|
December 27, 2001
PHADA sends letter to Harvard outlining concerns on Cost Model
PHADA recently sent correspondence to Harvard University representatives conducting the public housing operating cost study. The letter outlines some of PHADA's concerns regarding the methodology used to determine costs. PHADA's most pressing concerns revolve around the fact that Harvard is not accurately factoring the age of properties into its model. In addition, PHADA notes its concerns about being compared to private, for-profit landlords, worrying that this may ultimately result in inadequate public housing funding. The full text of the letter follows.
December 20, 2001
Mr. James Stockard
Principal Investigator
Public Housing Operating Cost Study
Harvard Graduate School of Design
48 Quincy St. S402
Cambridge, Ma. 02138
Re: Cost Study Model Variables
Dear Mr. Stockard:
The Harvard Public Housing Operating Cost Study team has presented the variables it has determined affect public housing costs to the research working group. This presentation engendered a great deal of discussion. At the November 30 meeting Greg Byrne, Project Director, informed the group that he hoped to have a final version of the cost model for the January 10, 2002 gathering.
PHADA would like to take this opportunity to put in writing the concerns it has voiced about the cost study model variables in order that they be fully considered before a finalized version is chosen. PHADA's concerns fall into two categories. The first includes variables for which the properties in the FHA database do not compare with the properties in public housing. The second comprises variables which affect the FHA database, but which do not apply to public housing.
PHADA also takes this opportunity to remind Harvard of the commitments it has made to provide further information or analysis regarding various issues. A separate letter is being submitted directly by Professor Alex White outlining his observations on two variables-age and ownership type. Until all these concerns are resolved, PHADA is not convinced Harvard is ready to finalize the cost model.
Variables for Which the FHA Database is not Comparable to Public Housing
1. Age
GSD has determined that costs increase with age. At the November 30 meeting, however, it had chosen not to measure age by date of construction or date of occupancy, but rather by the date of the last major renovation. Professor White points out that for the FHA database, there is no significant difference in the effect on costs between date of occupancy and date since the last major renovation, measured by the latest mortgage refinancing date. Since there is no observable difference, it is unclear why GSD would choose an artificial construct, such as latest mortgage refinance, rather than the actual age of the property. It is clear, however, that applying the same measurement to public housing, i.e. the date of the last comprehensive modernization, would have the effect of lowering the age of some properties and thus lowering their operating cost funding.
As you are aware, as well, using the date of last comprehensive modernization poses many difficulties. For one thing, this record might not be available from HUD, or might not be available uniformly. Secondly, there is no definition of what constitutes a comprehensive modernization. It is certain that renovations called by this name will have extremely different meanings in different locations. Thirdly, comprehensive modernizations are often multiyear projects making it difficult to select a precise date.
It is also true that the developments most likely to have undertaken comprehensive modernizations will be large, urban developments, while those least likely will be elderly ones. As a result GSD's system would have the effect of possibly assigning a younger age to developments generally considered the most difficult to operate, and thus providing them with less funding, and assigning an older age to developments considered to be more manageable, and thus providing them with more funding. GSD's system would thus provide funding in inverse proportion to need.
GSD has also made the point that some of the funding in the limited dividend properties' operating expenses is used on capital improvements. In addition, there are many properties in the FHA database with significant revenue streams allowing for continual upgrading, Thus, it does not seem likely that on average during the fifteen year period following a comprehensive modernization that public housing properties would be less costly to operate than properties in the FHA database. The fact that the capital backlog for the FHA database, only $3,000 per property, is so small also emphasizes the point that for all intents and purposes the condition of the FHA database as a whole may be similar to the condition of property after a major renovation.
If on average public housing properties which have undergone a comprehensive modernization in the past fifteen years are not in significantly better condition than properties in the FHA database which are more than fifteen years old, then conceptually it would not be right to fund them at a lower level. In sum, for all of these reasons, PHADA believes strongly that GSD should use date of occupancy or date of construction to measure the age of a property.
Finally, the FHA database is unable to gauge the effect of age beyond 30 years, because it does not have an adequate sample. Yet, 65 percent of public housing properties are older than 30 years, and some of them are now more than 60 years old. Harvard must develop a method to calculate the effect on a building's operating costs of having served low-income persons in the United States for the past 60 years. Harvard has mentioned using the "Mitchell-Lama" data, but it has provided no information concerning this database or how it is applicable to this issue.
2. Property Size
PHADA has long taken issue with Harvard's finding that developments with more than 150 units will be less expensive to manage in the public housing portfolio than developments with fewer than 150 units. PHADA has pointed out that the increased density in these larger complexes create additional costs for security, resident services, maintenance and property management. GSD's own senior advisor, Jeffrey Lines, has publicly stated that he believes efficiencies of scale end at the 300 unit level and that large, urban public housing developments do not lend themselves to the model. In addition, the FHA database does not have properties in the 1000 unit and above category to compare to public housing properties of this size.
GSD has committed to examining this issue in the context of the case studies. Without having a chance to review GSD's protocol for this investigation, PHADA cannot determine whether or not this method will be successful. It may be, though, that GSD will need to examine this issue separately and apart from the case studies by establishing a representative sample of large, urban family developments and specifically reviewing their operation in order to determine what costs are associated with these properties and including them in the eventual subsidy recommendation.
3. Building Type
The question has been raised whether Harvard has adequately accounted for the effects on costs of operating family, high rise public housing. Certain research working group members, such as Bill Steinmann from the New York City Housing Authority, felt that estimating a cost differential of 1 percent for these buildings did not reflect the true cost variances brought on by items such as elevator maintenance. Questions were then raised about whether or not there was an adequate supply of family, high rise buildings in the FHA database to measure this effect. Harvard agreed to supply this information to the research working group.
It is well known that public housing, family, high rises are so difficult and expensive to manage that starting with Pruitt-Igoe in St. Louis and continuing to today's Chicago Housing Authority many of these developments have had to be demolished. Thus, it seems critical that Harvard study this subgroup in more detail in order to make sure that the residents who live in these developments will have the adequate resources to live decent lives.
GSD has acknowledged that another building type, scattered site properties, are not represented in the FHA database and that an "out of model" adjustment will need to be made. Yet, it has provided no information to date as to how it will go about estimating these costs. PHADA would like GSD to describe this methodology, so that it will have an opportunity for review.
4. The Effect of the Capital Backlog on Costs
PHADA is concerned that Harvard has not yet captured the effect of the capital backlog on public housing operating costs. GSD has heard many specific criticisms of its method, ranging from the fact that REAC inspectors did not know their data would be used in this fashion to the documented problem of inspector inconsistency and the inability to differentiate between the seriousness of identified capital deficiencies.
In addition to these problems, PHADA has a more general concern. Abt Associates has identified a capital backlog of over $20 billion in public housing. For the REAC method to be valid, Harvard would need to document that REAC inspectors identified approximately $20 billion in capital deficiencies. In other words, have they listed enough capital deficiencies multiplied by an average cost to total $20 + billion? If the REAC marker is insufficient for public housing, the question arises of whether it is also capable of identifying FHA properties with capital deficiencies.
To the extent, then, that the hypothesis that unmet capital needs will increase operating costs holds true, PHADA is concerned that GSD's cost model has not been able to measure this effect. PHADA would like GSD to demonstrate that its method of identifying capital deficiencies has accounted for the full extent of the $20 + billion backlog in public housing. If it cannot, PHADA believes Harvard must undertake a new methodology to resolve this issue, preferably one which encompasses the actual examination of the effect a capital backlog has on the operating costs of actual public housing properties.
5. Elderly/Disabled
GSD has acknowledged that program differences between the FHA database and public housing mean that there is a far greater percentage of the young, disabled population living in public housing elderly developments than living in elderly properties in the FHA portfolio. It has also agreed that this fact may have an impact on costs in public housing properties, and it has said it will study this effect further. Again, PHADA would like to know how Harvard intends to proceed with this investigation. Perhaps the methods it develops to study issues such as this one and scattered site housing can be applied to other problems such as large developments, family high rises and capital backlog.
6. Resident Demographics
GSD has understood that resident demographics can affect costs. This effect appears to be born out in the FHA database. GSD has had a problem, however, because for a portion of the FHA database there are no tenant descriptions. As a result, it is not possible to make a direct comparison between the FHA population and that of public housing. For this reason, the cost model cannot be as accurate as it might be.
Harvard has contended, however, that other measurements of population will closely replicate resident demographics. After examining several possibilities, it determined that the percent of single parent households in the census tract correlated most closely with operating cost differences in the FHA database. Yet, when resident organizations were provided with this information, they objected because they felt the use of single parent households in this fashion would imply that it was the fact of being a single parent household which caused costs to increase, an assertion they could not support. Despite explanations that the use of single parent household was only a proxy for a variety of characteristics, these groups maintained their opposition.
Consequently, GSD explored the use of another indicator, poverty rate in the census tract. Unfortunately, this indicator did not correlate as closely with costs and did not benchmark expenses as accurately as did single parent households. Clearly, then, this solution is not acceptable. GSD has said it will run the poverty rate figures again in conjunction with some other changes in the model to see if the regression analysis will calculate figures which are different from the ones reported November 30.
The tenant demographic issue is a problem for the study, because it cannot use actual demographics. Since it has to rely on a proxy, it must make sure that that proxy best represents the costs of operating property. Harvard cannot make its decision based on unrelated objections. After all, it would be the residents of public housing who would be hurt the most by a model which underestimates the cost of operating public housing.
Variables Affecting the FHA Database, but Not Public Housing
1. Percent Assisted
GSD has shown that in the FHA database assisted properties are more expensive than unassisted. This finding does not directly affect public housing, for it is neither assisted nor unassisted. Nevertheless, GSD has assumed that public housing's operating costs will be affected similarly to a group it categorizes as between 80 and 100 percent assisted. It was pointed out that at the very least, Harvard should categorize public housing as 100 percent assisted, since it is 100 percent public housing. GSD said it would look at this variable using a category defined as 100 percent assisted.
2. Ownership Type
Harvard has reported that in the FHA database, for-profit owners manage their properties less expensively than not-for-profit owners. PHADA has long contested this finding, because it believes that the profit should be included as part of the operating cost. It made this point in its comments to the draft research design arguing that for-profit properties would neither have been built nor operated without the profit motive and that therefore it is an integral part of the operating cost. GSD responded that it would investigate the role of profit in its review of asset management costs.
As with the category of assisted and unassisted, this effect is one which has no direct relation to public housing, since public housing authorities are neither for-profit owners, nor are they non-profit ones.
GSD has three choices in this situation. It could either decide that since public housing corresponds to neither category, it should simply leave this variable out of the model, or it could arbitrarily assign public housing to one or the other-for-profit or non-profit.
In its mission and its organizational culture, public housing clearly is more similar to not-for-profit organizations than for-profit ones.
Nonetheless, Harvard has decided that public housing should be compared to for-profit housing rather than not-for-profit. Its rationale, which has not been clearly articulated, appears to be that because some not-for-profit property has restrictions on the amount of money which can be retained operating expenses are not truly reflective of operating costs. This case must be made much more fully before PHADA can evaluate its validity.
GSD's second argument that for-profit owners operate more efficiently ignores the role of the profit incentive in management. An owner who can make management decisions with the knowledge that the lower the costs the greater his or her profit has a different purpose than public housing administrators. It is neither fair nor good public policy to decide that public administrators conducting the public's business must imitate an expense level achieved by operators who are able to put every cost reduction directly into their own bank accounts. Therefore, PHADA urges GSD to reconsider its position on this important issue.
3. Rent/FMR
This variable is an interesting one. Although said to measure quality, basically, GSD has found that costs vary based on revenue. The higher the revenue in comparison with an average, the higher the expenses. Thus, within the FHA database, Harvard can predict costs when it knows the amount of revenue.
Yet, it is very difficult to see how this variable can apply to benchmarking public housing costs. To benchmark the cost, GSD must place a public housing property into an FHA-based revenue category. The whole point of the cost modeling exercise, however, is to determine how much revenue a public housing property will receive. So, to place the public housing property in a revenue category, before actually knowing the revenue needed for that property, is to predetermine that property's revenue. If Harvard knows in advance what the revenue is, what is the purpose of the study?
Also, is it right to assume that public housing could not cost more than the FMR? There are now many cases of tenant-based Section 8 units which far exceed the FMR. There are communities in California which have payments standards as high as 180 percent of FMR. The Cambridge, Massachusetts authority has the ability to waive payment standards altogether and approve any rent level.
There are many housing authorities in the 120 percent range. Furthermore, many communities, including 39 of the largest MSAs use an FMR at the 50th percentile rather than the 40th. As a result, HUD permits them to go to 110 percent of the 50th percentile without its approval. These housing authorities, therefore, can lease up units without HUD approval which are higher than the average category Harvard has decided applies to most public housing. By placing public housing in this group, Harvard is artificially restricting public housing costs below what is permissible in the Section 8 program in many cities.
Beyond the circular nature of this variable and GSD's questionable assumptions about revenue, it illustrates a problem with using the FHA database as a proxy. This variable measures the effect of revenue on costs. Its relationship with quality is hypothesized but not proven. Factors other than quality may be involved in determining revenue, such as the location of the complex or its program type. Yet, the whole idea of the cost model is to determine what effect various property characteristics have on costs. If this variable measures something other than property characteristics, it is extraneous to that design.
In sum, PHADA continues to have many concerns with the cost model. It has specific questions which remain to be answered concerning the variables involving the use of age, property size, building type, the capital backlog, elderly/disabled and resident demographics as predictors of cost for public housing. In addition, there are larger conceptual issues dealing with whether variables affecting the FHA database can be reasonably assigned to public housing.
PHADA appreciates GSD taking these comments into account as it progresses, and it looks forward to continuing working together to develop a fair and accurate determination of public housing costs.
Sincerely yours,
Timothy Kaiser
cc: Michael Liu, Assistant Secretary for Public and Indian Housing, H.U.D.
Gregory Byrne, Project Director
Jeffrey Lines, Senior Advisor
Phillip Clay, Senior Advisor
| 