 0.107011 4 0.250 0.026753 0.6409 0.7422 3.9567 0.105854 11260 0.9932 0.105137 1.268 0.131378 247 96798 HMO 311,465 0.104923 1 1.000 0.104923 0.2226 0.2578 0.3173 0.033291 27084 1.0000 0.033291 1.268 0.041600 N/A N/A IND N/A N/A N/A N/A N/A 0.1365 N/A N/A N/A N/A N/A N/A N/A N/A Table 5: NationalLevel Weight Calculation for 2012 Connecticut Small Group Market
ID (A) Carrier (B) Product (C) State (D) Final StateLevel Weight (E) Carrier MemberMonths from NAIC (F) State MemberMonths from NAIC (G) NationalLevel Adjustment (H) NationalLevel Single File Adjustment (I) Final NationalLevel Weight (J) 138,779,122 213 11209 HMO CT 0.012618 95,747 3,640,884 0.026235099 1 0.000331 214 11209 PPO CT 0.002116 95,747 3,640,884 0.026235099 1 0.000056 222 60217 HMO CT 0.132115 989,174 3,640,884 0.026235099 1 0.003466 231 78026 HMO CT 0.074014 554,159 3,640,884 0.026235099 1 0.001942 236 95675 HMO CT 0.076194 700,295 3,640,884 0.026235099 1 0.001999 237 95675 PPO CT 0.216226 700,295 3,640,884 0.026235099 1 0.005673 240 95935 HMO CT 0.005577 317,663 3,640,884 0.026235099 1 0.000146 241 95935 PPO CT 0.306417 317,663 3,640,884 0.026235099 1 0.008039 242 95935 HMO CT 0.001744 317,663 3,640,884 0.026235099 1 0.000046 243 95935 PPO CT 0.131378 317,663 3,640,884 0.026235099 1 0.003447 247 96798 HMO CT 0.041600 311,465 3,640,884 0.026235099 1 0.001091 ^{29 }While we aggregate conversion filings with those from the individual market for the purposes of analysis, we do not expect these conversion filings to measurably impact the findings as they are few in number (approximately 11% of the individual market sample by count) and they tend to have low enrollments.
30 In this case, G is a constant – the market type T is either small group (G) or individual (I), but the product adjustment is only possible for the small group (G) market type.


Imputations

Values for item nonresponse were not imputed. The weighting mechanism described above does make adjustments for nonresponding carriers in any given year. Most important, we have not imputed any values for dependent variables – premium increases or approval by the state regulatory authority.


“Any Size Group” Filings

Some filings were filed in the SERFF system under the designation “any size group.” Further examination showed that some but not all of the business covered in these filings was in the small group market. Following discussion with ASPE and analysis by the NORC team, “any size group” filings were only included in the study if they were submitted by a sampled carrier but no other data (i.e. small grouponly filings) from that carrier’s business in the given state and year were available. Because most “any size group” filings did not separate reported enrollment figures into small group and large group components, enrollment data for these observations were considered missing. As a result of this procedure, 53 “any size group” filings were incorporated in our analysis.


Adjustments to Filings with Greater Than Annual Increase Periods

For filings where the period between the effective date of the approved premium increase and the effective date of the last approved premium increase was greater than one year, adjustments were made to the increase such that they represented an annual period of increase, using exponentiation to account for compounding effects. For example if an insurer requested a 30% increase after three years of stable rates, this rate would be adjusted down to 9.1% for the year, which is 1.30 raised to the 1/3 power minus one.


Statistical Testing

We conducted descriptive analyses to address the study research questions. Ttests were used to determine whether means were significantly different, and estimates of variance were corrected for using finite population correction. The paper presents national and statewide results. We examine multiyear trends for dependent variables, and analyze variations in dependent variables by selected independent variables. Dependent variables are:
 Premium increases
 Percent of rate increases approved by state regulators
 Percent of premium rate increases modified by state regulators
We present findings separately for the individual insurance and small group markets. Key covariates are:
 State regulatory authority
 Prior approval
 File and use
 HMO review authority only
 Rate review authority
 No requirement for filing
 Product type (HMO, PPO/HDHP, indemnity)
 Carrier size (top three carrier in the state and market, other)
 Market concentration in the individual and small group markets
 High – Largest three carriers in state have 80 or more percent of the market
 Medium – Largest three carriers in state have 5079 percent of the market
 Low – Largest three carriers in state have less than 50 percent of the market
For statelevel estimates, we required filings to encompass a minimum proportion of 50 percent of NAICreported statewide enrollment to report results for a given state and year to ensure reliability. Multivariate modeling is beyond the scope of this study.

View full report
"20130607InsMktReformReportFnl.pdf" (pdf, 1.46Mb)
Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®