Modified Adjusted Gross Income (MAGI) Income Conversion Methodologies. Section II: Disregard Methods

03/01/2013

This section of the paper summarizes the Department’s analysis of the potential methods to reflect varying state disregards, and the reasons for its selection of the Standardized MAGI Conversion Methodology.  The Disregard Methods convert a state’s current Medicaid net income standard to an equivalent standard that accounts for disregards.  The Department evaluated three types of Disregard Methods: the Same Number Net and Gross Method, the Average Disregard Method, and the Marginal Disregard Method.

Same Number Net and Gross: The Same Number Net and Gross (SNNG) Method uses the assumption that the new converted standard is based only on gross income, defined as net income plus disregards.  (However, in 2014, the rules for assessing eligibility will account for other adjustments besides disregards; therefore, this method does not actually achieve the goal of establishing a standard where the same number of people are eligible in 2014 as were eligible under the existing standard.)  Under the SNNG method, the converted income standard is set so that the same number of people are eligible under a gross income standard, as defined above, as would be eligible under the existing net income standard.  This method uses SIPP data to estimate the number of individuals eligible in a particular state and category according to the current income eligibility rules.  All individuals in that state and category are then ranked from lowest to highest gross income, which is defined as the sum of net income plus disregards, and the converted standard is set at the point at which the number of individuals eligible based on gross income is equal to the number eligible under the existing rules.  This method is illustrated in the text box on the following page.


SNNG Example (Tables A and B)

The example relates to a specific eligibility category within a state. Under March 23, 2010 eligibility rules, 3 individuals in a group of 6 individuals are eligible for Medicaid: Person 1, Person 2, and Person 3 (see Table 1). When the gross income is ranked from lowest to highest and the new converted standard is drawn, 3 individuals are still eligible. However, these are not the same 3 individuals who were eligible under the net income standard (see Table 2).

Table A: Example before conversion, SNNG

Eligibility Using State-Eligibility Category Specific Net Income Standard on March 23, 2010

Individual

% FPL using Net Income

Eligible under net income standard of 100% FPL?

Person 1

70%

Yes

Person 2

85%

Yes

Person 3

90%

Yes

Person 4

110%

No

Person 5

120%

No

Person 2

130%

No

Table B: Example after conversion, SNNG

Eligibility Using State-Eligibility Category Specific Converted Gross Standard

Individual

% FPL using Net Income

%FPL using Gross Income

Eligible under converted income standard of 110%

Person 1

70%

90%

Yes

Person 2

85%

105%

Yes

Person 4

110%

110%

Yes

Person 3

90%

120%

No

Person 5

120%

130%

No

Person 6

130%

150%

No

In this example, the converted standard would be 110% FPL.  The bolded observations in Table 2 represent individuals who changed eligibility status. This method can only be implemented with SIPP data because state administrative data include information only about individuals who apply and/or are enrolled in Medicaid.


The SNNG method does not ensure that the exact same individuals retain eligibility, but aims to ensure that the same number, in the aggregate, would be eligible if gross income were used to assess eligibility.  For example, Person 4 gains eligibility under the converted standard of 110% FPL based on gross income, but was not eligible under the net income standard of 100% FPL.  Because his income level does not change in this example, whether it is calculated as net income or gross income, Person 4 had no disregards.  However, Person 3 loses eligibility under the converted standard because his gross income exceeds 110%, whereas his net income of 90% made him eligible under the net income standard of 100%.  This implies that Person 3 has disregards that affect his eligibility.  Under the converted standard, Person 4 gains eligibility while Person 3 loses eligibility, but they balance each other in the aggregate.

The SNNG Method has three important limitations:

  1. The method can only be used with SIPP data.  It cannot be used with state administrative data because state data do not provide any information on individuals who are not enrolled in Medicaid or CHIP.  
  2. The accuracy of the SNNG Method is heavily dependent on the accuracy of the eligibility simulations on which it is based.   While this is true to some extent for all the methods the Department analyzed, it is truer for the SNNG Method.  The SNNG approach depends on accurately approximating more target numbers than the other methods.  In addition, there are not good benchmarks for setting many of these targets.  Appendix 3 describes in detail the difference between SNNG and the Department’s recommended Standardized MAGI Conversion Methodology.
  3. The SNNG only accounts for disregards and does not address the impact of adjustments for other MAGI rules, such as household composition and income counting.  Because eligibility in 2014 will be determined using these other MAGI rules, SNNG does not set the converted standard at the point where the same number who are eligible under current rules will be eligible under 2014 rules.

Average Disregard Method: The Average Disregard Method (ADM) takes the value of disregards assigned to each individual, converts this disregard across all individuals to a percentage of the FPL (where the FPL is determined based on household size), sums up the individual’s disregards expressed as a percent of FPL, and then divides this number by the total number of eligibles or enrollees in that state and eligibility category.  This method can be implemented using either SIPP or state data.  If SIPP data are used, eligible individuals are evaluated.  If state data are used, enrolled individuals are evaluated.

ADM Example

The current net income standard for a specific eligibility group is 100% FPL. Among a group of eligible (or enrolled) individuals, the average value of the disregards is 3% of the FPL.  The converted standard is 100% + 3% = 103% FPL.

The ADM has two strengths: First, it adjusts for disregards, which are a major component underlying the difference between current net income standards and MAGI-based standards.  Second, it can be implemented using either SIPP or state data, at the option of the state.  Moreover, for the eligibility categories examined, the ADM produced eligibility standards that were generally consistent with the results produced by other methods.  (See Table 5 below, summarizing the income standards resulting from each method for various eligibility categories.)

However, the Department concluded that the ADM is likely to produce a systematically biased result in which, on average, more people lose eligibility than gain eligibility as a result of the conversion.  This bias exists because, as shown below, there is a systematic relationship between the average size of the disregard and net income.  People with higher levels of net income have, on average, higher disregards.  In contrast, the amount of the disregard is irrelevant for most people with low levels of net income.  They will be eligible regardless of the size of the disregard.  The people for whom the size of the disregard is most likely to affect eligibility are the people whose income is a bit above the net standard.  Because, as shown below, average disregards are greater for these people than for those with lower levels of net income, the average disregard method systematically estimates a lower average disregard than a method that captures disregards for those on the “margin,” and causes the converted threshold to be biased downwards from the level needed to be unbiased. 

The income disregard amounts for individuals whose income is within a relatively small margin of the eligibility standard may be higher than the income disregard amounts for individuals whose income is well below the eligibility standard, for at least two reasons: 

  • Certain income disregards, such as child care expenses and work expenses, are only available for individuals with earned income.
  • States do not always collect complete disregard information for all individuals. When states are assessing eligibility for individuals whose income is significantly below the net income standard, states have little or no reason to collect full disregard information.  The applicant’s income is low enough to meet the eligibility standard regardless of whether he or she has applicable disregards.  While not collecting complete information reduces both the state’s administrative burden and the burden on applicants, it results in a potentially biased converted standard if the average disregard method is used.   

For these reasons, the Department refined the Average Disregard Method by examining bands of income and taking the average disregard only for individuals for whom disregards affect eligibility.

Marginal Disregard Method (selected as the Department’s recommended method):  The Department tested the hypothesis that disregards are positively correlated with income and confirmed that a positive correlation exists.  As shown in Appendix 1, there is a correlation between level of income and level of disregards.  Higher income individuals tend to have higher disregards.  Therefore, the Department concluded that the conversion methodology should focus on those individuals for whom disregards are most likely to affect their Medicaid eligibility.  These are the individuals whose net income falls just at or below the current Medicaid eligibility standard.  It is this group for which the amount of their disregards is most important in determining whether they are eligible.

This refinement of the Average Disregard Methodology is called the Marginal Disregard Method (MDM).  It is similar to the Average Disregard Method in that disregards are expressed as a percentage of FPL and are totaled and divided by the number of enrollees (if state data is used) or eligibles (if SIPP data is used).  The difference between the methods is that, for the Marginal Disregard Method, this calculation is performed only for a subset of enrollees or eligibles.  The Marginal Disregard Method, like the Average Disregard Method (and unlike the SNNG Method), can be used with either state or SIPP data.

MDM/25 Example

Assume the current net income standard is 100% FPL. Using SIPP data and analyzing only those individuals with net incomes between 75% and 100% FPL, the average amount of disregarded income is 10% FPL. This results in a converted standard of 110% FPL.

Selecting a Marginal Income Band: The Department used two criteria to select an appropriate income band:

  • The income band should reasonably capture individuals whose eligibility is affected by disregards.
  • The population, or sample size included in the income band, should be large enough to precisely measure the average. In other words, the sample size used to compute the “marginal” average should be large enough that the converted standard is valid and reliable.

The Department analyzed the average size of disregards to understand how many individuals have disregards that could affect eligibility and to determine the marginal band size that would produce the most precise converted standard. 

Table 1 shows the percentile distribution of individuals’ disregards for a sample of 16 eligibility categories from the pilot states.  For each eligibility category, at least 5% of individuals have no disregards, and for all states and eligibility categories shown, except for Nebraska CHIP children, at least 25% have no disregards.  For 12 of the 16 categories, at least 50% of the individuals have no disregards. This shows that within most eligibility categories, disregards do not affect eligibility for at least 50% of individuals. 

For example, the first category shown in Table 1 is Arizona children less than age one.  Of the 490 observations in this category, more than 245 (the 50th percentile) have zero disregards.  The disregard for the 367th observation (the 75th percentile) is 5.9% of FPL.  The disregard for the 441stobservation (the 90th percentile) is 7.4% of FPL.  In summary, Table 1 shows that very few individuals have disregards, and of those who do, the size of the disregard is less than 25% of FPL for almost all observations.

TABLE 1: Distribution of Total Disregards as a Percent of FPL, by Percentile

Category

Number of Observations

Min.

1st pctl.

5th pctl.

10th pctl.

25th pctl.

Mean

50th pctl.

75th pctl.

90th pctl.

95th pctl.

99th pctl.

Max.

AZ:  Children < 1

490

0.0

0.0

0.0

0.0

0.0

3.1

0.0

5.9

7.4

9.8

19.0

40.4

AZ:  Children 1-5

2879

0.0

0.0

0.0

0.0

0.0

3.2

3.3

4.9

8.0

9.8

14.8

40.4

AZ:  Children 6-18

5684

0.0

0.0

0.0

0.0

0.0

3.0

0.0

4.9

8.4

10.7

15.1

44.5

AZ:  Parents

4171

0.0

0.0

0.0

0.0

0.0

3.0

0.0

4.9

8.4

11.5

17.4

44.5

AZ:  CHIP

1881

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

NE:  Children < 1

500

0.0

0.0

0.0

0.0

0.0

3.7

0.0

6.6

9.3

10.9

22.7

41.2

NE:  Children 1-5

2831

0.0

0.0

0.0

0.0

0.0

3.7

3.6

6.6

8.2

10.9

16.5

80.3

NE:  Children 6-18

5248

0.0

0.0

0.0

0.0

0.0

3.2

0.0

5.4

9.3

11.1

16.5

80.3

NE:  Parents

2430

0.0

0.0

0.0

0.0

0.0

1.8

0.0

0.0

6.6

9.3

16.5

49.4

NE:  CHIP

2085

0.0

0.0

0.0

2.9

5.4

7.4

6.6

9.3

13.1

14.0

19.7

70.0

NY:  Children < 1

575

0.0

0.0

0.0

0.0

0.0

3.1

3.3

4.9

7.4

8.8

15.7

40.4

NY:  Children 1-5

2812

0.0

0.0

0.0

0.0

0.0

3.2

0.0

4.9

7.4

10.3

17.4

79.6

NY:  Children 6-18

6294

0.0

0.0

0.0

0.0

0.0

2.9

0.0

4.9

7.4

10.3

15.7

79.6

NY:  Parents

5633

0.0

0.0

0.0

0.0

0.0

1.5

0.0

0.0

5.9

7.4

15.7

61.6

NY:  CHIP

2395

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

0.0

WV:  Parents

1616

0.0

0.0

0.0

0.0

0.0

0.4

0.0

0.0

0.0

3.9

9.8

39.3

Source:  Analysis of SIPP Data

The Department chose the marginal band of 25 percentage points of FPL for two primary reasons:

  • For all eligibility categories examined, most individuals have disregards of less than 25 percentage points of FPL.
  • A band of 25 percentage points creates a large enough sample of individuals for the converted standard to be stable. 

Because at least 99% of individuals have disregards less than 25 percentage points of FPL, the marginal band should capture virtually everyone who could be made eligible by disregards.  In addition, because the SIPP is a sample of individuals, further reducing the number of individuals used for conversion could adversely affect the stability of the conversion pools.  

The Department considered using a band smaller than 25 percentage points of FPL.  However, as shown in Tables 4A, 4B, 4C, 4D, 4E, and 4F in Appendix 4, within the 25% band, there is no systematic relationship between net income and the size of the disregard.  That is, people with income within 5 percentage points of the income standard do not have larger disregards than the average person within the 25% band.  Therefore, a band of 25 percentage points is broad enough to ensure that it captures virtually all of those for whom disregards matter, while a smaller band might have omitted some of them.  Also, as noted above, a broader band creates a large enough sample to ensure that the converted standard is stable.    

The Department concluded that the Marginal Disregard Method with a band of 25 percentage points of FPL (MDM/25) results in a converted income standard that is not systematically biased and that focuses on those individuals for whom disregards matter.  For these reasons, the Department recommended this method as the “Standardized MAGI Conversion Method” in the December 28, 2012 CMS Guidance.  As noted in the CMS Guidance, the preferred method can be used with either SIPP data or state administrative Medicaid and CHIP data.  States can choose which data they prefer to use.  States may also suggest an alternative methodology, entitled “State Proposal Option.”

View full report

Preview
Download

"rb.pdf" (pdf, 747.87Kb)

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®