Skip to main content

Table 5 Average posteriors of individual-level coefficients and relative treatment effects by HIL subgroup

From: Does targeted information impact consumers’ preferences for value-based health insurance? Evidence from a survey experiment

  

All

 

Low HIL

 

High HIL

  

Treatment 1

Treatment 2

 

Treatment 1

Treatment 2

 

Treatment 1

Treatment 2

Model B (alternative to status quo)

 

-1.587

  

-1.614

  

-1.411

 
  

+ 9.3%

+ 19.6%

 

+ 19.4%

+ 25.7%

 

+ 7.9%

+ 12.9%

  

(2.4%)

(2.5%)

 

(7.3%)

(7.6%)

 

(7.2%)

(7.5%)

VBID elements

         

Copayment

0% high-value care

 

0.570

  

0.561

  

0.572

 
   

-44.2%

-44.1%

 

-45.0%

-44.8%

 

-38.2

-40.2%

   

(1.2%)

(0.8%)

 

(3.5%)

(2.5%)

 

(3.8%)

(3.2%)

 

20% low-value care

 

-0.982

  

-0.941

  

-1.002

 
   

-30.6

-31.5%

 

-36.9

-46.3%

 

-29.9%

-28.8%

   

(2.7%)

(3.0%)

 

(9.5%)

(10.1%)

 

(8.1%)

(8.9%)

Drug disbursement based on value

 

-0.909

  

-0.919

  

-0.925

 
  

+ 21.8%

+ 17.7%

 

+ 25.1%

+ 20.5%

 

+ 21.4%

+ 17.3%

  

(0.8%)

(0.7%)

 

(2.4%)

(2.2%)

 

(2.3%)

(2.0%)

General cost-sharing

         

Deductible

CHF 0

 

1.178

  

1.141

  

1.152

 
   

-20.6%

-1.3%

 

-14.4%

+ 6.3

 

-25.2%

-8.3%

   

(1.8%)

(1.8%)

 

(5.7%)

(6.1%)

 

(5.3%)

(5.4%)

 

CHF 5000

 

-4.178

  

-4.198

  

-4.238

 
   

+ 0.0

-16.6%

 

+ 6.4

-12.9%

 

-1.6

-16.9%

   

(1.1%)

(1.3%)

 

(4.0%)

(4.3%)

 

(3.3%)

(3.7%)

Monthly premium

 

-0.034

  

-0.037

  

-0.033

 
  

-2.8

-3.4%

 

-3.6%

-2.7%

 

-2.5

-2.6%

  

(1.9%)

(1.9%)

 

(5.9%)

(6.3%)

 

(6.3%)

(6.4%)

  1. Source: Swiss Health Insurance Literacy Survey 2021. Notes: The table shows the mean values of the posterior estimates of the individual-level coefficients after the mixed logit regressions for the control group in italics (baseline values) for the overall sample (n = 6033), a subsample with low HIL (bottom 10% of the HIL scale; see (35) for details on the construction of the scale; n = 594), and a subsample with high HIL (top 10% of the HIL scale; n = 610). Treatment effects are estimated using linear regressions with robust standard errors from the posterior estimates using the data from the control group and the two treatment groups (as shown in Table 4), overall and by HIL subsample. Since baseline preferences vary by subsample, treatment effects are reported in percent values relative to the baseline estimates for better comparability. Standard errors (in parentheses) are calculated using the delta method from the estimated baseline and treatment effects in the linear regressions