Tax Credits: Reasons for Cost Differences in Housing Built by For-Profit
and Nonprofit Developers (Letter Report, 03/10/99, GAO/RCED-99-60).

Pursuant to a congressional request, GAO reviewed the characteristics of
the residents and properties that have benefited from the Low-Income
Housing Tax Credit program, focusing on assessing the impact of
variations in characteristics such as the type or location of the
property or the type of tenants served.

GAO noted that: (1) while tax credit units built by nonprofit developers
cost more, on average, than units built by for-profit developers,
nonprofit developers' costs were not necessarily higher when differences
in the units' characteristics were taken into account; (2) GAO
identified four characteristics that both increased average costs and
were more likely to be associated with units built by nonprofit
developers; (3) these characteristics were: (a) location in areas with
high poverty and unemployment rates; (b) location in areas eligible for
additional tax credits (because the costs of development were high
relative to incomes in these areas); (c) large units; and (d) units in
the Northeast or Pacific regions; (4) taking these and the other
characteristics GAO studied into consideration, GAO found that the
estimated per-unit cost was $5,600 more for nonprofit developers than
for for-profit developers; (5) however, because the analysis was based
on a sample and sampling introduces uncertainty, this cost difference
could range from $1,600 less to $12,700 more for units built by
nonprofit developers; and (6) consequently, the difference in estimated
per-unit costs for nonprofit and for-profit developers was not
statistically significant.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  RCED-99-60
     TITLE:  Tax Credits: Reasons for Cost Differences in Housing Built 
             by For-Profit and Nonprofit Developers
      DATE:  03/10/99
   SUBJECT:  Housing programs
             Cost analysis
             Community development programs
             Low income housing
             Tax credit
             Nonprofit organizations
             Comparative analysis
IDENTIFIER:  HUD Low Income Housing Tax Credit Program
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Report to the Chairman, Subcommittee on Housing and Community
Opportunity, Committee on Banking and Financial Services, House of
Representatives

March 1999

TAX CREDITS - REASONS FOR COST
DIFFERENCES IN HOUSING BUILT BY
FOR-PROFIT AND NONPROFIT
DEVELOPERS

GAO/RCED-99-60

Tax Credits

(385774)


Abbreviations
=============================================================== ABBREV

  GAO -
  RCED -

Letter
=============================================================== LETTER


B-281811

March 10, 1999

The Honorable Rick A.  Lazio
Chairman
Subcommittee on Housing
 and Community Opportunity
Committee on Banking and Financial
 Services
House of Representatives

Dear Mr.  Chairman,

In March 1997, we reported\1 on the characteristics of the residents
and properties that have benefited from the Low-Income Housing Tax
Credit program and made recommendations for improvements to the
program.  In our report, we estimated that the average cost of
developing tax credit units was about $60,000.\2 After issuing our
report, we further analyzed the data collected during our study and
estimated that the average cost of units built by nonprofit
developers was about $18,000 higher than the average cost of
for-profit developers' units.\3

Because the difference in average costs between nonprofit and
for-profit developers does not take into consideration variations in
the types of units built by each, you asked us to assess the impact
of variations in characteristics such as the type or location of the
property or the type of tenants served.  To assess these differences,
we analyzed unit cost data collected for our 1997 report and 1990
Census data on distressed Census tracts.  We also interviewed
officials from national associations representing nonprofit and
for-profit developers to gain their views on other factors that could
influence development costs.  Our methodology is described further in
appendix I. 


--------------------
\1 Tax Credits:  Opportunities to Improve Oversight of the Low-Income
Housing Program (GAO/GGD/RCED-97-55, Mar.  28, 1997). 

\2 Figures in this report are estimates based on a sample of 423
properties placed in service between 1992 and 1994 and described in
detail in our 1997 report. 

\3 We estimate that, on average, nonprofit developers' units cost
$18,000 ï¿½ $12,500 more than for-profit developers' units, or from
$5,500 to $30,500 more. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

While tax credit units built by nonprofit developers cost more, on
average, than units built by for-profit developers, nonprofit
developers' costs were not necessarily higher when differences in the
units' characteristics were taken into account.  We identified four
characteristics that both increased average costs and were more
likely to be associated with units built by nonprofit developers. 
These characteristics were (1) location in areas with high poverty
and unemployment rates, (2) location in areas eligible for additional
tax credits (because the costs of development were high relative to
incomes in these areas), (3) large units, and (4) units in the
Northeast or Pacific regions.  Taking these and the other
characteristics we studied into consideration, we found that the
estimated per-unit cost was $5,600 more for nonprofit developers than
for for-profit developers.  However, because our analysis was based
on a sample and sampling introduces uncertainty, this cost difference
could range from $1,600 less to $12,700 more for units built by
nonprofit developers.  Consequently, the difference in estimated
per-unit costs for nonprofit and for-profit developers was not
statistically significant. 


   BACKGROUND
------------------------------------------------------------ Letter :2

In the Tax Reform Act of 1986, the Congress replaced existing tax
incentives for the construction of low-income housing, such as
accelerated depreciation, with tax credits to encourage the
development of affordable rental housing for households whose incomes
are at or below specified income levels.  An incentive was needed for
such housing to be built because rental income and other returns from
investment in low-income housing would generally not be sufficient to
cover the costs of developing and maintaining such properties.  One
provision of the law establishing the program set aside 10 percent of
each state's allotted credits for properties built by nonprofit
developers.  In practice, nonprofit developers have received
significantly more credits than were set aside for them. 
Specifically, our study showed that about 22 percent of the
properties that were placed in service between 1992 and 1994 were
developed by nonprofit developers.  We estimate that the average cost
of units developed by nonprofit builders during this period was about
$73,000 compared with $55,000 for for-profit developers.  Figure 1
shows how the major components of the costs of developing a unit were
distributed for nonprofit and for-profit developers. 

   Figure 1:  Comparison of
   Development Costs for Nonprofit
   and For-Profit Developers

   (See figure in printed
   edition.)

Source:  GAO's analysis of data provided by tax credit allocating
agencies. 

As shown in figure 1, the proportion of total development costs spent
on the different cost components was similar for both types of
developers.  For both, construction-related expenses accounted for
over half of the total development costs while general development
costs--which include the developer's fees, profit, and overhead, as
well as various fees for professional services--accounted for about a
fifth of the total costs.  The remaining expenses were for the
acquisition of land and buildings (where applicable); operating,
replacement and other prefunded reserves; and other costs related to
the development of specific properties, such as the costs of applying
for the tax credit, conducting a market analysis, and insuring the
property during construction. 

Our earlier report noted, however, that differences in the types and
locations of properties can lead to substantial variations in their
costs.  In the report, we estimated that the average per-unit cost of
developing tax-credit-supported units placed in service from 1992
through 1994 was about $60,000; however, about 10 percent of the
units cost less than $20,000 to develop while about 10 percent cost
more than $100,000.  We noted that differences in the physical
characteristics of properties--including the costs of acquiring land
and existing buildings, the types of buildings constructed, the
geographic location, the size of the units, the amenities provided,
and the construction standards used--accounted for some of the
variation in development costs.  We estimated, for example, that the
average per-unit cost for newly constructed buildings was about
$68,000 and the average cost for substantially rehabilitated
buildings was approximately $48,000.  We further noted that other
physical characteristics--such as unusually high local construction
costs, local seismic standards, or requirements to address
environmental issues--contributed to the higher development costs of
some properties. 


   NONPROFIT ORGANIZATIONS
   DEVELOPED DIFFERENT TYPES OF
   UNITS
------------------------------------------------------------ Letter :3

Our analysis showed that certain characteristics increased the costs
of housing units for both types of developers and, in some cases,
nonprofit developers were substantially more likely to build units
with such characteristics than for-profit developers.  We did not
find any instances in which for-profit developers' units had higher
cost characteristics. 

We identified characteristics of tax credit properties in our
database that appeared likely to us to have an impact on for-profit
and nonprofit developers' costs.  The eight characteristics we
analyzed were whether the unit was (1) located in an urban, suburban,
or rural area; (2) located in a distressed or nondistressed Census
tract;\4 (3) eligible to receive additional tax credits;\5 (4) in a
garden style, town house, or high-rise building or in a mixed type of
development; (5) built to serve families, the elderly, or others
(i.e., persons with special needs); (6) newly constructed or
rehabilitated; (7) under 700 square feet, between 700 and 1,000
square feet, or over 1,000 square feet; and (8) located in the
Pacific, Mountain/West Central, East North Central, Southeast, or
Northeast region of the country. 

For each of these characteristics, we identified (1) its influence on
cost for all tax credit units and (2) the relative proportion of
for-profit and nonprofit developers' units with the characteristic. 
For example, we determined that garden style units cost less than
other types of units and that for-profit developers were more likely
to build garden style units than nonprofit developers.  Our findings
for each of these characteristics can be found in appendix I. 

To determine the extent to which the difference in the average cost
of units built by nonprofit and for-profit developers could be
explained by these eight characteristics, we performed a statistical
procedure called a regression analysis.\6 This analysis showed that
some or all of the difference could be explained by differences in
the eight characteristics.  Specifically, this analysis showed that,
after accounting for differences in what for-profit and nonprofit
developers built, the cost difference was $5,600 ï¿½ $7,200.\7 This
means that, if all other factors in building these housing units were
equal, the average per-unit cost for nonprofit developers was between
$1,600 lower and $12,700\8 higher than the average per-unit cost for
for-profit developers. 

The results of our regression analysis showed that the following
characteristics had a statistically significant relationship with the
per-unit cost:  (1) the property's location in a distressed Census
tract, (2) the property's eligibility for additional tax credits, (3)
the type of building (high-rise, garden style, town house, or
other/mixed), (4) the type of construction (new construction versus
rehabilitation), (5) the number of square feet per unit, and (6) the
region of the country.  Characteristics that we did not find to be
statistically significant were (1) the location of the property
(urban, suburban, or rural) and (2) the population primarily served
(the elderly versus families).  This analysis and its results are
explained in more detail in appendix I, and details are provided in
table I.1. 

Our analysis of the cost implications of the various characteristics
of properties built by for-profit and nonprofit developers explains
75 percent of the variation in the per-unit costs observed in our
data.  Had information on additional characteristics that may affect
per-unit costs, such as unusually high local construction costs or
stringent seismic standards, been available, we might have been able
to explain some or all of the remaining variation in unit costs. 
Also, with additional information to explain the remaining 25 percent
of the variation, our conclusion about the effect of nonprofit
developers might be different. 


--------------------
\4 The criteria we used to identify distressed Census tracts are
described on p.  11. 

\5 Properties developed in neighborhoods where development costs are
high relative to incomes are entitled to receive supplemental tax
credits. 

\6 We developed a statistical model, called a regression model, to
examine the factors associated with unit cost.  A regression model is
used to investigate the relationships among variables.  For this
study, we used the model to predict the amount of change to unit cost
that would accompany changes in other factors.  For example, we
predicted the amount of change to unit cost that would result from a
change in unit size, after accounting for other factors that
influence cost, such as unit location and type. 

\7 This difference was not statistically significant. 

\8 Because of rounding, the upper bound estimate is $12,700, not
$12,800. 


   AGENCY COMMENTS
------------------------------------------------------------ Letter :4

Because this report does not discuss any aspect of the Low-Income
Housing Tax Credit program's implementation or administration by the
federal government or the states, we did not solicit comments from
either the Department of the Treasury or the state-level tax credit
allocating agencies.  We did, however, seek the views of national
organizations representing both for-profit and nonprofit developers
on which characteristics we should include in our analysis and
incorporated their suggestions to the extent possible. 


   SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :5

To help understand why per-unit costs were higher for nonprofit
developers, we looked at the types of units they were building and
asked whether these types of units were generally more or less costly
to build than the types of units built by for-profit developers. 
First, we examined characteristics of housing units that we thought
might be related to per-unit costs and for which data were available
in our database.  These characteristics were location (i.e., urban,
suburban, or rural), the economic condition of the area, the
property's eligibility for additional tax credits, the type of
building, the type of tenants, the type of construction, the number
of square feet in the unit, and the region of the country.  Then we
performed a regression analysis to estimate the influence of the type
of developer on per-unit costs while controlling for these other
factors. 

Most of this analysis used data that we collected in 1996 from a
statistical sample of 423 properties placed in service between 1992
and 1994.  We also used certain 1990 Census data and a definition of
distressed Census tracts that considers local poverty and
unemployment rates.  A detailed description of our methodology
appears in appendix I. 

In describing the results of our analysis, we generally present the
upper and lower bound of the confidence interval around each point
estimate.  Where the confidence interval is not presented with the
estimate itself (see fig.  1, for example), we included this
information in table I.2 in appendix 1. 

We conducted our work from July to December 1998 in accordance with
generally accepted government auditing standards. 


---------------------------------------------------------- Letter :5.1

We are sending copies of this report to the appropriate congressional
committees.  We will make copies available to others on request. 

Please contact me at (202) 512-7631 if you or your staff have any
questions.  Major contributors to this report are listed in appendix
II. 

Sincerely yours,

Judy A.  England-Joseph
Director, Housing and Community
 Development Issues


COST IMPLICATIONS OF TAX CREDIT
PROPERTIES' CHARACTERISTICS
=========================================================== Appendix I

To understand why the average per-unit costs of properties developed
with low-income housing tax credits were higher for nonprofit
developers than for for-profit developers, we studied the data in our
database of tax credit units placed in service between 1992 and
1994.\1 Specifically, we looked at the following eight
characteristics that we believed could have had an impact on the cost
of developing these properties: 

  -- whether the property was in an urban, suburban, or rural
     location;

  -- whether the property was in an economically distressed area;

  -- whether the property was in a location that made it eligible for
     additional tax credits;

  -- whether the development was a high-rise building, garden style
     building, town house, or mixture of these unit types;

  -- whether the units were built for families in general, the
     elderly, or a special needs population, such as the mentally
     disabled or the recently homeless;

  -- whether the property being developed was all new construction or
     the rehabilitation of an existing property;

  -- the size of the units developed; and

  -- the region of the country. 

This analysis consisted of three steps:  First, we identified the
impact of each characteristic on the unit cost for all of the
properties by determining the relative cost of units in properties
with or without each characteristic.  Second, we determined the
degree to which for-profit or nonprofit developers were more or less
likely to develop properties with each characteristic.  Finally, we
performed a regression analysis, taking into account our sample
design, to estimate the influence of the type of developer on
per-unit costs while controlling for these other factors. 

The following describes the results of the first two analyses for
each of the eight characteristics.  For each statistical estimate, we
computed the upper and lower bounds of the 95-percent confidence
interval.  We also tested for the statistical significance of
differences between estimates.\2 Each graph contains several vertical
lines.  The top of each line, which we designate as our higher
estimate, represents the upper bound of the 95-percent confidence
interval, and the bottom of the line, which we designate as our lower
estimate, represents the lower bound of the 95-percent confidence
interval.  The circle near the center of the line shows our single
point estimate.  Note that graphs presenting information on the cost
implications of the different characteristics describe all units
(those produced by both types of developers) while the graphs
describing the portion of developers' units with various
characteristics have separate lines for nonprofit and for-profit
developers.  On the graphs that describe differences between
nonprofit and for-profit developers, shaded areas indicate that a
difference is statistically significant at the 95-percent confidence
level.  For the remaining graphs, the results of testing for
statistical significance are described in the figures' titles. 


--------------------
\1 Our database contains information from a probability sample of 423
properties.  We used this sample to represent our total study
universe of about 4,100 properties.  These 4,100 properties,
containing over 170,000 low-income units, were placed in service in
the 48 contiguous states and the District of Columbia from January 1,
1992, through December 31, 1994.  Our probability sample of 423
properties was drawn from two strata, a large property stratum and a
small property stratum.  The large property stratum consisted of 29
properties with more than 300 units in each property.  All 29 of
these properties were included in our sample.  The remaining
properties were in the small property stratum.  We selected 394
properties from this stratum into the sample with probabilities
proportionate to their size, as measured by their numbers of
low-income housing tax credit units. 

\2 These tests are more useful than relying on overlapping confidence
intervals to rule out statistically significant differences. 


      URBAN/SUBURBAN/RURAL
      LOCATION
------------------------------------------------------- Appendix I:0.1

For all tax credit units, urban and suburban units cost more than
rural units.  See figure I.1. 

   Figure I.1:  Urban and Suburban
   Units Cost More Than Rural
   Units

   (See figure in printed
   edition.)

Our analysis shows that nonprofit developers were more likely to
build units in urban areas--61 percent ï¿½ 10 percent compared with 44
percent ï¿½ 6 percent for for-profit developers.  At the same time,
for-profit developers were more likely to build in rural areas--33
percent ï¿½ 6 percent compared with 15 percent ï¿½ 6 percent for
nonprofit developers.  There was no significant difference in the
proportion of building done by for-profit developers and nonprofit
developers in suburban areas.  See figure I.2. 

   Figure I.2:  Nonprofit Units
   Were More Likely to Be in Urban
   Areas and Less Likely to Be in
   Rural Areas

   (See figure in printed
   edition.)


      DISTRESSED/NONDISTRESSED
      CENSUS TRACT
------------------------------------------------------- Appendix I:0.2

To determine the relationship between the economic health of an area
and the cost of developing low-income housing tax credit properties,
we used information that we created in 1998 for a study on the
designation of economically distressed areas as "renewal"
communities.\3 This study used 1990 Census data to identify Census
tracts in which (1) the poverty rate was at least 20 percent, (2) the
unemployment rate was 9.45 percent or higher, and (3) at least 70
percent of the households had incomes of less than 80 percent of the
local area's median income.\4

We were able to identify the Census tracts of most of the properties
in our database of tax credit properties.\5 For these, we determined
whether the tracts were classified as distressed according to the
criteria described above.  We then analyzed the cost implications of
building in distressed areas and the proportion of such units built
by for-profit and nonprofit developers. 

As figure I.3 shows, unit costs did not vary significantly with the
economic condition of the neighborhood. 

   Figure I.3:  Differences in
   Unit Costs, by the
   Neighborhood's Economic
   Condition, Were Not
   Statistically Significant

   (See figure in printed
   edition.)

Nonprofit developers were also more likely than for-profit developers
to build units in economically distressed areas.  See figure I.4. 

   Figure I.4:  Nonprofit
   Developers Were More Likely to
   Build in Distressed Areas

   (See figure in printed
   edition.)


--------------------
\3 We have issued one report and two testimonies on the American
Community Renewal Act of 1998--Community Development:  Identification
of Economically Distressed Areas (GAO/RCED-98-158R, May 12, 1998),
Community Development:  Information Related to H.R.  3865, the
American Community Renewal Act of 1998 (GAO/T-RCED-98-196, May 19,
1998), and Community Development:  The American Community Renewal Act
of 1998 (GAO/T-RCED-98-263, Aug.  19, 1998). 

\4 Although there are various ways of quantifying distress, we
selected these criteria because they were included in proposed
legislation related to renewal communities. 

\5 We were not able to determine the Census tracts for 15 percent of
the nonprofit units and 21 percent of the for-profit units in our
database and termed these "unknown" for the purposes of this
analysis. 


      ELIGIBILITY FOR ADDITIONAL
      CREDITS
------------------------------------------------------- Appendix I:0.3

Properties developed in neighborhoods where development costs are
high relative to incomes are entitled to receive supplemental tax
credits.  For all tax credit units, those eligible for additional
credits were more costly to develop than those that were not
eligible.  See figure I.5. 

   Figure I.5:  Units Eligible for
   Additional Credits Cost More
   Than Other Units

   (See figure in printed
   edition.)

We found that nonprofit developers were more likely than for-profit
developers to build units eligible for additional credits:  About 49
percent (ï¿½ 10 percent) of nonprofit units qualify, while only 21
percent (ï¿½ 5 percent) of the for-profit units qualify, as figure I.6
shows. 

   Figure I.6:  Nonprofit Units
   Were More Likely to Be Eligible
   for Additional Credits

   (See figure in printed
   edition.)


      TYPE OF BUILDING
------------------------------------------------------- Appendix I:0.4

For all tax credit units, as figure I.7 shows, the cost of building
was significantly higher for high-rise units than for units in town
house, garden style, or mixed developments.  At the same time, the
cost of building was lower for garden style units. 

   Figure I.7:  High-rise Units
   Cost More and Garden Style
   Units Cost Less Than Other
   Styles

   (See figure in printed
   edition.)

As figure I.8 shows, we did not find significant differences between
the proportion of high-rise or town house units built by for-profit
and nonprofit developers, but we did find statistically significant
differences in the proportion of garden style units and other/mixed
types of developments.  Specifically, for-profit developers were more
likely to develop garden style units (62 percent ï¿½ 6 percent) than
nonprofit developers (47 percent ï¿½ 10 percent), while nonprofit
developers were more likely to build mixed developments (22 percent ï¿½
8 percent) than for-profit developers (12 percent ï¿½ 3 percent). 

   Figure I.8:  Nonprofit
   Properties Were Less Likely to
   Have Garden Style Units and
   More Likely to Have Other or
   Mixed Units

   (See figure in printed
   edition.)


      TYPE OF TENANTS
------------------------------------------------------- Appendix I:0.5

For all units, the per-unit costs did not vary significantly with the
type of tenant served--families, the elderly, or other groups with
special needs.  See figure I.9. 

   Figure I.9:  Differences in
   Unit Costs, by the Property's
   Primary Use, Were Not
   Statistically Significant

   (See figure in printed
   edition.)

We did not find significant differences between nonprofit developers
and for-profit developers in the proportion of units they built to
serve either families or elderly tenants, but we found that nonprofit
developers were significantly more likely to build units intended to
serve other groups with special needs.  Specifically, we estimate
that 12 percent ï¿½ 6 percent of the units built by nonprofit
developers were targeted to serve tenants with special needs compared
with 4 percent ï¿½ 2 percent of the units built by for-profit
developers.  See figure I.10. 

   Figure I.10:  Nonprofit Units
   Were More Likely to Serve
   Groups With Special Needs

   (See figure in printed
   edition.)


      NEW
      CONSTRUCTION/REHABILITATION
------------------------------------------------------- Appendix I:0.6

For all units, we also found that new construction tended to cost
more than rehabilitation, as shown in figure I.11. 

   Figure I.11:  New Construction
   Cost More Than Rehabilitation

   (See figure in printed
   edition.)

As figure I.12 shows, we did not find a significant difference
between for-profit and nonprofit developers in the proportion of
units developed through rehabilitation and new construction. 

   Figure I.12:  Percentages of
   Rehabilitated and Newly
   Constructed Units Showed No
   Statistically Significant
   Differences

   (See figure in printed
   edition.)


      SIZE OF UNITS
------------------------------------------------------- Appendix I:0.7

For all units, we found that the cost to develop larger units was
greater than the cost to develop smaller units.  See figure I.13. 

   Figure I.13:  Large Units Cost
   More Than Other Units

   (See figure in printed
   edition.)

We found that nonprofit developers were less likely than for-profit
developers to build units of between 700 and 1,000 square feet. 
However, nonprofit developers were more likely to build units of over
1,000 square feet--47 percent ï¿½ 15 percent for nonprofit developers
compared with 24 percent ï¿½ 12 percent for for-profit developers. 
There was no significant difference in the proportion of units of
under 700 square feet built by either type of developer.  See figure
I.14. 

   Figure I.14:  Nonprofit
   Developers Were Less Likely to
   Build Medium-Sized Units and
   More Likely to Build Large
   Units

   (See figure in printed
   edition.)


      REGION
------------------------------------------------------- Appendix I:0.8

To determine the effect of regional differences, we combined some of
the Census Bureau's nine geographical regions into five regions, for
the purposes of this analysis.  See figure I.15. 

   Figure I.15:  Regions Used in
   GAO's Analysis of Tax Credit
   Unit Costs

   (See figure in printed
   edition.)

As figure I.16 shows, unit costs varied by geographical region. 
Costs were higher in the Northeast and Pacific regions than in the
other three.  Per-unit costs were lowest in the Mountain/West Central
region. 

   Figure I.16:  Development Costs
   Were Higher in the Pacific and
   Northeast Regions

   (See figure in printed
   edition.)

We also found that nonprofit and for-profit developers' activity
varied by region:  Nonprofit developers' units were more likely to be
in the Pacific and Northeast regions and less likely to be in the
Mountain/West Central and Southeast regions, as shown in figure I.17. 

   Figure I.17:  Nonprofit Units
   Were More Likely to Be Found in
   the Pacific and Northeast
   Regions

   (See figure in printed
   edition.)


      REGRESSION ANALYSIS
------------------------------------------------------- Appendix I:0.9

The above comparisons were between units in nonprofit and for-profit
developers' properties that were the same for only one of the eight
factors we examined.  Ideally, we would compare properties that were
the same for all factors thought to influence per-unit costs.  The
regression analysis that follows simultaneously considers the effects
on unit costs of nonprofit or for-profit development and of the eight
other factors we examined. 

To estimate how much of the $18,000 average per-unit cost difference
could be explained by these characteristics, we performed a
regression analysis.  For this analysis we combined the cost
implications for each of the eight characteristics we had already
examined to determine their collective implications for the costs of
units developed by for-profit and nonprofit builders.  This analysis,
which accounted for the differing proportions of higher- and
lower-cost characteristics associated with the units built by the two
types of developers, did not detect a statistically significant
difference, at the 95-percent confidence level, between the costs for
the two types of developers.  Specifically, it showed that the cost
difference between them was $5,600 ï¿½ $7,200.  In other words, if all
other factors had been equal, nonprofit developers' units could have
been expected to be from $1,600 cheaper to $12,700\6 more expensive
than for-profit developers' units. 

Our analysis explained about 75 percent of the variation in the
per-unit costs observed in our data.  If information on additional
characteristics affecting per-unit costs, such as unusually high
local construction costs or stringent seismic standards, had been
available, we might have been able to explain more of the variation
in unit costs.  With information to explain the remaining 25 percent
of the variation, our conclusion about the effect of nonprofit
development might have been different.  Table I.1 contains the
detailed results of our regression analysis. 



                               Table I.1
                
                Detailed Results of Regression Analysis
                   on Units' Total Development Costs

                                                  Standard
                                                 errors of
                                    Coefficien  coefficien
Independent variables and effects            t           t     P-value
----------------------------------  ----------  ----------  ----------
Intercept                            $52654.22     5693.63      0.0000
Sponsor type
 Nonprofit                             5562.79     3659.69      0.1293
 For-profit                               0.00        0.00
Geographic area
 Urban                                -2631.92     3547.61      0.4586
 Suburban                             -2524.09     3147.06      0.4230
 Rural                                    0.00        0.00           .
Construction type
 New only                             16646.92     3074.18      0.0000
 Rehabilitation only                      0.00        0.00           .
Building type
 High-rise only                       13583.53     4964.91      0.0065
 Other only and mixed                 -6507.51     3186.06      0.0417
 Walkup/garden only                       0.00        0.00           .
Primary use
 Elderly                               5039.75     6428.70      0.4335
 Family                                   0.00        0.00           .
Eligibility for additional
 credits                              16325.02     4211.37      0.0001
 Eligible                                 0.00        0.00           .
 Not Eligible
Economic Condition
 Distressed                           12883.65     4714.36      0.0065
 Unknown                               4908.41     2410.66      0.0424
 Not distressed                           0.00        0.00           .
Region
 Pacific                              11222.78     7217.23      0.1207
 Mountain/West Central                       -     5395.40      0.0000
 East North Central                   25229.98     4592.45      0.0000
 Southeast                                   -     4857.45      0.0000
 Northeast                            23841.43        0.00           .
                                             -
                                      20268.61
                                          0.00
Square feet above or below 900          101.33       19.41      0.0000
----------------------------------------------------------------------
Note:  The multiple R-Square for this model was 0.76480.  The total
number of apartment units represented in this analysis was 162,385. 

We also added two variables to the regression--(1) whether a Rural
Housing Service loan was obtained on the property and (2) the number
of units in the property.\7 Because neither of these characteristics
proved to be statistically significant, we excluded them from the
final model. 

Table I.2 provides details on the point estimates used elsewhere in
this report. 



                                        Table I.2
                         
                          Sampling Errors of Estimates From Low-
                            Income Housing Tax Credit Database

                                                               Confidence
                                                               interval--      Confidence
Unit characteristic              Estimate  Sampling error            from    interval--to
-------------------------  --------------  --------------  --------------  --------------
Average per-unit cost of           59,489           8,087          51,402          67,576
 all units
Average per-unit cost of           72,855           6,587          66,268          79,442
 nonprofit developers'
 units
Average per-unit cost of           55,015          10,640          44,376          65,655
 for-profit developers'
 units
Percentage of tax credit               22               6              16              28
 properties developed by
 nonprofit developers
Percentage of all units                10               3               7              13
 costing less than
 $20,000
Percentage of all units                10               4               6              14
 costing more than
 $100,000
Average per-unit cost of           67,246          12,789          54,457          80,035
 all newly constructed
 units
Average per-unit cost of           48,068           5,298          42,770          53,366
 all rehabilitation units
Average land cost for               4,019           1,279           2,739           5,298
 nonprofit developers
Average land cost for               2,855             564           2,290           3,419
 for-profit developers
Average acquisition cost            5,172           2,031           3,141           7,204
 for nonprofit developers
Average acquisition cost            4,424             892           3,352           5,315
 for for-profit
 developers
Average construction cost          42,177           4,276          37,901          46,453
 for nonprofit developers
Average construction cost          29,237           3,483          25,754          32,720
 for for-profit
 developers
Average general                    14,509           2,073          12,435          16,582
 development cost for
 nonprofit developers
Average general                    10,792           1,155           9,638          11,947
 development cost for
 for-profit developers
Average cost of reserves            2,292             861           1,432           3,153
 for nonprofit developers
Average cost of reserves              989             423             565           1,412
 for for-profit
 developers
Average of other costs              1,710             320           1,390           2,031
 for nonprofit developers
Average of other costs              3,641           5,401        -1,759\a           9,042
 for for-profit
 developers
-----------------------------------------------------------------------------------------
Note:  The 95-percent confidence level is used throughout this table. 
We found no statistically significant differences between nonprofit
and for-profit developers in the percentage of unit costs
attributable to (1) construction-related expenses; (2) general
development costs (including developers' fees, profit and overhead)
and various fees for professional services, such as accounting; and
(3) other expenses, including the cost of acquiring land or buildings
(where applicable); operating, replacement, and other prefunded
reserves; and other costs, such as the cost of retiring an existing
mortgage or the cost of a market analysis. 

\a Negative sign indicates that the estimate is unreliable. 


--------------------
\6 Because of rounding, the upper bound estimate is $12,700, not
$12,800. 

\7 The P-values for these variables were 0.63 and 0.64, respectively. 


MAJOR CONTRIBUTORS TO THIS REPORT
========================================================== Appendix II

RESOURCES, COMMUNITY, AND ECONOMIC
DEVELOPMENT DIVISION, WASHINGTON,
D.C. 

Stan Czerwinski, Associate Director
Karen Bracey, Assistant Director
Dennis Fricke, Assistant Director
David Lewis, Evaluator-in-Charge
Sara Ann Moessbauer, Referencer
Lynne Goldfarb, Graphics Adviser
Elizabeth R.  Eisenstadt, Communications Analyst


*** End of document. ***