OCP Summary

OCP - CA SUMMARY

ABC FEDERAL CREDIT UNION

LOW-INCOME DESIGNATION REQUEST SUMMARY

Name:	ABC	Charter #:	12345
Location:	Anyplace, USA	Total Members:	58,504
Assets:	$718,055,041	Net Worth:	8.47%
Shares:	$611,677,814	Return on Assets:	.72%
Loans:	$610,097,403	Delinquency:	.69%
SE/EX:	Jackson/Chase	Net Charge-Offs:	.74%
Last Exam:	03/31/13	CAMEL:	1-2-2-1-2 = 2

Financial Data: June 30, 2013 FPR

BACKGROUND

XXXXX requested a low-income designation (LID) in accordance with Part 701.34 of NCUA’s Rules and Regulations (Rules).

XXXXX is a community charter serving San Luis Obispo County, California and northern Santa Barbara County, California.

Based on the LID Assessment Tool, the credit union does not qualify as low-income. The initial review of all credit unions in July 2012 showed 40.65 of the credit union’s members qualified as low income with 8.71 percent of the addresses listed as a P.O. Box.

As the total number of low-income members and P.O. Boxes is less than 50 percent, the credit union submitted a request for the LID using an alternative methodology.

They wanted to use Experian’s Income Insight W2 valuation model. Using the model, Experian analyzed the credit union’s entire loan portfolio of 30,665 borrowers. These borrowers represent 50.9 percent of the credit union’s total membership. The income estimates Experian provided showed 50.3 percent of the credit union’s members qualify as low-income.

We told the credit union we cannot use Experian’s model. It does not meet the requirements in the Rules as an alternative method.

We discussed alternatives with the credit union and provided them with draft information on sampling procedures. They hired a marketing firm and decided to complete a telephonic member income survey.

The Office of Chief Economist (Economist) reviewed the credit union’s survey methodology. The Economist stated we can approve the methodology; however, we will decide on the applicability of their survey for getting the LID designation when we see the actual results and the response rates. We could not guarantee applicability until we see the results.

Their main concerns where:

1. People tend to under-report their income on surveys, thereby introducing a possible bias and over-estimate of the share of members with incomes below any threshold. To help with the under-reporting of income levels, they recommended the credit union ask for adjusted gross income from the member’s latest tax return.

2. There may be a high non-response rate on the survey’s income questions. When completed, the Economist wanted to review response rates by question and by region. They stated high non-response on the income question and/or over-representation of responses in regions with lower-income members may render the survey unusable for the purposes of LID designation.

Based on the Economist’s review and recommendations, we told the credit union, the methodology as proposed is consistent with regulatory requirements. However, we would wait to make our final conclusion regarding the validity of the survey until after we were able to evaluate actual survey results and response rates.

We noted their methodology did not include members under 18, those living outside of California, business accounts, and P.O. Boxes in the survey. Although they did not include them, they must still factor in the entire membership when determining the share of members that are low-income.

Thus, they needed to include in the total membership those excluded from the survey when determining the number of members that are low-income. The exception was for those members in high school.

As the Rules automatically deem high school students qualify as low-income. They can count the portion of members between the ages of 14 and 18 as low-income if they could support they were enrolled in high school. If not, they could reduce the number by an appropriate, supported dropout rate.

They could include those under 14 if they met the credit union’s definition of member as defined in the bylaws. If they did not meet the definition, they were not to include them in the membership count.

We asked them to provide the results in the following format

Credit Union Members by Survey Group		Low-Income Response %		Survey Group % of Mbrs		% CU Members Low-Income
Mbrs in the 7 regions		X%	*	77.86%	=	X%
Mbrs Outside 7 regions live in CA	+	X%	*	4.77%	=	X%
Students (14 to 18)	+	X%	*	3.33%	=	X%
P.O. Boxes	+	0.00%	*	2.53%	=	0.00%
Businesses	+	0.00%	*	0.37%	=	0.00%
Not Surveyed (outside CA)	+	0.00%	*	5.85%	=	0.00%
Under 14	+	0.00% or X%	*	5.29%	=	0.00% or X%
Total				100.00%		X%

They needed to change the methodology for income comparisons. They planned to compare income to city median incomes for each respondent city rather than by metropolitan or non-metropolitan area as the case warranted.

Finally, for greater consistency, we suggested they ask for the member’s adjusted gross income from the last tax return in lieu of their annual income before taxes.

SUPPORTING DOCUMENTATION

Credit Union Submittal: The credit union provided the completed survey stating 52.07 percent of the membership qualified as low-income.

To arrive at that percentage, the credit union counted all 14 and under year olds as low-income. They also used a California dropout rate of 14.4 percent for high school students rather than at the lowest geographic level possible as stated in our letter. The credit union further compounded the error as they have members between 14 and 18 in 103 locations in 21 states.

Consequently, we asked them to provide updated dropout information at the lowest geographic level possible for each location. Using the information on the locations, XXXXX calculated a weighted average dropout rate of 2.57 percent.

Using 2.57 percent as the dropout rate, the credit union showed 52.46 percent of its members qualified as low-income.

Analysis: The review found XXXXX incorrectly calculated the percent of its members that qualified as low-income. The Office of Chief Economist also expressed concerns with the survey. We discuss their issues later under the Office of Chief Economist section.

The review found the credit union does not qualify for the LID. Our calculation shows 49.80 percent of the credit union’s members qualify as low-income.

The credit union included all accounts for those under 14 as low-income in their total. This is contrary to the instructions in our letter. We stated to include those under 14 if they meet the credit union’s definition of member as defined in the bylaws. If not, they were not to include them in the membership count.

The credit union did not show they meet their definition of member. For their support they provided Article XV, Minors. The article allows credit unions to issue shares in the name of a minor thus making them account holders. It does not provide for the members fundamental right to vote. Article V. Elections, Section 7 Minimum age requirement sets the minimum age to vote.

Not only did they include them in the membership count, they also reported 100 percent qualified as low-income. As they were not included in the survey, the instructions stated set the response rate to zero percent.

Our review of the survey indicated members may not have accurately reported their income. When the credit union asked if a member’s income figure was an estimate or exact, 24.17 percent reported the income was exact. The remaining 75.83 percent stated it was an estimate or did not answer the question.

However, we observed only 10.83 percent of the members reported their income as a whole number indicating it is the exact income figure. We found the remaining 89.16 percent reported their income in thousands indicating it is an estimate.

Office of Chief Economist: The Economist expressed concerns with the survey and its results.

The credit union’s response rate was 6 percent. The Economist stated while this is in line with standard industry practices, it is a concern. The low response rate can introduce bias potentially undermining the validity of the results.

The Economist stated they can explain the low response rate, in part; by the way the credit union conducted the survey. The company they used appears to have called new numbers when they could not reach a member initially identified.

This is not the generally accepted approach recommended by the American Association for Public Opinion Research’s “Standard Definitions” (7th edition). It can induce important biases based on the survey primarily reaching individuals who are at home and willing to answer the phone at a particular time of day. To limit this bias, the recommended approach is to focus on making multiple attempts to reach individuals in the survey frame.

A quote from NYT/CBS on how they conduct polls effectively summarizes the recommended approach.

"Interviewers made multiple attempts to reach every phone number in the survey, calling back unanswered numbers on different days at different times of both day and evening."

The Economist had the following observations and recommendations:

Data quality issues

As they emphasized in their initial review, income surveys tend to produce inaccurate and underestimated incomes. In this case, only 24 percent of respondents said that incomes were exact when asked.

The Economist stated it is important to understand that any result based on these incomes is at best an estimate. They recommended OCP consider whether it is comfortable making decisions based on incomes that can potentially result in a higher share of low income members than warranted.

Data assumptions and benchmarking

a) The credit union used county incomes for comparisons when the Rules require they use MSA incomes. Changing these benchmarks might produce more accurate results. The credit union’s survey is correct. The counties surveyed are the MSA. Thus, while their terminology is incorrect, the income figures used are correct.

b) The credit union used what they call wage-earner income for comparison for members they identify as single filers. This is median earnings, which is not the best benchmark to use for comparisons to AGI income, which can be considerably higher than earnings and would tend to overestimate the share of low-income members. MSA individual income would be a better benchmark in this case. This is correct. The credit union survey used AGI as the income benchmark for median earnings. For wage earners, they should have asked for W-2 income and not AGI.

c) The survey questionnaire does not distinguish between individual and joint tax return filing. The credit union made inferences about which benchmark to apply based on answers to marital status and presence of dependent children. This inference may or may not be correct. For example, there may be a higher likelihood of counting a member as low-income when you compare incomes of single respondents with dependent children to median family income. This is correct. The credit union made the inference and was not consistent in its application. You can see where they counted single respondents with dependent children using both median family income and median wages.

The very low response rate and the unconventional way in which the credit union conducted the survey might be reason enough to render results of this survey unacceptable. The credit union was aware that we would make the final conclusion based on the results and that acceptance of the results was conditional on the response rate.

If OCP wants to proceed with this application, we recommend asking the surveying firm to provide us with reassurance on the validity of the methodology and the reliability of the results.

This would include obtaining data on response rates by time of day, results by time of day and perhaps even resurveying some non-respondents to see if their answers differ systematically.

Generally the question is how they justify that this approach is not biased.

RECOMMENDATION

I recommend deferring the request. We provide them with the concerns outlined by the Office of Chief Economist, our calculation of the percent of membership that qualifies as low-income, and provide them with other options. Using the LI tool with updated data, the percent of credit union members that qualify as low-income is over 48 percent with over 8 percent in P.O. Boxes. If the credit union obtained addresses for the boxes, they would probably quality for the LID.