TO: QUADRENNIAL GUIDELINE REVIEW PANEL

FROM: RICHARD J. BYRD

RE: ANALYSIS OF PSI STUDY AND RECOMMENDATION

DATE: May 26, 1999

I have been asked by the Review Panel to analyze and comment on the proposal for changes in the Virginia Child Support Guidelines done by Policy Studies, Inc., (PSI). The cover letter from Bob Owen stated that several members of the Panel had concerns regarding the economic data and methodologies used by PSI. His letter indicated that the Panel would have to "grapple with the validity of the data used by PSI." I share the concern of the Panel. I have serious reservations about the economic models used and will set forth those in this Memorandum.

For those on the Review Panel who may not know me, and would wonder about the qualifications of someone asked to do such a review, I have attached a curriculum vitae to this Memo.

We are all aware that the current Virginia support guidelines, first drafted in 1987, were based upon a national study of the costs of raising children in intact households, the basic data having been gathered and analyzed by Dr. Thomas Espenshade. The Espenshade study was incorporated into a study done by Dr. Robert Williams, who is one of the authors of the PSI study subject of this Memorandum. Although the national data used by Dr. Espenshade was drawn from a survey done by the U.S. Bureau of Labor Statistics in the early-1970’s, the child support guideline table is naturally indexed, hence so long as family consumption reasonably tracks family incomes, the original guideline table is not necessarily made invalid by the passage of time and the effects of inflation. If inflation does not affect items consumed by children differently than it affects items consumed by parents, the proportionality of their relationship should remain consistent.

A. The "Rothbarth" Estimator: The PSI study discusses the difficulty of determining the amount that a family household spends on its children. The problem is basically one of marginal expenditures versus average expenditures. That is, a family with no children spends a certain amount to heat and provide the other utilities to their home. Those utility expenses would not be increased significantly if the household also had two children. However, if the two parents and the two children were considered equally to be the consumers of the utilities, then one-half of their costs would be allocated to the children. In this situation, while the marginal increase in utilities cost due to adding the children is minimal, the average cost is great. The PSI study reviewed several analytical methods in an attempt to resolve this problem. PSI reviewed a study done by Dr. David Betson of the University of Notre Dame. Betson used the "Rothbarth estimator", which PSI found "to have the most economic validity and plausibility."

It is difficult to understand the reliance of PSI on the Rothbarth estimator. This analysis concludes that the well-being of a family depends on the amount that the family spends on tobacco and alcohol! What an absurd measuring stick to use to determine the quality of life of a family. Particularly, since the consumption of tobacco and alcohol by adults has decreased significantly since many of these studies were done, one could question this methodology on several fronts. Dr. Betson apparently found that if he used family expenditures for tobacco and alcohol, he got the same results as if he used expenditures for adult clothing and tobacco and alcohol, therefore he could just use family expenditures on tobacco and alcohol as a measurement of the well-being of a family. To accept the PSI analysis is to agree that the well-being of Virginia households is measured by their consumption of tobacco and alcohol. I think that is an irrational standard, and I certainly would not want to justify it to the General Assembly.

B. The "Engle" Estimator: The study done by PSI considered a methodology based on century-old findings of an economist, Ernst Engle, to the effect that as a family’s income rises, total spending on food increases, but the percentage of income spent on food decreases. This conclusion is certainly understandable, but where the PSI analysis goes from there is a non sequitur. The PSI reports that: "Thus, if two families of different size spend the same proportions of their incomes on food, they are deemed to be equally well off." The absurdity of such a conclusion must be evident to anyone who reads this sentence. According to PSI, from their reliance on the Engle estimator, a family earning $300,000 per year who spends $30,000 (10%) on food, is "equally well off" as a family who earns $20,000 per year and spends $2,000 (10%) on food. If you think that the $300,000 and the $20,000 families are equally well off, then you will not be bothered by the use of the Engle estimator as a basis for the Virginia child support guidelines. I find this to be an irrational basis on which to determine which family is more "well-off."

PSI admits that "no consensus has emerged that any single estimator is better than another. All have their limitations and biases." PSI goes on to find that "only the Rothbarth and Engle methodologies are without serious problems of empirical specification." Hence, PSI has based its proposal for Virginia’s child support guidelines upon the assumptions that the well-being of Virginia families depends on whether they spend the same share of their income on food, and on whether they consume the same amount of tobacco and alcohol. I am amazed that these are the best methodologies available for determining a support guideline table.

C. Intact versus Separated Households: The PSI report discusses the problem that all of the methodologies used to determine appropriate child support are based on studies of intact families, where both parents are resident in the same household. Yet, the application of the child support tables that result from the use of these methodologies is applicable solely to separated families. PSI states that the income shares model, upon which the Virginia guidelines are based, "seeks to apportion to the child the amount that the parents would have spent if the household were intact." I do not believe that conclusion necessarily follows from the use of the income shares model. The income shares model simply directs that whatever guideline amount is determined to be appropriate for the child, that amount is allocated between the parents in accordance with their relative proportions of incomes. The adoption of an income shares model does not give any mandate as to what is the appropriate total amount of support for a child. The PSI study observes that "single-parent households actually do spend less income on children than two-parent households…" Even though PSI acknowledges this fact, no adjustment is made in the data to account for this different allocation of consumption by separated families. In simple terms, PSI is collecting data on apples and using the data to suggest how to regulate oranges.

A. Needed Adjustments: As the PSI report discusses, the Betson-Rothbarth survey data only indicate the relative consumption for children in intact households. To make this data relevant to the Virginia child support guidelines, many adjustments are needed. First of all, the Virginia support guidelines add the amounts needed for daycare, for health insurance, and for extraordinary medical expenses to the guideline support determined from the statutory table. Therefore, since these costs are inherently embedded in the consumption data, these costs must be backed-out in order to generate the guideline table useful in Virginia’s statute. Additionally, as is discussed further above, the consumption data must then be related to net income data and the net income data then must be related to gross income. An analysis of these adjustments is discussed in this section.

B. Gross Income versus Net Income: All of the various methodologies were based upon consumption, that is what families actually spend for themselves and their children. Since the Virginia guideline has always been based upon a gross income standard, all of the data used as the basis for the PSI conclusions must be adjusted so as to be relevant to gross income. The study data used by PSI does not relate to gross income nor even to net income. As income rises, more of a family’s use of income is not on consumables, but may go to savings or investments. Savings are not accounted for in consumption and hence there are two levels of abstraction needed to go from consumption to gross income. First of all, there has to be some relationship used to convert from consumption to net income and then some relationship to go from net income to gross income. The assumptions used by PSI regarding taxes are probably as reasonable as one could use, but the committee should understand that the range of "error" in transposing from consumption to net income to gross income probably exceeds the "error" range of any of the other assumptions used to generate the proposed guideline table. As will be referred to later, since the table proposed by PSI is, in most income ranges, not vastly different from the existing table, such a range of error in determining gross income itself totally obscures the relatively small difference between the proposed table and the existing table.

C. Adjustments for Daycare and Medical Costs: In the introduction to the PSI study, it cites the federal requirements that the state quadrennial review of child support "must include an assessment of the most recent economic data on child-rearing costs…" It is noteworthy that the figures used to convert consumption data to net income data used in the PSI study are based on the Consumer Expenditure Survey (CEX) data from 1980 through 1986. The CEX income data used constant 1983 dollars which were updated to January, 1999 dollars by use of the Consumer Price Index. One should question whether the continued use of 1980 data really meets the federal mandate of "…the most recent economic data…"

The economic consumption data is adjusted for child care costs by using the CEX database which identifies expenditures for each commodity. From this database, child care costs range from .62% of consumption for households with net incomes less than $10,997 to 1.28% of consumption for households with net incomes of approximately $50,000. Apparently, no data exists for families in higher income ranges. However, the proposed support guideline table is taken to gross incomes of $240,000 per year. It is unknown how PSI adjusted these higher incomes for daycare costs, since the CEX database they were using did not cover that range.

I question the deduction used by PSI for child care costs from a common sense standpoint. PSI used an adjustment of 1.28% of consumption for households with net incomes of $50,000. I don’t know what factor PSI used to adjust between consumption and net income, but for the sake of argument, let’s use 95% indicating a savings rate of 5%. The family with a net income of $50,000 then spends $47,500 on consumables. PSI says that this family then spends $608 per year, or $51 per month on child care. It is very hard to believe that any family in Virginia that has child care has a child care cost as low as $51 per month. If this figure used by PSI is greatly in error, it means that the child support table still includes too much consumption for daycare. This is unfair to both families in which daycare is added on to the figure derived from the support table, and to families who do not actually have daycare. In other words, it would appear to me that the guideline table figures used by PSI may include some embedded daycare costs. In order to conclude that the table does not include daycare costs, one has to believe that the average daycare for families in Virginia with children is only $51 per month. Maybe they can find cheap daycare in Wise County, but I will guaranty you don’t get much daycare for $51 per month in Fairfax!

D. Extraordinary Medical Expenses: The PSI study seems to confuse extraordinary medical expenses with unreimbursed medical expenses.. On page 16 of the Report, paragraph 3 is entitled "Deducting the Child’s Share of Unreimbursed Medical Expenses." In the paragraph that follows, PSI indicates how they determine how much of a family’s total medical expenditures are for children. They then go on to conclude that that is the proper amount for "extraordinary medical expenses." This is incorrect. PSI has equated unreimbursed medical expenses with extraordinary medical expenses, and the two are not congruent.

Later in the PSI report (Chapter 4, Summary of Key Assumptions), PSI indicates that the guideline schedule table they propose "…is based on the assumption that expenditures on ordinary medical care are $250 per child per year." They give no source for this assumption. When PSI discussed "unreimbursed medical expenses" on Page 16 of the report, they indicated that the Rothbarth studies indicated that for families with a net income of $100,000, they spent 0.58% on extraordinary medical expenses. However, in that paragraph, it is unclear whether they really meant extraordinary medical expenses, or unreimbursed medical expenses. The 0.58% would correspond to $600 per year. PSI also indicated that for a low income family with a net income between $10,000 and $16,000 per year, the amount spent for extraordinary medical expenses (or unreimbursed medical expenses) would be $476 per year. It is unknown how these calculations relate at all to PSI’s later "assumption", which is the basis of their proposed table, that expenditures on ordinary medical care are $250 per year, irrespective of the income of the family. It is unreasonable to believe that any expenditures for children, including medical expenditures, are a fixed dollar amount irrespective of family income. Such a conclusion is totally contrary to everything said by PSI based on the studies and methodologies used in the report. It would seem more realistic to use a percentage amount for ordinary medical care. When the Board of Governors of the Family Law Section of the Virginia State Bar was considering proposing legislation to eliminate the extraordinary medical expense addition, and replace it with a statutory mandate that all unreimbursed medical expenses would be shared by the parents in relation to their relative incomes, we concluded that ordinary medical expenses were between 5% and 10% of guideline table child support. That is, for a child support of $500, ordinary medical expenses would be considered to be between $25 - $50 per month. For a child support of $1,000, ordinary medical expenses would be considered to be between $50 - $100 per month. This would seem a more rational approach to deal with the issue of extraordinary and unreimbursed medical expenses.

My proposal, based upon the studies done by the Board of Governors and by the VBA Coalition, would be to eliminate the extraordinary medical expense addition to support in the Virginia guideline, reduce all supports in the guideline by a percentage between 5% and 10%. The statute would mandate that all unreimbursed medical expenses be shared by the parents in proportion to their income shares. This would be the most child support-neutral method of handling this. The PSI study retains the difficult to interpret extraordinary language of our statute, and makes an inexplicable deduction from the child support otherwise determined by Betson and Espenshade of $250 per year, irrespective of income.

E. The relationship between consumption and net income: This item was discussed in a general way earlier. PSI used the CEX data which indicated that the ratio of consumption spending to net income vary from 100% for annual net incomes below $32,991, to about 65% of net income for households with annual net incomes in excess of $109,973. It is unknown what data PSI used in the proposed table, which goes to gross incomes of $240,000 per year. One would not expect a linear relationship between income and savings rate, but PSI may well have counted for a non-linear rate in its approximations.

F. Larger Household Sizes: PSI notes that the CEX data only extend to households with three children. Hence, PSI had to extrapolate the data to get estimates for families of four, five and six children. For this, PSI (and Dr. Betson) used the Espenshade data on consumption from the early 1980’s. This again brings into question the federal mandate that the quadrennial review be based upon "the most recent economic data."

The PSI study gives several comparisons of their proposed guideline support table and the current Virginia guideline table. However, I found most of these comparisons did not exemplify the actual differences between the two guideline tables. Attached to this outline, is a comparison of the PSI table and the existing Virginia table for incomes from $1,500 per month to $20,000 per month for one, two and three children. To make the table easier to review on one page, the support is shown for each $500 increment of income, rather than for each $50 of income as shown in the PSI table. This table shows the dollar difference between the proposed and existing schedules and the percentage difference. The support differences are shown as positive numbers where the proposed table is higher than the existing table, and are negative numbers where the proposed table is lower than the existing table. As is evidenced by the few numbers having negative signs, the proposed table is higher than the existing table most levels of income. Looking at the table for one child, the proposed schedule is 10-16% higher than the existing table for low incomes, is 3-10% higher than the existing schedule for moderate incomes, and is 20-26% higher for higher incomes.

Also enclosed with this outline is a graph showing the child support for one child as proposed by PSI, and the existing guideline table. On this graph, the discrepancy between the two tables at the higher incomes is quite striking. In the current Virginia statute, the regular guideline table ends at $10,000 per month, and a series of three levels of percentages are added to the income in excess of $10,000 per month to calculate higher support. I did the original calculations that were adopted by the General Assembly for this extension of the guidelines. When I did this extension, I chose a mathematical curve which would be a "logical extension" of the guideline table from the point where it ended at $10,000 per month. As is seen by the green line on the chart, the higher level child support is a smooth curve from where the table ends at $10,000 per month. The PSI proposed support in the higher income ranges also follows a smooth curve from the level at $10,000 per month, the difference is that the slope of the PSI table is demonstrably higher at $10,000 per month. Therefore, that higher slope is translated to all incomes higher than $10,000 per month, giving a great discrepancy between the two tables in support at $20,000 per month. Although the PSI report says that the data was smooth so that there would not be "notches" in the proposed table, yet there is obviously a notch in the table in the $4,800 - $6,800 income range. The existing Virginia curve is a much smoother relationship between child support and income.

The bottom-line in comparing the two tables is that for most families in Virginia having total family incomes between $4,000 and $9,000 per month, the change in support between the existing table and the proposed table would not be large, but nor could it be called insignificant. For one child, and $4,000 per month, the support would go from $553 under the current statute to $641 under the proposed statute. This is an increase of $88 per month, or almost 14%. Remember, this is the total support paid for by both parents, and would be allocated between the parents in accordance with their income shares. Still, for the payor parent, the increase is 14%.

The largest increases in support which would result under the PSI proposed table would be for the support payable for one child when the parents make more than $12,000 per month in combined income. This income level represents less than 10% of the families in Virginia.

For families in the moderate income ranges of $5,000 to $8,000 per month in combined income, having three children, the new table would result in reductions in support of as much as 16%. However, for families in the high income ranges with three children, there would be increases in support of as much as 13%.

In Appendix 1 of the PSI report, the authors admit that determination of an appropriate level for child support to be paid by separated parents is difficult to determine both philosophically and statistically. There are policy and lifestyle considerations on which people do differ in their opinion. Should one family be considered to be more well-off than another family because the one family spends a lot of money on tobacco and alcohol, whereas the other family spends an equal amount on home computers and trips to Europe? PSI clearly casts its vote with the tobacco-alcohol family. I think I would definitely cast my vote that the family spending money for computers and trips to Europe is more well-off, but as I said, reasonable people might differ.

Apparently, the statistical economic data which might help to resolve some of these conflicts has simply not been done. The PSI study is still basing its proposals on economic data which is older than the first passage of support guidelines in Virginia. This would hardly seem to comply with the federal mandate for this quadrennial review commission to use "the most recent economic data on child rearing costs…" As PSI points out in Appendix 1, the only items available in the CEX study that would constitute adult expenditures were adult clothing, alcohol and tobacco. Hence, despite the absurdity of using such a definition to establish whether one family is more well-off than another, perhaps PSI had no choice since that was the only data that it had. As PSI pointed out further in its study, even this distinction in expenditures was not valid, since the adult clothing expenditures also included teenage clothing, and hence the three categories do not necessarily constitute expenditures by adults as opposed to children. PSI freely admits the difficulties with some of its data and the arbitrariness of some of its assumptions in Appendix 1 to the report.

I recommend that the table of guideline child support proposed in the PSI study be rejected by the Quadrennial Review Commission and that the commission recommend that we retain the present child support table. I recommend this for the following reasons:

1. In my discussion of how the current child support table is working, I have not heard any strong, unified opinion that the Virginia table yields support levels with are too high or too low. My colleagues in the domestic relations field and the many judges that I have discussed this with seem to indicate that the guideline child support seems to be either "just about right", or as many think that it is somewhat too low as think it is somewhat too high.
2. The economic data used by PSI is no more recent than the data upon which the original Virginia guideline table was based.
3. The methodology of Dr. Betson and Dr. Espenshade which was adopted by PSI makes assumptions as to their judgment of when two families are equally well-off, which assumptions would be difficult to justify to the citizens of the Commonwealth of Virginia or to our General Assembly. I would certainly not want to be the one trying to convince the General Assembly that Virginia families who spend more on tobacco and alcohol are the most well-off families in Virginia!
4. The potential error range in results due to arbitrary assumptions in the PSI study is much greater than the differences between the proposed support table and the present one. That is, if all of the assumptions made in the PSI study were made at one end of the range and then at the other, the two different tables which would result would be vastly more different than is the difference between the proposed table and the existing one. To put this another way, the current Virginia table is well within the error of the PSI study.