Font Size:

Respondent Payment Experiment with MEPS Panel 13

October 6, 2010

The Medical Expenditure Panel Survey (MEPS) is a nationally representative study of health care use and expenses, sponsored by the Agency for Healthcare Research and Quality (AHRQ) and the Centers for Disease Control & Prevention (CDC). The sample selected for MEPS for a given panel is a subsample of those households that participated in the National Health Interview Survey (NHIS) conducted by CDC during the prior year. The household component of the MEPS has an overlapping panel design, in which a new panel of households is introduced each year and interviewed 5 times over a 2.5 year period, providing data for two full calendar years. Data from the MEPS allow those in health care policy to study how changes in the structure of American health care delivery, private insurance, Federal health care programs, and the demographic composition of the country affect:

the kinds and amounts of healthcare Americans use,
how the care is paid for,
how much is paid for it, and
the implications for changes in health care policy.

This report discusses the results of an experiment designed to compare the effect of different respondent payment amounts offered to MEPS reporting units (RUs) within the household component (HC) of MEPS. The experiment was implemented for MEPS Panel 13, first fielded in 2008. This report covers results from all five rounds of data collection for Panel 13.

Return To Table Of Contents

Background

With each successive year the household component of the MEPS has experienced increased difficulty in obtaining cooperation for the first round of data collection and maintaining response and retention rates across all rounds of data collection. Lower response and retention rates across rounds can potentially affect bias in final survey estimates. For example, Cohen et al (2006) discuss how attrition across rounds affects estimates of health insurance coverage for 2001-2003.

To encourage survey participation, data collection procedures for the household component have included a monetary gift to respondents in appreciation for the time and effort they spend keeping records and participating in the survey. Since 2007, the amount of the respondent gift, or payment has been $30 per household per interview. In general, the use of respondent payments in survey research has become fairly common, and survey practitioners agree that respondent payments increase overall response rates (Goyder, 1994; Willimack, Schuman, Pennell and Lepkowski, 1995; Singer, 1999, 2002).

At the request of the Office of Management and Budget, AHRQ introduced an experiment for the 2008 MEPS panel that compared the effect of three different respondent payment amounts ($30, $50 and $70) on three different categories of measures:

cooperation with the survey request, as measured by response rates and refusal rates
the level of effort needed, and thus field costs necessary to obtain cooperation in each round, and
the quality of the response data

Each sampled household in the 2008 panel was randomly assigned to one of the three different levels of payment. In all cases, the field interviewer provided the respondent payment at the completion of the interview in each round. The full panel sample of 9,939 households was included in the experiment, and the results through all five rounds of MEPS data collection are reported.

Return To Table Of Contents

Experimental Design and Implementation

All households in the Panel 13 sample were randomly assigned to one of three respondent payment groups. The $30 group served as the control group since MEPS offered that amount to responding households starting in 2007 (prior to that, a respondent payment of $25 was offered). Assignments to groups were made at the NHIS segment level to help reduce, but not eliminate, the risk that neighboring sample households in the same MEPS panel would receive different amounts (sampled households in MEPS panels 12 and 14 had only the $30 respondent payment amount but were located in the same segments as Panel 13).

Each MEPS panel represents a subsample of the NHIS sampled households, and the MEPS subsample oversamples various domains of analytic interest. Membership in a domain is based on at least one member having a given characteristic, reflecting race/ethnicity or income status (2002-2008). Income status is based on a model where households are "predicted to be poor" or not, based on data available at the time of sample selection. If multiple characteristics are associated with a given household, the "highest priority" characteristic is chosen to represent the household. For Panel 13 the domains, in priority order were Asians, "predicted to be poor", Hispanic, Black, and Other (basically, those white and not "predicted to be poor"). We will refer to the "predicted to be poor" as low income for the purpose of these discussions¹. From past history those households characterized as Hispanic, Black, or Low Income had substantially higher unweighted response rates than Asians or Others.

Statistical staff categorized the segments into one of two strata based on expected response propensity, prior to assignment to the respondent payment condition. Segments with a majority of black, low income, and Hispanic households were assigned to the "expected higher response" stratum and the remaining households were assigned to the "expected lower response rate" stratum. Allocation to respondent payment groups by response propensity strata was proportional to the sample distribution of MEPS households. Roughly, 35 percent of the households in the MEPS sample were in the "expected higher response rate" stratum, so roughly 35 percent of the segments in each respondent payment group came from the "expected higher response" stratum.

At AHRQ's request, Westat allocated slightly more cases to the $30 respondent payment group as compared to the other two groups. The experiment used this allocation strategy in an effort to obtain close to the same level of precision for each group since we expected the response rates for the $30 group to be lower than the other two groups. The table below shows the assignment to respondent payment group for the full sample, as well as for each of the sampling domains.

Panel 13 Assignment to Respondent Payment Group
Respondent Group	$30	$50	$70	Total
Asian	364	235	267	866
Low Income	725	614	571	1,910
Hispanic	751	599	538	1,888
Black	541	490	477	1,508
Other	1,392	1,166	1,209	3,767
Full sample	3,773	3,104	3,062	9,939

¹The "predicted to be poor" or "low income" domain as labeled here was dropped as a sampling domain in 2009 and 2010.

To enhance comparison of the results from the experiment with prior MEPS panels, project staff restricted procedural changes for the year in which the experiment was fielded to those necessary for implementing the experiment. In general, field procedures and operational activities remained unchanged from those used in the previous panel. Pre-field activities, including advance letter mail outs, advance contact calls, and assignment material preparation remained unchanged from prior years, other than updating the respondent payment amount for those households in the $50 and $70 groups. Home office tracking, disseminating information from the respondent calls to the help line, mailing of refusal letters, and other data collection support activities also remained the same as with the previous panel. This was done so that differences detected in the response rates or other evaluative measures by respondent payment group could more readily be generalized beyond this panel.

Procedural changes necessary for the implementation of the respondent payment experiment included only the following minor adjustments to the case materials and reporting forms:

labels on case folders and RU folders contained a code to indicate the respondent payment amount.
the interviewer's weekly status report and the interviewer assignment sheet were both updated to indicate the respondent payment amount.
the check for the appropriate amount was included in each case folder to reduce the risk of paying the respondent the wrong amount.
in training, interviewers were reminded to pay attention to the version of the advance letter handed to respondents in Panel 13 since the letter indicated the payment amount.

In addition, home office receipt processing included a step to verify that interviewers were implementing the respondent payment amounts as assigned. About a dozen households in Panel 13 received the incorrect respondent payment amount and thus are excluded from this analysis.

To avoid any possibility of influencing the outcome of the experiment, home office and field supervisors and managers were blinded to the production and response rate status by respondent payment group throughout the Round 1 field period. Although the respondent payment amount for each case was clearly visible on the round 1 materials, staff monitoring production did not view outcomes by respondent payment group until the end of the data collection round. In subsequent rounds, home office did track response outcomes by respondent gift amount.

Return To Table Of Contents

Analytic Measures

As noted above, the experimental analysis addresses three evaluative categories: cooperation, level of effort and data quality.

Return To Table Of Contents

Cooperation

The primary measures of cooperation are weighted response rates and refusal rates by sampling domain and across the full sample. Response and refusal rates were calculated by individual round and a composite weighted response rates across all 5 rounds was also calculated. The weights reflect the probability of selection but no other adjustments. The analysis examines whether the higher level respondent payments result in improved response rates and diminished refusal rates relative to the $30 payment. The response rate and refusal rate calculations use edited data that reflect updates in an RU status (i.e., complete, refused, ineligible, etc) for each round, based on the knowledge gained about that RU and the persons within the RU across the first full year of data collection.

Round 1 response rates include two components in the calculation. The first component of the overall Round 1 response rate reflects the degree of participation among those RUs found in the sampled households from the NHIS. The second component of the response rate accounts for the new RU's that occur with the MEPS data collection. This component is needed to account for nonrespondents in any new RU's that MEPS misses in the field because the parent RU is a nonrespondent. The Round 1 response rate is the product of both components.

The response rate calculation for subsequent rounds is essentially the same but takes into account whether a case is newly fielded in the round, as is the case with split or student RUs.

Refusal rates reflect a simple ratio of cases coded as final refusals over all eligible cases for the round. The analysis also looks at the proportion of cases that ever gave a refusal, even if later converted to a complete in a round.

Additionally, the analysis includes two other measures of cooperation from just the responding Round 1 RU's. In an advance mailing just prior to Round 2, eligible persons within RUs are asked to complete a self-administered questionnaire and return it to the interviewer at the round 2 interview. Interviewers also accept completed SAQ's returned in the next round. The analyses assesses whether the proportion of RU's that complete and return all eligible SAQ's within an RU, complete at least some of the eligible SAQ's within an RU, or do not complete any SAQs for the RU changes depending on the amount of the respondent payment.

As a last cooperation measure, the analysis assesses compliance in the first three rounds of data collection with the request to authorize MEPS to contact RU members' medical providers for the MEPS follow-up medical provider component. In each round of data collection, interviewers ask eligible persons within an RU to sign an authorization form that allows MEPS to contact his/her medical provider and obtain records indicating medical services received and the costs of those services. The analysis assesses whether cooperation with either of these requests, to complete the SAQ in round 2 or sign the medical authorization form (AF) in the first three rounds improves with the higher respondent payments relative to the $30 payment. While the response rates and refusal rates were calculated based on final edited data files, these data came from the status at initial receipt.

Return To Table Of Contents

Level of effort and estimated costs

The cost of data collection operations is difficult to control and is on the rise, in a large part because of the amount of effort needed to contact and resolve the sampled housing units. This analysis examines the extent to which the higher respondent payments reduces the average number of contacts required, in-person, by phone and overall, to contact and finalize the status of each sampled RU. In addition, the analysis looks at whether the amount of the respondent payment reduces the number of calendar days on average it takes to close a case once an interviewer starts working it.

Return To Table Of Contents

Data quality

The data quality assessment focuses on the first three rounds of data collection only since the full range of data editing and final processing are not yet complete for rounds 4 and 5. This analysis assesses the effect of the respondent payment level on data quality using the following metrics:

two survey estimates based on responses to MEPS questions in particular rounds of data collection
item nonresponse rates, or more specifically the proportion of eligible responses for a given item, that do not have a valid answer (i.e., missing, don't know or refuse). These data come from the unedited response files.
the distribution of several different demographic characteristics for the MEPS respondents. These serve as indicators of potential response bias. These distributions also come from the unedited response files.

Return To Table Of Contents

Results

All significance testing reflects test of proportions (t-tests) of each pair wise comparison, without adjustments for multiple comparisons. The tables and corresponding discussion cover statistically significant differences at α=0.01 and α=0.05 levels. All tables are at the back of the report.

Return To Table Of Contents

Weighted Response Rates (Tables 1-6)

In general, these data indicate that response rates do increase with the higher incentive payment amounts. Across the full sample in Round 1 (Table 2), both the $50 and $70 respondent payment amounts resulted in a significantly higher response rate (76.60% and 77.60% respectively) than the control group that received the $30 payment (71.60%). Results were the same in round 2 (Table 3) with the $50 and $70 groups (95.54%, 96.99%) both attaining significantly higher response rates than the $30 group (92.33%). In each of the final three rounds of data collection starting in the second year, only the $70 respondent payment resulted in a significantly higher response rate than the $30 payment group (Tables 4-6).

Ultimately, as shown in Table 1 below, the composite response rate across all five rounds of data collection for the full sample for both the $50 and $70 respondent payment groups (66.74%, and 71.13%, respectively) was higher than the composite response rate for the $30 group (58.84%). Additionally, the composite response rate for the $70 payment group was significantly higher than that of the $50 group.

Table 1: Composite response rate (R1-R5) by sampling domain and respondent payment group
Respondent Group	$30 Incentive Weighted RR	$30 Incentive s.e.	$50 Incentive Weighted RR	$50 Incentive s.e.	$70 Incentive Weighted RR	$70 Incentive s.e.
Asian	51.25⁺	3.22	62.32^$	3.45	63.35^@	3.13
Low Income	71.82^@	2.01	72.55	2.23	78.88^@	2.20
Hispanic	57.14⁺	2.11	65.52^!	2.33	73.15^!	2.35
Black	60.84⁺	2.51	70.03^@	2.24	74.39^@	1.90
Other	57.11⁺	1.53	65.71^@	1.55	69.97^@	1.55
All Domains	58.84⁺	1.13	66.74^{^}	1.19	71.13^{^}	1.12

Footnotes

⁺Significant at the 0.01 level and both $50 and $70 are significantly different from $30 at the 0.01 level
^$Significant at the 0.05 level
^!Significant at the 0.01 level and $50 and $70 are significantly different from one another at the 0.05 level
^@ Significant at the 0.01 level
^{^} Significant at the 0.01 level and $50 and $70 are significantly different from one another at the 0.01 level
^# Significant at the 0.05 level and $50 and $70 are significantly different from one another at the 0.05 level
^* $50 and $70 are significantly different from one another at the 0.01 level
^" $50 and $70 are significantly different from one another at the 0.05 level
^> Significant at the 0.05 level and both $50 and $70 are significantly different from $30 at the 0.05 level
^: Significant at the 0.05 level and $50 and $70 are significantly different from one another at the 0.01 level

In looking at the response rates for the individual sampling domains, the pattern of response rates by respondent payment level for the "other" domain, the largest sample domain and historically with the lowest response rate in the MEPS, mirrors that of the overall sample. In each of the first two rounds of data collection, the response rates among the "other" domain for both the $50 and $70 payment groups are significantly higher than the response rates in the $30 group, but for the last three rounds of data collection, only the response rate for the $70 respondent payment group was significantly higher than the $30 group. However, the composite response rate across all 5 rounds of data collection for both $50 and $70 respondent payment groups within the "other" domain was significantly higher than the $30 group (8.60 and 12.86 percentage points higher, respectively).

Like the "other" domain, the Asian domain also typically has one of the lowest response rates in the MEPS. Ultimately, in looking at the composite response rates as shown in table 1 above, both the $50 and $70 respondent payment groups resulted in noticeably higher response rates than did the $30 group for the Asian domain. Within the individual rounds (see tables 2-6), the response rates for the two higher respondent payment levels tended to be higher than those of the $30 payment group, but only in Round 2 did the difference in response rates between the $30 and the two higher respondent payment amounts attain statistical significance.

For each of the remaining three sampling domains, Low Income, Hispanic and Black, the $70 respondent payment group attained a composite response rate that was significantly higher (p≤0.01) than the $30 respondent payment group's composite response rate. The $50 respondent group also attains a significantly higher composite response rate relative to the $30 group, but for only the Hispanic and Black domains.

In addition to looking at the response rates by sampling domain, we also looked at the effect of the respondent payment amount on the response rates of cases sampled for the MEPS that had final outcome codes as "partial completes" in the NHIS. The NHIS partial complete cases consistently have lower response rates in the MEPS, higher refusal rates, and are more difficult and thus more costly to work. Table 7 shows the response rates for the NHIS partial completes and fully complete cases by respondent payment group for each round of data collection, and across all rounds. As can be seen in table 7, the composite response rate for both the $50 group (52.31%) and the $70 group (58.78%) are higher than the composite rate for the $30 group (45.97%) though only the 13 percentage point difference between the $30 and $70 group attains statistical significance. In general, in all but the 4^th round of data collection, the response rate for the $70 group was statistically significantly higher than the response rate of the $30 group. The response rate for the $50 group was also statistically significantly higher than the $30 group, but only in two of the five rounds, rounds 2 and 3. Overall, the $70 respondent payment resulted in greater improvements in the response rate for the NHIS partials, the hardest cases to complete, than did the $50 respondent payment relative to the control group.

To summarize the response rate findings, the response rate tables indicate significant gains in the overall or composite response rate for the full sample with the $70 respondent payment relative to the $30 group, an increase of 12 percentage points. In fact, the composite response rate for the $70 respondent payment group was four percentage points higher than the composite response rate for the $50 group, also a statistically significant difference. Similarly, in looking at the individual round level response rates for the full sample, only the $70 group continued to attain response rates higher than the $30 group through the last three rounds of data collection (p≤0.01).

In terms of subgroup differences in response rates, the composite response rate for both the $50 and $70 respondent payment groups were higher than the composite response rate for the $30 group for all sample domains other than low income, though the only statistically significant differences occurred with the Hispanic subdomain between the $50 and $70 groups. The subgroup, NHIS partials attained almost a 13 percentage point gain in the composite response rate with the $70 respondent payment relative to the $30 group, a statistically significant difference. The difference in the overall response rate for the NHIS partials receiving $50 and $30 did not attain statistical significance.

Return To Table Of Contents

Refusal Rates (Tables 8-12)

The refusal rates follow a very similar pattern to that of the response rates in terms of attaining statistically significant differences between the two higher respondent payment groups relative to the control or $30 respondent payment group. Across the full sample in round 1, the final refusal rate and the percent of RU's that refused at some time during the round were significantly lower in the two higher respondent payment groups relative to the $30 group. Specifically, the final refusal rate for the $70 group was 17.36%, for the $50 group it was 17.03%, but for the $30 group, the final refusal rate was 21.70% (Table 8). This is in accordance with the fact that the two higher respondent payment groups had significantly higher response rates relative to the $30 group, suggesting that the higher response rates in fact reflect lower refusal rates rather than drops in other types of noninterviews.

In round 2 (Table 9), across the full sample, the two higher respondent payment groups continued to achieve a statistically significant lower refusal rate as compared to the $30 group. However, in each of rounds 3, 4 and 5, only the $70 respondent payment group continued to achieve significantly lower refusal rates (and higher response rates) relative to the $30 group.

This same pattern occurs within the "other" domain, which historically has the highest refusal rates of all sampled domains. In each of the first two rounds of data collection the increased response rate in both the higher respondent payment groups relative to the $30 group reflects a significantly lower refusal rate in both the $50 and $70 respondent groups. However, for each of rounds 3, 4 and 5, only the $70 respondent payment group had a significantly lower final refusal rate relative to the $30 group within the "other" domain. The percent of cases in the "other" domain that ever refused during a round of data collection followed a very similar pattern looking across rounds. In the first round, both the $50 and $70 respondent payment groups had significantly lower proportions of cases that 'ever-refused' as compared to the $30 group. However, by the last two rounds of data collection, only the $70 group still had a significantly smaller percentage of RU's that 'ever-refused' during the round as compared to the $30 group.

Within the Asian domain, another domain with a traditionally high refusal rate, the respondent payment amount only affects the final refusal rate in the second round of data collection. In the second round both the $50 and $70 groups achieve a final refusal rate of about 2%, but the $30 respondent group had a final refusal rate of almost 11%. However, the significant drop in refusal rates in the Asian domain with the higher respondent payments does not continue into subsequent rounds, other than in Round 4 where the difference between the $50 and $30 groups attains statistical significance (p≤0.05).

Only round 2 shows a statistically significant drop in final refusal rates for all sampling domains other than the low income domain with the two higher respondent payment amounts relative to the control (see table 9). By round 5 though (table 12), only the 'other' and the Black domain continued to see reduced refusal rates relative to the $30 respondent payment group, and only for the $70 respondent payment group. This is also the case in terms of response rates (tables 1-6). Round 2 shows statistically higher response rates in almost all domains, as well as the full sample, for the higher respondent payment amounts relative to the control. But by round 5, only the Black and the Other domains, along with the full sample show significantly improved response rates relative to the control, and only for the $70 group. This inverse relationship between response and refusal rates reinforces the finding that the higher respondent payment amount improves the response rate primarily by decreasing the refusal rate.

Return To Table Of Contents

Among respondents, cooperation with requests for additional information (Tables 13, 14)

The analysis included two additional measures of cooperation. These measures assess whether the higher respondent payment amount increased the number of eligible adults within a cooperating RU who:

provided authorization to contact their medical provider in any of the first three rounds of data collection , or
completed the Self-Administered Questionnaire (SAQ) in the second round of data collection

Table 13 shows the proportion of households for whom the eligible adults signed and returned some, none or all of the requested authorization forms, by respondent payment amount in each of the first three rounds. As seen in the table, none of the percentage point differences in respondents' willingness to sign the authorization form by respondent payment amount attained statistical significance. However, there may be a trend or pattern indicating that fewer RU's return "none" of the requested authorization form in the $70 group relative to the $30 group. In Round 1, the table shows about a 5 percentage point decrease in the number of households returning 'none' of the requested authorization forms between the $70 and $30 group. Likewise, there is about a 5 percentage point increase in the proportion of households in which all of the requested authorization forms are completed and returned in the $70 group relative to the $30 group. In rounds 2 and 3, the same pattern can be observed, but the percentage point differences are less dramatic.

In terms of completing and returning the SAQ in Round 2 (table 14), a significantly higher percentage of households return all of the requested SAQ's in the $70 group (87.55%) relative to the $30 group (83.42%). Similarly, a significantly lower percentage of households return "none" of the requested SAQs in the $70 group (6.26%) relative to the $30 group (8.64%). The percentage point difference between the $50 respondent payment group and the $30 group does not attain statistical significance. These results suggest that the higher respondent payment does help motivate respondents to follow-through and complete the SAQ.

Return To Table Of Contents

Level of Effort (Tables 15, 16)

These two tables address how much time and effort interviewers must use to close a sampled case and reach a final outcome albeit a complete or some other final outcome. The amount of effort spent closing out the full sample in the data collection period translates to the cost of conducting data collection.

In table 15, the first row of data with each round shows whether interviewers start to work cases differently depending on the respondent payment amount. In round 1, interviewers start all their cases, regardless of the assigned respondent payment amount within about the same time period, about 34 days after the start of the new year. However, in round 2, 3 and 4 the data suggest that interviewers did tend to start working the cases in the higher respondent payment groups a little earlier than the $30 group. This may in part reflect the fact that they realize after the first round that it is somewhat easier to close a case at the higher respondent payment levels than the $30 level (discussed in more detail below - data on table 16).

Of greater consequence, in all rounds, regardless of when the interviewer starts working a case, the cases in the two higher respondent payment groups both require fewer days to close out than the cases in the $30 respondent payment group (Table 15). In rounds 3 and 4, interviewers close cases in the $70 respondent payment group in significantly fewer days than both the $50 and $30 respondent payment groups. Though by round 5, it takes interviewers about 25 days to close a case for both of the higher respondent payment groups, and about 30 days to close a case in the $30 group. These differences have a practical impact in terms of cost.

Table 16 shows the average number of total contacts, and the average number of contacts by mode of contact needed to close a case. In all rounds, the two higher respondent payment amounts require significantly fewer contacts on average to close a case. In each round, it takes close to one less contact on average to close a case for the $70 respondent payment group relative to the $30 group.

Return To Table Of Contents

Data Quality - Item missing data rates (Table 17)

As can be seen in Table 17, the amount of the respondent payment did not affect the percentage of item missing data across any of the variables included in this analysis, for any of the first three rounds of data collection. A variable is considered to have missing data if the question was eligible for the person to answer, and the person either did not provide any answer, or answered with "don't know" or "refused." The percentage of missing data was calculated from the raw response data, prior to any editing or other internal processing.

Four of the variables included in the analysis are those for which it seemed that in most cases the household respondent would know and could provide an answer for any other adult household member. Thus, we did not expect to see any differences by respondent payment amount, unless the higher response rates in the higher respondent payment groups tended to bring in largely uncooperative or less motivated respondents.

We included the final two variables, the date of the person's last flu shot and last check up, as a way to assess whether differences in item missing data rates might reflect an increase in respondents' willingness to check records or talk with other household members in the higher respondent payment groups in order to get a response. However, as with the other variables, the analysis did not find any significant differences in item missing data rates by respondent payment amount for these two variables either.

Return To Table Of Contents

Data Quality - Selected estimates (Table 18)

To examine whether the respondent payment amount affects how people answer questions, and thus the resulting survey estimates, the table includes an analysis of two estimates calculated for each of the first three rounds. Again, this analysis uses the raw response data prior to any editing or cleaning.

The first estimate reflects the proportion of RU's reporting at least one office-based medical visit during the round. As Table 18 shows, the proportion does not differ by the amount of the respondent payment.

The second estimate reflects the proportion of RU's with all members younger than 65 years of age that report at least one person in the household having health insurance on the day of the MEPS interview. Again, as seen in table 18, this proportion does not differ by the amount of the respondent payment.

These findings along with the item missing data findings suggest that the respondent payment amount does not affect how respondents answer the MEPS survey questions. In other words, these data do not support concerns that higher respondent payment amounts could bring in less motivated respondents.

Return To Table Of Contents

Demographic and Geographic Characteristics by Respondent Payment Group (Tables 19-27)

This analysis assesses whether the demographic characteristics of respondents to the MEPS survey differed by the respondent payment amount. In other words, the analysis looks at whether the responding population differs when a higher respondent payment amount is provided relative to the control group. This analysis used the raw, unedited data.

We looked at several standard demographic characteristics of the actual MEPS respondent thought to be correlated with the decision to respond to the MEPS: age, race, education level, employment status, marital status, self-reported health status, and the number of people living in the RU. In addition, we looked at several geographic characteristics of the RU known to be related to response propensity based on data from prior panels of MEPS: region of the country and MSA status (in-MSA or not). Across all variables, and across all rounds, the analysis found no evidence that the respondent payment amount has an effect on who chooses to complete the MEPS interview. In other words, similar distributions across the demographic and geographic variables were observed in each of the respondent payment groups.

Return To Table Of Contents

Summary and Discussion

The primary objective of the MEPS Panel 13 respondent payment experiment was to assess whether increasing the respondent payment would increase response rates to the MEPS in the first and subsequent rounds of data collection, as well as across all five rounds of data collection. Another objective was to learn whether any observed increase in response rates reflected a simultaneous drop in refusal rates, indicating that the respondent payment affects response rates by decreasing refusal to the survey request.

Both of these hypotheses were realized in the experiment. Across the full sample, in rounds 1 and 2, both levels of higher respondent payments resulted in statistically higher response rates than the $30 respondent payment group. In the last three rounds of data collection, only the $70 respondent payment group continued to attain higher response rates (p≤0.01) relative to the $30 group. This same pattern across rounds emerged in regards to refusal rates, with both of the higher respondent payment amounts resulting in lower refusal rates than the $30 group (p≤0.01) in rounds 1 and 2, but only the $70 respondent payment group attaining lower refusal rates in all subsequent rounds (p≤0.01). This negative correlation between the response and refusal rates strongly suggests that the gains in response rates come from a decrease in refusals rather than a change in other noninterview dispositions (e.g., noncontacts or unable to locate/access).

The composite response rate across all 5 rounds of data collection was higher for both the $50 and $70 respondent payment group relative to the $30 group across the full sample, and for all sample domains (p≤0.01) other than the low income domain. In addition, in looking across the full sample, the difference in response rates between the $70 and $50 respondent payment groups was also statistically significant, with a composite response rate of 71.13% and 66.74% for the $70 and $50 groups, respectively.

A second category of analysis examined whether data quality was affected by using higher respondent payments. Specifically, we wanted to understand whether any observed increase in response rates with the higher respondent payment amount resulted in an increase in the cooperation of less enthusiastic respondents who chose to participate in reaction to the monetary payment alone and thus might only put limited effort into responding. To assess this, we looked at item missing data rates, a couple of specific survey estimates, and levels of cooperation with a request to complete a self-administered questionnaire or authorize MEPS to contact their medical provider to collect additional data. The results indicated no negative impact on these particular data quality measures with the higher respondent payments and corresponding higher response rates. Specifically, there were no statistically significant differences in item nonresponse rates, or in either of the two survey estimates (proportion of RUs reporting at least one office-based visit, and proportion of RU's with at least one person with health insurance on the day of the MEPS interview, restricted to households with only members younger than 65 years old). Similarly, RU members were no more or less likely to give authorization to contact their medical provider to collect additional information outside of the MEPS survey in the higher respondent payment groups. However, a higher proportion of RUs returned all of the requested SAQs in the $70 group (87.6%) as compared to the $30 group (83.4%), indicating a potential gain in quality as related to the information collected on that self-administered questionnaire.

We also assessed whether the demographic characteristics of respondents differed by the respondent payment amounts. This component of the analysis helped assess whether the higher respondent payment yields differential responses for certain groups, another indicator of data quality. Ideally, the higher respondent payment amount would result in an increase in the respondents with demographic characteristics of those who typically respond at lower levels. However, neither of the respondent payment groups resulted in any changes in the distribution of age, race, marital status, educational attainment, employment status, or self-reported health status as compared to the $30 respondent payment group. These results suggests that while significant gains in response rates occurred with the higher respondent payment amounts, those high payments tended to garner response and cooperation from respondents very similar to those that already respond at the $30 level, as measured in this analysis. Thus, we found no evidence of negative effects on data quality associated with the higher respondent payment amounts and the increased incentive payments appeared to have a similar effect across the full spectrum of the demographic groups examined. Higher respondent payments have the potential for reducing bias through added survey participation across many subgroups of analytic interest. Further evaluation of other estimates would be called for to help in the assessment of bias reduction.

As a last measure of the effect of the higher respondent payments, the analysis looked at whether the amount of time and resources required to close out the sample in each round of data collection would decrease with the higher respondent payment amounts. The data suggest that in all rounds, interviewers closed cases in fewer days on average with the higher respondent payment levels, and with fewer contacts on average. So the efficiency of survey operations was improved with both of the higher respondent payment levels.

In order to understand how these gains in efficiency effect data collection costs, we estimated the cost per case, by round, using the hours per case reported in panel 11 (prior to any changes in sample design or the CAPI instrument) and the estimated average number of contacts to close a case as observed in the experiment. Field staff did not track their hours by respondent payment group directly, so using number of contacts to complete a case allowed us to reflect the differential level of effort required between the respondent payment groups in the estimated data collection costs.

The table below shows the estimated net change in cost per case relative to the current $30 respondent payment after accounting for the higher respondent payment in round one, by the third round of data collection, and then the net change in total by the final round of data collection relative to the $30 payment. The estimates reflect the current sample allocation only.


Respondent payment	After Round 1	After Round 3	After Round 5
$50	-$3	-$12	+$0.32
$70	-$2	+$5	+$28

As indicated in the table, the $20 increase in the respondent payment, or the $50 gift to respondents, ultimately (after Round 5) results in almost no estimated increase in cost per case. However, the estimated cost per case increases more dramatically with the $70 gift to respondents. Much of the increase in costs in the $70 group reflects the positive aspect of the$70 payment group - there are simply more cooperating households in the $70 group.

While the data in these analyses indicate that the greatest gains in response come with the $70 respondent payment, especially in the last three rounds of data collection, they suggest that both of the two higher respondent payment amounts elicit cooperation from respondents who are very similar to those already participating with the $30 respondent payment according to select demographic characteristics. More specifically since the respondents to the MEPS are self-selected among eligible adults based on the NHIS survey outcome, these data suggest that the higher respondent payment level does not change the demographics of who agrees to serve as the respondent to the MEPS within an RU. One potential concern with paying higher respondent payment amounts is that respondents motivated only by the monetary gift will cooperate with the survey request, but will not put forth adequate effort to provide good data. However, these results clearly indicate that the higher respondent payments did not bring in respondents who participate at only minimal levels of effort.

Return To Table Of Contents

References

Cohen, Steven B.; Trena Ezzati-Rice; W. Yu (2006) "The impact of survey attrition on health insurance coverage estimates in a national longitudinal health care survey" Health Services Outcomes Research Methods, 6; 111-125.

Dillman, Don A. (2007) Mail and Internet Surveys: The Tailored Design Method (2007 Update with new Internet, Visual and Mixed Mode Guide.) John Wiley & Sons

Goldenberg, Karen L.; David McGrath, Lucilla Tan (2009) "Incentives in the Consumer Expenditure Interview Survey: One payment, lasting effects" Paper presented at the 2009 American Association of Public Opinion Research Conference, Hollywood, Florida.

Goyder, John (1994) "An experiment with cash incentives on a personal interview survey." Journal of the Marketing Research Society. 36(4)

Singer, Eleanor; John Van Hoewyk, Nancy Gebler, Trivellore Raghunathan and Katherine McGonagle "The Effect of Incentives on Response Rate in Interviewer-Mediated Surveys" Journal of Official Statistics Vol. 15, no. 2

Willimack, Diane K., Howard Schuman, Beth-Ellen Pennell and James Lepkowski (1995) "Effects of a Prepaid NonMonetary Incentive on Response Rates and Response Quality in a Face-to-Face Survey" Public Opinion Quarterly, 59.

Return To Table Of Contents

Table 1: Composite response rate (R1-R5) by sampling domain and incentive group
Respondent Group	$30 Incentive Unweighted	$30 Incentive Weighted RR	$30 Incentive s.e.	$50 Incentive Unweighted	$50 Incentive Weighted RR	$50 Incentive s.e.	$70 Incentive Unweighted	$70 Incentive Weighted RR	$70 Incentive s.e.
Asian	50.92	51.25⁺	3.22	62.98	62.32^$	3.45	59.93	63.35^@	3.13
Low Income	72.06	71.82^@	2.01	69.72	72.55	2.23	79.06	78.88^@	2.2
Hispanic	57.82	57.14⁺	2.11	65.64	65.52^!	2.33	72.64	73.15^!	2.35
Black	61.01	60.84⁺	2.51	69.77	70.03^@	2.24	74.76	74.39^@	1.9
Other	57.01	57.11⁺	1.53	65.88	65.71^@	1.55	68.89	69.97^@	1.55
All Domains	59.92	58.84⁺	1.13	66.92	66.74^{^}	1.19	71.45	71.13^{^}	1.12

Return To Table Of Contents

Table 2: Round 1 response rate by sampling domain and incentive group
Respondent Group	Asian	Low Income	Hispanic	Black	Other	All Domains
$30 Incentive, Complete	254	605	600	452	1,023	2,934
$30 Incentive, Net Sample	363	725	799	574	1,449	3,910
$30 Incentive, Unweighted RR	68.36%	82.38%	72.96%	77.44%	69.15%	73.46%
$30 Incentive, Weighted RR	68.77%	82.22%	72.37%^$	78.22%	69.05%⁺	71.67%⁺
$30 Incentive, s.e.	2.82	1.64	1.89	1.65	1.35	1.01
$50 Incentive, Complete	180	483	511	431	928	2533
$50 Incentive, Net Sample	245	599	648	521	1,207	3,220
$50 Incentive, Unweighted RR	71.55%	79.02%	76.92%	81.78%	75.66%	77.20%
$50 Incentive, Weighted RR	70.25%	81.35%^$	76.85%	81.95%	75.36%^@	76.60%^@
$50 Incentive, s.e.	3.52	1.93	2.26	1.82	1.36	1.06
$70 Incentive, Complete	209	504	451	416	964	2544
$70 Incentive, Net Sample	284	570	565	505	1,258	3,182
$70 Incentive, Unweighted RR	71.03%	87.14%	78.45%	82.11%	75.14%	78.61%
$70 Incentive, Weighted RR	73.83%	87.02%^#	78.87%^$	81.14%	75.63%^@	77.60%^@
$70 Incentive, s.e.	2.79	1.79	2.15	1.77	1.49	1.09

Return To Table Of Contents

Table 3: Round 2 response rate by sampling domain and incentive group
Respondent Group	Asian	Low Income	Hispanic	Black	Other	All Domains
$30 Incentive, Complete	233	581	557	426	978	2,775
$30 Incentive, Net Sample	261	619	622	463	1,046	3,011
$30 Incentive, Unweighted RR	88.61%	93.54%	89.13%	91.80%	93.24%	91.84%
$30 Incentive, Weighted RR	86.33%⁺	93.61%^$	88.95%⁺	90.28%^@	93.36%^>	92.33%⁺
$30 Incentive, s.e.	2.93	1.18	1.35	2.05	0.83	0.62
$50 Incentive, Complete	175	478	508	426	902	2489
$50 Incentive, Net Sample	183	500	530	449	944	2,604
$50 Incentive, Unweighted RR	95.57%	95.45%	95.69%	94.68%	95.47%	95.38%
$50 Incentive, Weighted RR	95.59%^@	96.16%	95.84%^@	94.44%^"	95.53%^$	95.54%^{^}
$50 Incentive, s.e.	1.42	0.81	0.92	1.15	0.71	0.49
$70 Incentive, Complete	201	506	454	418	964	2543
$70 Incentive, Net Sample	214	517	469	430	996	2,626
$70 Incentive, Unweighted RR	93.79%	97.22%	96.70%	97.13%	96.69%	96.63%
$70 Incentive, Weighted RR	95.04%^@	97.03%^$	97.24%^@	97.20%^!	97.06%^@	96.99%^{^}
$70 Incentive, s.e.	1.5	0.73	0.69	0.88	0.52	0.38

Return To Table Of Contents

Table 4: Round 3 response rate by sampling domain and incentive group
Respondent Group	Asian	Low Income	Hispanic	Black	Other	All Domains
$30 Incentive, Complete	215	574	547	413	944	2,693
$30 Incentive, Net Sample	234	587	574	442	993	2,830
$30 Incentive, Unweighted RR	91.81%	97.58%	95.15%	93.23%	94.95%	95.01%
$30 Incentive, Weighted RR	93.74%	97.84%	94.92%^@	92.94%^@	94.98%^@	95.13%^@
$30 Incentive, s.e.	1.36	0.6	1.06	1.52	0.74	0.53
$50 Incentive, Complete	172	462	505	416	883	2438
$50 Incentive, Net Sample	178	477	532	435	914	2,536
$50 Incentive, Unweighted RR	96.55%	96.80%	94.73%	95.52%	96.43%	96.00%
$50 Incentive, Weighted RR	96.62%	97.21%	94.42%^*	95.43%	96.65%	96.37%
$50 Incentive, s.e.	1.31	0.72	1.33	1	0.7	0.5
$70 Incentive, Complete	201	498	466	424	953	2542
$70 Incentive, Net Sample	208	514	472	435	975	2,604
$70 Incentive, Unweighted RR	96.55%	96.83%	98.68%	97.39%	97.71%	97.56%
$70 Incentive, Weighted RR	96.13%	96.58%	98.45%^{^}	97.55%^@	97.65%^@	97.51%^@
$70 Incentive, s.e.	1.41	1.34	0.59	0.67	0.54	0.4

Return To Table Of Contents

Table 5: Round 4 response rate by sampling domain and incentive group
Respondent Group	Asian	Low Income	Hispanic	Black	Other	All Domains
$30 Incentive, Complete	204	571	533	402	921	2,631
$30 Incentive, Net Sample	214	585	557	421	958	2,735
$30 Incentive, Unweighted RR	95.31%	97.55%	95.59%	95.39%	96.06%	96.12%
$30 Incentive, Weighted RR	95.90%	97.31%	96.23%^$	95.83%	96.28%^$	96.36%^@
$30 Incentive, s.e.	1.45	0.87	0.78	1.01	0.71	0.45
$50 Incentive, Complete	172	461	501	414	865	2,413
$50 Incentive, Net Sample	177	471	516	425	896	2,485
$50 Incentive, Unweighted RR	97.09%	97.84%	97.04%	97.36%	96.48%	97.05%
$50 Incentive, Weighted RR	97.40%	97.67%	96.92%	97.55%	96.55%	96.86%
$50 Incentive, s.e.	1.31	0.94	0.87	0.87	0.62	0.43
$70 Incentive, Complete	193	496	467	422	937	2,515
$70 Incentive, Net Sample	202	506	475	434	957	2,574
$70 Incentive, Unweighted RR	95.52%	97.98%	98.29%	97.18%	97.89%	97.67%
$70 Incentive, Weighted RR	96.35%	98.01%	98.31%^$	97.72%	97.95%^$	97.90%^@
$70 Incentive, s.e.	1.22	0.7	0.64	0.68	0.54	0.38

Return To Table Of Contents

Table 6: Round 5 response rate by sampling domain and incentive group
Respondent Group	Asian	Low Income	Hispanic	Black	Other	All Domains
$30 Incentive, Complete	197	562	531	388	894	2,572
$30 Incentive, Net Sample	205	572	543	402	922	2,644
$30 Incentive, Unweighted RR	96.06%	98.24%	97.75%	96.50%	96.94%	97.25%
$30 Incentive, Weighted RR	96.04%	98.00%	97.18%	96.74%^$	96.86%^@	97.00%^@
$30 Incentive, s.e.	1.57	0.81	0.91	0.79	0.71	0.49
$50 Incentive, Complete	170	456	490	414	848	2,378
$50 Incentive, Net Sample	173	467	505	427	865	2,437
$50 Incentive, Unweighted RR	98.25%	97.60%	97.01%	96.89%	98.03%	97.55%
$50 Incentive, Weighted RR	98.60%	97.69%	97.22%	97.20%^"	97.81%^"	97.70%^*
$50 Incentive, s.e.	0.82	0.73	0.79	0.70	0.64	0.43
$70 Incentive, Complete	190	491	462	422	933	2,498
$70 Incentive, Net Sample	196	499	468	426	941	2,530
$70 Incentive, Unweighted RR	96.89%	98.37%	98.72%	99.05%	99.14%	98.72%
$70 Incentive, Weighted RR	97.48%	98.69%	98.54%	98.96%^#	99.22%^!	99.00%^{^}
$70 Incentive, s.e.	1.02	0.51	0.64	0.55	0.28	0.21

Return To Table Of Contents

Table 7: Response rate by NHIS outcome (partial or complete) by round, and overall
Respondent Group	R1	R2	R3	R4	R5	All
$30 Incentive, NHIS Completes, Net sample, NHIS completes	2,904	2335	2211	2145	2074	3101
$30 Incentive, NHIS Completes, Weighted RR	75.40	92.86	95.82	96.30	97.17	62.78
$30 Incentive, NHIS Completes, s.e.	1.00	0.63	0.55	0.51	0.47	1.12
$30 Incentive, NHIS Partials, Net Sample, NHIS partials	1,006	676	619	590	570	1072
$30 Incentive, NHIS Partials, Weighted RR	59.34^$	90.21^>	92.26⁺	96.64	96.3^$	45.97^$
$30 Incentive, NHIS Partials, s.e.	2.04	1.55	1.369	0.906	1.217	2.154
$50 Incentive, NHIS Completes, Net sample, NHIS completes	2439	2063	2012	1975	1924	2618
$50 Incentive, NHIS Completes, Weighted RR	80.64	95.85	96.71	96.59	98.1	70.83
$50 Incentive, NHIS Completes, s.e.	1.13	0.53	0.54	0.48	0.35	1.28
$50 Incentive, NHIS Partials, Net Sample, NHIS partials	781	541	524	510	513	845
$50 Incentive, NHIS Partials, Weighted RR	62.17	94.21^$	94.88^{^}	98.02	96.02^"	52.31
$50 Incentive, NHIS Partials, s.e.	2.32	1.13	1.15	0.76	1.27	2.32
$70 Incentive, NHIS Completes, Net sample, NHIS completes	2365	2028	2019	1991	1961	2550
$70 Incentive, NHIS Completes, Weighted RR	81.27	97.18	97.37	98.35	99.05	74.93
$70 Incentive, NHIS Completes, s.e.	1.19	0.39	0.48	0.33	0.24	1.25
$70 Incentive, NHIS Partials, Net Sample, NHIS partials	817	598	585	583	569	878
$70 Incentive, NHIS Partials, Weighted RR	65.55^$	96.28^@	98.02^{^}	96.18	98.78^#	58.78^$
$70 Incentive, NHIS Partials, s.e.	2.06	0.99	0.63	1.19	0.53	1.99

Return To Table Of Contents

Table 8: Percent of RU's that ever refused, and the final refusal rate in Round 1, by sampling domain and incentive group
Respondent Group	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive (%)	$70 Incentive s.e.
Asian - Percent ever refused	27.45	1.65	30.11	2.49	27.09	3.40	24.08	2.49
Asian - Final R1 Refusal Rate	21.33	1.50	24.08	2.24	20.50	3.03	18.25	2.47
Low Income - Percent ever refused	13.47	0.86	13.91	1.50	14.13	1.67	12.22	1.91
Low Income - Final R1 Refusal Rate	9.29	0.67	10.74	1.27	8.89	1.33	7.83	1.52
Hispanic - Percent ever refused	19.87	1.09	23.23^>	1.80	17.80^$	1.80	17.43^$	1.79
Hispanic - Final R1 Refusal Rate	14.44	0.91	16.01	1.34	13.41	1.52	13.38	1.71
Black - Percent ever refused	18.26	1.12	20.94	1.72	16.78	1.68	16.59	1.92
Black - Final R1 Refusal Rate	12.32	0.89	14.47^$	1.47	10.80^$	1.28	11.34	1.44
Other - Percent ever refused	29.45	0.82	32.86⁺	1.46	27.42^@	1.26	27.34^@	1.57
Other - Final R1 Refusal Rate	21.97	0.70	25.30⁺	1.28	19.71^@	1.24	20.18^@	1.39
Across all sample domains - Percent ever refused	25.62	0.59	28.54⁺	1.07	23.91^@	1.02	23.72^@	1.14
Across all sample domains - Final R1 Refusal Rate	18.92	0.50	21.70⁺	0.94	17.03^@	0.97	17.36^@	1.14

Return To Table Of Contents

Table 9: In Round 2, percent of RU's that ever refused, and the final refusal rate, by sampling domain and incentive group
Respondent Group	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive (%)	$70 Incentive s.e.
Asian - Percent ever refused	7.89	1.27	13.10⁺	2.67	3.04^@	1.09	5.35^@	1.47
Asian - Final R2 Refusal Rate	5.83	1.23	10.78⁺	2.66	2.24^@	1.00	2.54^@	1.00
Low Income - Percent ever refused	3.89	0.55	5.61^>	1.13	3.22	0.84	2.47^$	0.71
Low Income - Final R2 Refusal Rate	2.09	0.41	2.89	0.87	1.64	0.55	1.57	0.54
Hispanic - Percent ever refused	6.32	0.69	11.09⁺	1.51	3.64^@	0.88	3.03^@	0.82
Hispanic - Final R2 Refusal Rate	4.21	0.51	7.93^>	1.15	2.13^$	0.57	1.62^$	0.57
Black - Percent ever refused	6.88	0.92	11.01⁺	1.87	4.95^@	1.17	4.20^@	2.15^$
Black - Final R2 Refusal Rate	4.71	0.76	8.27^>	1.74	3.25^$	1.01	1.19	0.73
Other - Percent ever refused	5.84	0.49	7.23^$	0.83	5.75	0.85	4.40^$	0.66
Other - Final R2 Refusal Rate	3.91	0.38	5.59⁺	0.76	3.52^$	0.58	2.46^@	0.47
Across all sample domains - Percent ever refused	5.82	0.36	8.03⁺	0.64	5.03^@	0.54	4.06^@	0.49
Across all sample domains - Final R2 Refusal Rate	3.86	0.27	5.96⁺	0.54	3.07^@	0.38	2.25^@	0.33

Return To Table Of Contents

Table 10: In Round 3, percent of RU's that ever refused, and the final refusal rate, by sampling domain and incentive group
Respondent Group	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive (%)	$70 Incentive s.e.
Asian - Percent ever refused	5.48	0.91	6.95	1.78	4.93	1.60	4.28	1.45
Asian - Final R3 Refusal Rate	3.57	0.76	3.83	1.24	2.98	1.29	3.78	1.44
Low Income - Percent ever refused	2.04	0.39	1.65	0.57	2.58	0.67	1.96	0.70
Low Income - Final R3 Refusal Rate	0.74	0.26	0.56	0.38	0.83	0.44	0.85	0.44
Hispanic - Percent ever refused	3.52	0.46	4.75^$	1.02	3.30	0.74	2.28^$	0.74
Hispanic - Final R3 Refusal Rate	2.43	0.40	3.68^@	0.90	2.38	0.66	0.97^@	0.56
Black - Percent ever refused	4.38	0.60	4.21	0.99	5.37	1.15	3.56	0.88
Black - Final R3 Refusal Rate	2.59	0.42	2.42	0.63	3.12	0.93	2.23	0.61
Other - Percent ever refused	4.79	0.42	6.90^>	0.91	4.54^$	0.73	2.81^@	0.54
Other - Final R3 Refusal Rate	2.94	0.33	4.33⁺	0.70	2.79	0.53	1.63^@	0.40
Across all sample domains - Percent ever refused	4.30	0.29	5.74^@	0.60	4.26^"	0.51	2.78^!	0.38
Across all sample domains - Final R3 Refusal Rate	2.60	0.23	3.56⁺	0.48	2.54	0.37	1.62^@	0.29

Return To Table Of Contents

Table 11: In Round 4, percent of RU's that ever refused, and the final refusal rate, by sampling domain and incentive group
Respondent Group	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive (%)	$70 Incentive s.e.
Asian - Percent ever refused	3.56	0.82	5.50	1.64	2.12	1.01	2.70	1.12
Asian - Final R4 Refusal Rate	2.53	0.75	4.09^$	1.46	1.15^$	0.80	2.04	0.94
Low Income - Percent ever refused	2.44	0.46	3.89^$	0.92	1.30^$	0.58	1.80	0.70
Low Income - Final R4 Refusal Rate	1.19	0.31	1.88	0.59	0.77	0.49	0.77	0.44
Hispanic - Percent ever refused	2.79	0.46	3.63	0.82	2.70	0.83	1.88	0.69
Hispanic - Final R4 Refusal Rate	1.52	0.32	2.05	0.54	1.13	0.52	1.33	0.59
Black - Percent ever refused	3.20	0.50	4.89^$	1.17	2.63	0.77	2.05^$	0.69
Black - Final R4 Refusal Rate	1.73	0.38	2.77	0.87	0.98	0.49	1.43	0.56
Other - Percent ever refused	3.96	0.43	4.36^$	0.69	5.09*	0.80	2.48^:	0.58
Other - Final R4 Refusal Rate	2.44	0.35	3.00^$	0.60	2.74	0.55	1.58^$	0.49
Across all sample domains - Percent ever refused	3.56	0.29	4.32^@	0.45	4.02*	0.53	2.31^{^}	0.41
Across all sample domains - Final R4 Refusal Rate	2.13	0.22	2.78^@	0.37	2.10	0.37	1.46^@	0.34

Return To Table Of Contents

Table 12: In Round 5, percent of RU's that ever refused, and the final refusal rate, by sampling domain and incentive group
Respondent Group	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive (%)	$70 Incentive s.e.
Asian - Percent ever refused	1.89	0.59	2.27	1.17	1.24	0.68	2.04	0.92
Asian - Final R5 Refusal Rate	1.16	0.40	1.21	0.68	0.84	0.61	1.38	0.80
Low Income - Percent ever refused	1.20	0.33	1.28	0.54	1.42	0.58	0.88	0.48
Low Income - Final R5 Refusal Rate	0.69	0.23	1.11	0.53	0.80	0.36	0.10	0.10
Hispanic - Percent ever refused	1.47	0.35	1.95	0.66	1.49	0.67	0.89	0.49
Hispanic - Final R5 Refusal Rate	1.00	0.31	1.17	0.55	1.17	0.57	0.62	0.45
Black - Percent ever refused	1.49	0.34	3.40⁺	0.88	0.97^@	0.45	0.14^@	0.14
Black - Final R5 Refusal Rate	0.95	0.26	1.90^@	0.62	0.97^"	0.45	0.00^!	0.00
Other - Percent ever refused	1.98	0.31	2.85^@	0.66	2.11	0.58	0.98^@	0.33
Other - Final R5 Refusal Rate	1.27	0.24	1.88^@	0.52	1.47	0.51	0.49^@	0.22
Across all sample domains - Percent ever refused	1.78	0.20	2.56^@	0.41	1.81^"	0.37	0.92^!	0.23
Across all sample domains - Final R5 Refusal Rate	1.14	0.15	1.67^@	0.33	1.28^"	0.32	0.44^!	0.15

Return To Table Of Contents

Table 13: Cooperation with request to sign the Authorization Form by incentive group, for Rounds 1, 2, and 3
Of RU's w/1 or more AF's requested	Across Incentive Groups count	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive count	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive count	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive count	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - Some, but not all returned	48	2.38	0.48	22	2.64	0.70	11	1.96	0.69	15	2.47	1.03
Round 1 - None Returned	516	30.37	1.64	200	32.19	2.75	174	31.16	2.90	142	26.95	2.36
Round 1 - All returned	1182	67.25	1.62	454	65.17	2.64	376	66.87	2.93	352	70.58	2.37
Round 2 - Some, but not all returned	993	14.99	0.59	365	14.44	0.90	304	14.75	0.97	324	15.83	0.99
Round 2 - None Returned	1143	17.68	0.77	409	17.75	1.27	383	18.96	1.35	351	16.36	1.03
Round 2 - All returned	4433	67.33	1.04	1565	67.81	1.53	1408	66.29	1.59	1460	67.81	1.48
Round 3 -Some, but not all returned	813	12.66	0.55	288	12.88	0.93	249	12.10	0.89	276	12.97	0.85
Round 3 -None Returned	1280	20.70	0.91	469	20.95	1.34	419	21.72	1.41	392	19.47	1.39
Round 3 -All returned	4024	66.64	1.01	1398	66.17	1.47	1268	66.18	1.58	1358	67.56	1.62

Return To Table Of Contents

Table 14: Cooperation with request to complete and return the SAQ by incentive group, in Round 2
Of RU's with 1 or more SAQ requested	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Some, but not all completed	596	6.94	0.36	228	7.94 $	0.67	178	6.60	0.54	190	6.19$	0.54
None completed	609	7.45	0.39	256	8.64@	0.71	189	7.35	0.79	164	6.26@	0.57
All completed	6595	85.61	0.50	2289	83.42@	0.93	2120	86.05	0.88	2186	87.55@	0.74

Return To Table Of Contents

Table 15: For each round, the avg. number of days between the start of data collection and the first contact, and the avg. number of days between the first contact and final case resolution
Respondent Group	Across Incentive Groups Avg. number of days	Across Incentive Groups s.e.	$30 Incentive Avg. number of days	$30 Incentive s.e.	$50 Incentive Avg. number of days	$50 Incentive s.e.	$70 Incentive Avg. number of days	$70 Incentive s.e.
Round 1 - Between Jan 1 and First Contact	34.36	0.60	34.42	0.86	34.71	0.89	33.91	0.79
Round 1 - Between First Contact and Final Disp.	49.10	0.97	54.54+	1.83	47.31@	1.27	44.59@	1.49
Round 2 - Between July 28 and First Contact	25.19	0.52	26.12@	0.70	25.71"	0.74	23.61^!	0.74
Round 2 - Between First Contact and Final Disp.	22.34	0.44	26.85+	0.84	20.46@	0.79	19.03@	0.67
Round 3 - Between Jan 1 and First Contact	35.36	0.67	36.82^$	0.85	34.33^$	1.05	34.77	1.14
Round 3 - Between First Contact and Final Disp.	24.35	0.51	28.45+	0.91	23.55^!	0.97	20.70^!	0.82
Round 4 - Between June 28 and First Contact	32.37	0.51	34.68+	0.69	31.02@	0.77	31.21@	0.80
Round 4 - Between First Contact and Final Disp.	27.14	0.45	29.12^@	0.72	27.29"	0.70	24.88!	0.81
Round 5 - Between Jan 1 and First Contact	39.87	0.70	40.32	0.96	40.16	0.90	39.13	0.93
Round 5 - Between First Contact and Final Disp.	26.92	0.47	30.34+	0.85	25.64@	0.82	24.55@	0.73

Return To Table Of Contents

Table 16: For each round, the average number of in-person, telephone contacts and total contacts to close a case, by sampling domain and incentive group
Respondent Group	Across Incentive Groups Average Num of contacts to close R1 case	Across Incentive Groups s.e. of avg.	$30 Incentive Average Num of contacts to close R1 case	$30 Incentive s.e. of avg.	$50 Incentive Average Num of contacts to close R1 case	$50 Incentive s.e. of avg.	$70 Incentive Average Num of contacts to close R1 case	$70 Incentive s.e. of avg.
Round 1 - Total contacts	7.18	0.11	7.74⁺	0.20	7.04^!	0.18	6.62^!	0.17
Round 1 - In-person contacts	4.90	0.10	5.25⁺	0.17	4.85^$	0.14	4.52^@	0.13
Round 1 - Telephone contacts	2.28	0.05	2.49^>	0.10	2.19^$	0.08	2.10^@	0.08
Round 2 - Total contacts	5.73	0.10	6.50⁺	0.18	5.42^@	0.14	5.16^@	0.12
Round 2 - In-person contacts	2.18	0.04	2.48⁺	0.08	2.05^@	0.06	1.96^@	0.06
Round 2 - Telephone contacts	3.55	0.07	4.02⁺	0.12	3.36^@	0.10	3.21^@	0.10
Round 3 - Total contacts	5.16	0.07	5.60⁺	0.12	5.03^@	0.12	4.81^@	0.12
Round 3 - In-person contacts	1.93	0.03	2.08^>	0.06	1.90^$	0.06	1.78^@	0.05
Round 3 - Telephone contacts	3.23	0.06	3.52⁺	0.10	3.13^@	0.10	3.02^@	0.09
Round 4 - Total contacts	5.18	0.08	5.54⁺	0.14	5.11^$	0.13	4.86^@	0.12
Round 4 - In-person contacts	1.78	0.03	1.87^@	0.06	1.80	0.05	1.67^@	0.05
Round 4 - Telephone contacts	3.40	0.06	3.66⁺	0.12	3.32^$	0.10	3.19^@	0.09
Round 5 - Total contacts	4.57	0.07	4.97⁺	0.13	4.45^@	0.12	4.27^@	0.07
Round 5 - In-person contacts	1.67	0.03	1.83⁺	0.07	1.66^:	0.05	1.51^{^}	0.04
Round 5 - Telephone contacts	2.90	0.05	3.14^>	0.10	2.79^$	0.10	2.76^@	0.06

Return To Table Of Contents

Table 17: Item missing data rates for select variables by incentive group, for Rounds 1, 2 and 3
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - Educational Attainment	244	2.30	0.18	75	1.97	0.30	79	2.28	0.31	90	2.71	0.32
Round 1 - Mental Health Status	129	1.34	0.15	44	1.25	0.26	47	1.64	0.28	38	1.15	0.19
Round 1 - General Heath Status	126	1.32	0.14	44	1.25	0.26	47	1.56	0.26	35	1.15	0.20
Round 1 - Employment Status	245	2.66	0.22	104	3.30^@	0.41	80	2.71	0.38	61	1.88^@	0.27
Round 2 - Educational Attainment	42	2.68	0.59	18	3.35	0.95	16	2.56	0.86	8	2.02	0.90
Round 2 - Mental Health Status	126	1.50	0.15	48	1.73	0.31	44	1.30	0.26	34	1.43	0.23
Round 2 - General Heath Status	116	1.38	0.15	44	1.59	0.29	38	1.10	0.25	34	1.43	0.23
Round 2 - Employment Status	178	3.36	0.33	72	4.11	0.55	56	2.90	0.49	50	2.95	0.56
Round 3 - Educational Attainment	30	2.08	0.45	9	2.60	1.04	9	1.49	0.55	12	2.17	0.71
Round 3 - Mental Health Status	126	1.67	0.17	44	1.96	0.35	45	1.75	0.29	37	1.30	0.21
Round 3 - General Heath Status	122	1.60	0.17	42	1.83	0.35	43	1.66	0.28	37	1.30	0.21
Round 3 - Employment Status	163	3.25	0.28	64	3.55	0.51	55	3.41	0.48	44	2.75	0.44
Last Check-Up	683	7.62	0.35	243	7.79	0.65	218	7.56	0.62	222	7.50	0.61
Last Flu Shot	606	7.06	0.38	215	7.30	0.63	201	7.66	0.69	190	6.22	0.50

Return To Table Of Contents

Table 18: Comparison of selected estimates by incentive group, for Rounds 1, 2 and 3
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - Reported at least one office-based visit	5,760	83.68	0.59	2,146	84.72	0.98	1,815	82.71	1.07	1,799	83.44	1.02
Round 1 - At least one person in RU had health insurance on the day of MEPS interview	5,296	83.99	0.60	1,974	85.17	1.03	1,673	82.90	1.07	1,649	83.71	1.05
Round 2 - Reported at least one office-based visit	5,688	84.35	0.59	2,048	84.97	1.05	1,824	84.00	0.97	1,816	84.02	0.97
Round 2 - At least one person in RU had health insurance on the day of MEPS interview	5,234	84.67	0.60	1,882	85.14	1.12	1,680	84.44	0.94	1,672	84.39	0.96
Round 3 - Reported at least one office-based visit	5,611	84.21	0.54	1,980	84.43	1.02	1,792	83.54	0.93	1,839	84.61	1.00
Round 3 - At least one person in RU had health insurance on the day of MEPS interview	5,147	84.61	0.55	1,821	84.65	1.08	1,644	84.19	0.95	1,682	84.97	1.00

Return To Table Of Contents

Table 19: Age distribution of respondents in rounds 1, 2 and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7950	100	0	2915	100	0	2512	100	0	2523	100	0
Round 1 - 24 or younger	803	9.30	0.42	268	8.19	0.74	279	10.60	0.82	256	9.29	0.77
Round 1 - 25-34	1616	18.62	0.67	597	19.10	1.14	523	19.15	1.15	496	17.55	1.01
Round 1 - 35-44	1624	18.59	0.50	629	19.47	0.93	503	18.44	0.87	492	17.72	0.87
Round 1 - 45-54	1566	19.97	0.53	594	20.49	1.01	452	18.24	0.87	520	21.09	0.88
Round 1 - 55-64	1133	15.60	0.57	384	14.79	0.91	382	15.87	1.01	367	16.28	0.96
Round 1 - 65 or older	1208	17.91	0.67	443	17.96	0.95	373	17.71	1.16	392	18.06	1.19
Round 2 - All	7738	100	0	2754	100	0	2461	100	0	2523	100	0
Round 2 - 24 or younger	805	9.75	0.41	264	8.78	0.74	268	10.39	0.82	273	10.20	0.71
Round 2 - 25-34	1535	18.34	0.70	545	18.71	1.13	502	18.61	1.17	488	17.68	1.04
Round 2 - 35-44	1580	18.28	0.48	590	19.12	0.98	501	18.38	0.90	489	17.27	0.87
Round 2 - 45-54	1516	19.89	0.53	562	20.45	0.96	450	18.73	0.84	504	20.41	0.89
Round 2 - 55-64	1108	15.52	0.59	373	14.92	0.95	361	15.42	0.95	374	16.27	0.95
Round 2 - 65 or older	1194	18.22	0.70	420	18.03	0.92	379	18.47	1.21	395	18.17	1.23
Round 3 - All	7605	100	0	2673	100	0	2414	100	0	2518	100	0
Round 3 - 24 or younger	792	9.64	0.42	253	8.62	0.74	261	10.12	0.85	278	10.27	0.74
Round 3 - 25-34	1502	18.50	0.72	514	18.58	1.19	501	19.14	1.20	487	17.80	1.03
Round 3 - 35-44	1549	18.14	0.49	578	18.97	1.01	475	17.86	0.92	496	17.51	0.84
Round 3 - 45-54	1475	19.72	0.52	535	20.07	0.97	441	18.65	0.82	499	20.38	0.86
Round 3 - 55-64	1113	15.81	0.60	382	15.68	0.98	362	15.69	0.95	369	16.06	0.94
Round 3 - 65 or older	1174	18.19	0.70	411	18.08	0.94	374	18.54	1.26	389	17.98	1.20

Return To Table Of Contents

Table 20: Race distribution of respondents in Rounds 1, 2, and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7950	100	0	2915	100	0	2512	100	0	2523	100	0
Round 1 - Hispanic	1944	11.67	0.67	745	12.10	0.94	621	11.82	1.05	578	11.03	1.01
Round 1 - Black, Non-Hisp	1707	11.93	0.69	610	11.98	0.95	547	12.04	0.97	550	11.77	1.06
Round 1 - Asian, Non-Hisp	513	3.57	0.34	202	3.89	0.49	143	3.15	0.47	168	3.62	0.45
Round 1 - Other, Non-Hisp	3786	72.83	0.89	1358	72.04	1.38	1201	72.99	1.54	1227	73.59	1.48
Round 2 - All	7738	100	0	2754	100	0	2461	100	0	2523	100	0
Round 2 - Hispanic	1897	11.71	0.67	706	12.11	0.99	619	12.19	1.08	572	10.80	0.99
Round 2 - Black, Non-Hisp	1672	11.94	0.72	574	11.79	0.97	540	12.19	0.99	558	11.86	1.10
Round 2 - Asian, Non-Hisp	485	3.41	0.34	185	3.58	0.48	139	3.12	0.47	161	3.53	0.45
Round 2 - Other, Non-Hisp	3684	72.94	0.90	1289	72.52	1.47	1163	72.50	1.53	1232	73.82	1.49
Round 3 - All	7605	100	0	2673	100	0	2414	100	0	2518	100	0
Round 3 - Hispanic	1885	11.88	0.70	701	12.40	1.07	604	12.09	1.11	580	11.11	0.98
Round 3 - Black, Non-Hisp	1651	12.03	0.73	559	11.83	0.99	529	12.16	1.00	563	12.11	1.10
Round 3 - Asian, Non-Hisp	466	3.35	0.34	167	3.42	0.44	137	3.11	0.49	162	3.51	0.46
Round 3 - Other, Non-Hisp	3603	72.74	0.92	1246	72.35	1.51	1144	72.64	1.57	1213	73.27	1.49

Return To Table Of Contents

Table 21: Distribution of educational attainment for respondents in Rounds 1, 2, and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7928	100	0	2908	100	0	2504	100	0	2516	100	0
Round 1 - No degree	1587	13.22	0.47	610	14.10	0.74	485	13.07	0.88	492	12.37	0.84
Round 1 - High school degree, no college	2521	32.46	0.71	909	31.90	1.12	787	32.13	1.30	825	33.43	1.27
Round 1 - Some college, associate degree	1922	25.76	0.65	693	25.81	1.04	626	25.88	1.19	603	25.59	0.96
Round 1 - Bachelors degree	1236	18.44	0.63	462	18.68	0.96	405	19.64	1.12	369	16.99	1.03
Round 1 - Graduate degree	662	10.11	0.52	234	9.51	0.77	201	9.28	0.81	227	11.62	0.91
Round 2 - All	7708	100	0	2744	100	0	2453	100	0	2511	100	0
Round 2 - No degree	1544	13.61	0.48	572	14.17	0.83	483	13.74	0.91	489	12.87	0.84
Round 2 - High school degree, no college	2472	32.50	0.72	873	32.43	1.16	771	31.88	1.33	828	33.19	1.27
Round 2 - Some college, associate degree	1894	26.09	0.67	669	26.53	1.12	615	25.84	1.21	610	25.85	1.01
Round 2 - Bachelors degree	1178	17.98	0.63	416	17.55	0.97	396	19.67	1.20	366	16.82	1.00
Round 2 - Graduate degree	620	9.82	0.50	214	9.33	0.74	188	8.88	0.77	218	11.27	0.87
Round 3 - All	7568	100	0	2657	100	0	2404	100	0	2507	100	0
Round 3 - No degree	1547	13.93	0.46	570	14.57	0.78	483	14.14	0.94	494	13.06	0.81
Round 3 - High school degree, no college	2431	32.69	0.70	849	32.98	1.18	764	32.32	1.32	818	32.72	1.23
Round 3 - Some college, associate degree	1875	26.29	0.68	647	26.28	1.15	598	25.72	1.27	630	26.85	1.02
Round 3 - Bachelors degree	1134	17.54	0.62	394	17.13	0.97	379	18.92	1.19	361	16.66	1.01
Round 3 - Graduate degree	581	9.55	0.52	197	9.04	0.73	180	8.90	0.78	204	10.71	0.91

Return To Table Of Contents

Table 22: Distribution of employment status of respondents in rounds 1, 2 and 3, by incentive level
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7924	100	0	2906	100	0	2502	100	0	2516	100	0
Round 1 - Emplyed on Interview Date	4853	61.75	0.84	1782	61.43	1.20	1513	61.08	1.30	1558	62.78	1.35
Round 1 - Emplyd during rnd, not on Int. Date	53	0.52	0.09	21	0.49	0.14	16	0.63	0.19	16	0.45	0.15
Round 1 - Not emplyd this rnd, but emplyd in NHIS	175	2.22	0.22	51	1.80	0.33	61	2.68	0.41	63	2.24	0.33
Round 1 - Not emplyd this rnd and no past emplmnt	2843	35.51	0.81	1052	36.28	1.19	912	35.61	1.28	879	34.52	1.31
Round 2 - All	7728	100	0	2750	100	0	2457	100	0	2521	100	0
Round 2 - Emplyed on Interview Date	4656	60.82	0.81	1628	59.90	1.17	1480	60.82	1.28	1548	61.81	1.32
Round 2 - Emplyd during rnd, not on Int. Date	11	0.16	0.06	2	0.07	0.06	4	0.19	0.13	5	0.23	0.12
Round 2 - Not emplyd this rnd, but emplyd last rnd	407	4.73	0.25	155	4.95	0.45	127	4.72	0.49	125	4.51	0.50
Round 2 - Not emplyd this rnd and no past emplmnt	2654	34.29	0.80	965	35.08	1.08	846	34.27	1.30	843	33.44	1.41
Round 3 - All	7595	100	0	2667	100	0	2410	100	0	2518	100	0
Round 3 - Emplyed on Interview Date	4465	59.82	0.78	1547	58.70	1.09	1411	59.49	1.29	1507	61.32	1.47
Round 3 - Emplyd during rnd, not on Int. Date	19	0.29	0.07	10	0.41	0.14	1	0.03	0.03	8	0.40	0.16
Round 3 - Not emplyd this rnd, but emplyd last rnd	413	4.90	0.28	138	4.94	0.50	127	4.96	0.44	148	4.81	0.50
Round 3 - Not emplyd this rnd and no past emplmnt	2698	34.99	0.78	972	35.95	1.07	871	35.52	1.28	855	33.47	1.44

Return To Table Of Contents

Table 23: Distribution of respondent marital status for rounds 1, 2 and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7950	100	0	2915	100	0	2512	100	0	2523	100	0
Round 1 - Married	3713	48.93	0.84	1392	49.90	1.25	1145	46.98	1.42	1176	49.74	1.43
Round 1 - Widowed	635	8.63	0.46	212	7.89	0.62	221	9.55	0.81	202	8.56	0.79
Round 1 - Separated	1105	14.87	0.56	419	15.82	1.03	340	14.77	0.89	346	13.88	0.92
Round 1 - Divorced	363	3.51	0.23	142	3.75	0.45	104	3.18	0.45	117	3.57	0.47
Round 1 - Never Married	2134	24.05	0.64	750	22.63	1.13	702	25.52	1.30	682	24.24	1.04
Round 2 - All	7738	100	0	2754	100	0	2461	100	0	2523	100	0
Round 2 - Married	3540	47.95	0.78	1291	48.94	1.28	1099	46.63	1.43	1150	48.16	1.35
Round 2 - Widowed	627	8.76	0.47	209	8.17	0.61	218	9.71	0.85	200	8.48	0.77
Round 2 - Separated	1129	15.38	0.58	415	16.32	1.01	348	15.29	0.92	366	14.43	0.94
Round 2 - Divorced	337	3.45	0.24	125	3.43	0.46	110	3.42	0.46	102	3.52	0.54
Round 2 - Never Married	2105	24.45	0.65	714	23.14	1.15	686	24.95	1.30	705	25.41	1.06
Round 3 - All	7604	100	0	2672	100	0	2414	100	0	2518	100	0
Round 3 - Married	3433	47.41	0.79	1231	48.20	1.29	1072	46.49	1.54	1130	47.44	1.34
Round 3 - Widowed	616	8.74	0.45	201	8.24	0.60	211	9.48	0.81	204	8.56	0.76
Round 3 - Separated	1122	15.39	0.61	412	16.32	1.05	342	15.00	0.94	368	14.77	1.02
Round 3 - Divorced	341	3.80	0.28	130	3.83	0.48	113	3.96	0.54	98	3.62	0.54
Round 3 - Never Married	2092	24.66	0.67	698	23.40	1.17	676	25.08	1.31	718	25.61	1.13

Return To Table Of Contents

Table 24: Distribution of self-reported Health Status for respondents in rounds 1, 2, and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7936	100	0	2912	100	0	2507	100	0	2517	100	0
Round 1 - Excellent	1847	25.75	0.65	669	26.06	1.04	580	25.05	1.15	598	26.08	1.03
Round 1 - Very good	2385	31.48	0.61	834	29.49	1.03	790	33.05	1.15	761	32.22	1.22
Round 1 - Good	2217	25.88	0.60	839	26.33	1.00	681	25.72	1.02	697	25.52	1.00
Round 1 - Fair	1095	12.18	0.49	419	13.06	0.88	334	11.61	0.80	342	11.72	0.72
Round 1 - Poor	392	4.72	0.29	151	5.06	0.53	122	4.58	0.51	119	4.46	0.59
Round 2 - All	7732	100	0	2752	100	0	2459	100	0	2521	100	0
Round 2 - Excellent	1617	22.88	0.69	572	23.28	1.32	516	23.08	1.10	529	22.26	0.98
Round 2 - Very good	2483	33.76	0.74	860	31.90	1.20	814	35.19	1.23	809	34.41	1.11
Round 2 - Good	2284	28.21	0.60	822	28.94	1.16	716	27.17	1.06	746	28.42	0.91
Round 2 - Fair	1034	11.46	0.41	380	11.81	0.70	316	11.03	0.72	338	11.48	0.73
Round 2 - Poor	314	3.69	0.26	118	4.07	0.46	97	3.54	0.43	99	3.43	0.48
Round 3 - All	7601	100	0	2669	100	0	2414	100	0	2518	100	0
Round 3 - Excellent	1617	23.52	0.66	544	22.49	1.14	517	23.78	1.17	556	24.36	1.00
Round 3 - Very good	2369	33.11	0.70	809	32.23	1.26	771	33.31	1.13	789	33.86	1.16
Round 3 - Good	2310	28.37	0.64	843	29.80	1.13	724	28.11	1.01	743	27.09	1.09
Round 3 - Fair	1011	11.08	0.44	363	11.29	0.76	311	10.87	0.78	337	11.06	0.72
Round 3 - Poor	294	3.93	0.27	110	4.20	0.50	91	3.94	0.54	93	3.62	0.49

Return To Table Of Contents

Table 25: Distribution of RU size for responding RU's in Rounds 1, 2, and 3, by incentive level
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	7950	100	0	2915	100	0	2512	100	0	2523	100	0
Round 1 - 1	2200	30.40	0.73	790	29.30	1.19	705	31.91	1.29	705	30.16	1.21
Round 1 - 2	2172	30.72	0.70	783	31.27	1.12	691	29.90	1.06	698	30.90	1.22
Round 1 - 3	1320	15.24	0.45	488	14.94	0.76	432	15.81	0.85	400	15.02	0.80
Round 1 - 4	1200	13.52	0.45	443	13.64	0.74	377	13.17	0.91	380	13.73	0.77
Round 1 - 5+	1058	10.13	0.47	411	10.86	0.71	307	9.21	0.68	340	10.19	0.74
Round 2 - All	7738	100	0	2754	100	0	2461	100	0	2523	100	0
Round 2 - 1	2196	31.12	0.68	750	29.49	1.16	703	32.32	1.37	743	31.74	1.10
Round 2 - 2	2114	30.74	0.70	766	32.17	1.22	660	29.29	1.16	688	30.59	1.19
Round 2 - 3	1228	14.52	0.48	434	13.72	0.78	414	15.60	0.82	380	14.34	0.86
Round 2 - 4	1152	13.25	0.49	412	13.40	0.81	378	13.44	0.94	362	12.91	0.71
Round 2 - 5+	1048	10.36	0.47	392	11.22	0.75	306	9.35	0.72	350	10.41	0.75
Round 3 - All	7605	100	0	2673	100	0	2414	100	0	2518	100	0
Round 3 - 1	2197	31.77	0.69	747	30.49	1.10	694	32.54	1.46	756	32.40	1.15
Round 3 - 2	2093	30.69	0.73	744	31.70	1.26	661	29.85	1.19	688	30.42	1.21
Round 3 - 3	1185	14.36	0.47	408	13.53	0.81	395	15.43	0.87	382	14.22	0.85
Round 3 - 4	1107	12.98	0.49	404	13.64	0.83	371	13.03	0.88	332	12.23	0.73
Round 3 - 5+	1023	10.20	0.47	370	10.65	0.70	293	9.15	0.77	360	10.73	0.77

Return To Table Of Contents

Table 26: Distribution of MSA status for responding RU's in Rounds 1, 2 and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	8002	100	0	2930	100	0	2530	100	0	2542	100	0
Round 1 - MSA	6890	84.04	1.32	2488	83.95	2.12	2201	84.02	1.87	2201	84.17	2.00
Round 1 - Non-MSA	1112	15.96	1.32	442	16.05	2.12	329	15.98	1.87	341	15.83	2.00
Round 2 - All	7797	100	0	2771	100	0	2486	100	0	2540	100	0
Round 2 - MSA	6686	83.50	1.31	2339	83.18	2.17	2154	83.48	1.94	2193	83.87	1.98
Round 2 - Non-MSA	1111	16.50	1.31	432	16.82	2.17	332	16.52	1.94	347	16.13	1.98
Round 3 - All	7671	100	0	2692	100	0	2437	100	0	2542	100	0
Round 3 - MSA	6566	83.40	1.28	2264	83.13	2.24	2110	83.36	1.86	2192	83.72	1.97
Round 3 - Non-MSA	1105	16.60	1.28	428	16.87	2.24	327	16.64	1.86	350	16.28	1.97

Return To Table Of Contents

Table 27: Distribution of responding RU's by region of country for Rounds 1, 2 and 3, by incentive group
Respondent Group	Across Incentive Groups n	Across Incentive Groups (%)	Across Incentive Groups s.e.	$30 Incentive n	$30 Incentive (%)	$30 Incentive s.e.	$50 Incentive n	$50 Incentive (%)	$50 Incentive s.e.	$70 Incentive n	$70 Incentive (%)	$70 Incentive s.e.
Round 1 - All	8002	100	0	2930	100	0	2530	100	0	2542	100	0
Round 1 - Northeast	1291	16.30	0.58	475	17.03	1.24	423	16.69	1.51	393	15.06	1.53
Round 1 - Midwest	1556	24.28	0.90	568	24.61	1.90	489	24.36	2.07	499	23.83	1.89
Round 1 - South	3086	37.60	0.85	1120	37.39	1.78	959	36.60	2.07	1007	38.84	1.99
Round 1 - West	2069	21.82	0.74	767	20.97	1.29	659	22.35	1.65	643	22.27	1.51
Round 2 - All	7797	100	0	2771	100	0	2486	100	0	2540	100	0
Round 2 - Northeast	1227	15.73	0.57	438	16.45	1.26	416	16.45	1.46	373	14.26	1.45
Round 2 - Midwest	1543	24.73	0.88	552	25.28	1.93	479	24.44	2.09	512	24.42	1.92
Round 2 - South	3022	37.82	0.89	1050	37.21	1.94	953	37.04	2.10	1019	39.24	1.96
Round 2 - West	2005	21.72	0.77	731	21.07	1.32	638	22.08	1.66	636	22.08	1.49
Round 3 - All	7671	100	0	2692	100	0	2437	100	0	2542	100	0
Round 3 - Northeast	1191	15.55	0.59	414	15.96	1.23	408	16.46	1.44	369	14.24	1.46
Round 3 - Midwest	1531	25.03	0.88	544	25.76	2.03	479	24.94	2.15	508	24.35	1.91
Round 3 - South	2973	37.70	0.88	1035	37.43	2.00	919	36.38	2.16	1019	39.25	1.95
Round 3 - West	1976	21.72	0.77	699	20.84	1.36	631	22.22	1.73	646	22.17	1.47