Estimating Downtime from Data on

Residential Buildings after the Northridge

and Loma Prieta Earthquakes
Mary C. Comerio,a) M.EERI, and Howard E. Blecherb)

The performance-based earthquake engineering (PBEE) methodology

developed by the Pacific Earthquake Engineering Research (PEER) center
uses data from recent earthquakes to calibrate its loss models. This paper
describes a detailed review of building department permit data from the 1989
Loma Prieta earthquake and the 1994 Northridge earthquake. Although the
data is limited to wood-framed residential structures, it provides some insight
into the length of time between an event and re-occupancy. Based on a review
of approximately 4,900 records, the typical repair of damaged multifamily
residential buildings required two years and building replacement required
almost four years. When this data is supplemented with additional case studies
from other events, the capacity to better calibrate downtime models will
improve, particularly if construction-repair times are separated from estimates
of the time gap between closure and start-of-repair.
关DOI: 10.1193/1.3477993兴

INTRODUCTION: DEFINING DOWNTIME

When buildings or infrastructure are damaged in a natural or manmade disaster,
there is a period of time needed to inspect the facility, ascertain the degree of structural,
nonstructural, or hazardous materials damage, make decisions to close, repair, or re-
open the facility. Estimating the downtime between damage and re-occupancy is critical
in understanding the economic impacts of a disaster. In statistically-based loss estimat-
ing models such as HAZUS (NIBS/FEMA 1999), or in facility-specific models (Com-
erio 2000, Porter et al. 2001, Taghavi et al. 2002), the common approach to estimating
downtime is to use repair time for construction, derived from the structural and non-
structural damage probabilities for buildings.
The time needed to repair facility-specific damage is the traditional definition of
downtime, but experience in recent disasters suggests that the time to repair is only one
component of downtime. Often there is a large time gap between the damage and closure
of a facility and the beginning of construction for repairs. Krawinkler and Miranda
(2004) note that the basic difficulties in quantifying the length of downtime are the un-
certainties associated with the availability of labor, materials and capital following a di-
saster, and the difficulties of relating quantifiable damage with the need for repair and
the loss of function. Comerio (2006) defined downtime as having “rational” and “irra-

a)
Professor, Dept. of Architecture, University of California, Berkeley, CA 94720
b)
Former Graduate Research Assistant, Building Designer, Urban Designer, San Francisco CA 94110

951
Earthquake Spectra, Volume 26, No. 4, pages 951–965, November 2010; © 2010, Earthquake Engineering Research Institute
952 M. C. COMERIO AND H. E. BLECHER

tional” components. The “rational” components are those which are more predictable
and more easily quantifiable. These include construction costs and the time needed to
repair damaged facilities and refinish space for use. The “irrational” components are
situationally specific and difficult to quantify as they account for the time needed to mo-
bilize resources and make decisions. These include four issues which dominate the pre-
construction period: (1) the need to secure financing, (2) the ability to relocate building
functions, (3) the availability of manpower, and (4) economic and regulatory uncertain-
ties.
The amount of time added to the downtime equation by any of these components can
vary based on the unique and specific economic position of a building owner as well as
the operational needs of the tenants in a damaged facility. However, in general, down-
time can be defined as a function of the time needed to repair physical damage based on
damage conditions, plus a time increment derived from the presence of any of the “ir-
rational” factors based on the scale of the disaster. Comerio (2006) published a case
study of downtime at Stanford University after the Loma Prieta earthquake which sup-
ports this definition.

DATA SOURCES
The predominant downtime data available in the United States is on earthquake-
related damage to wood-framed residential buildings. The Association of Bay Area Gov-
ernments conducted a preliminary empirical study of downtime as a measure of time to
occupancy for residential buildings as part of a housing needs assessment (ABAG
2000). In the research presented here, the ABAG study data was expanded by adding
information from additional sources to refine the understanding of each building’s post-
earthquake status—whether it was repaired, demolished or rebuilt. Once these were es-
tablished, a statistical analysis of building characteristics and downtime information was
made from the expanded data set for most Northridge-affected areas and several Loma
Prieta affected cities.
The data used in the PEER study is a combination of information acquired from lo-
cal and county governments (particularly building departments) by ABAG subsequent to
the Loma Prieta and Northridge earthquakes, census and county Assessor’s data as of
2006 for the same properties. Combining information from multiple data sets was es-
sential to improving the accuracy of the data for each building. ABAG provided the base
data for the following Northridge-affected areas: the City of Los Angeles, unincorpo-
rated Los Angeles County and the City of Santa Monica. Additionally, data was provided
for the following Loma Prieta-affected areas: the City and County of San Francisco, the
City of Hollister, and the City of Watsonville. Notably absent from the Loma Prieta data
is information on buildings in Oakland and Santa Cruz which both suffered severe dam-
age. Unfortunately no data was available for these locations.
The ABAG data was partially verified through interviews with building departments
and site inspections of approximately 10% of the data set. The building description in-
formation from the municipalities was either based on permit application data or was
assembled from the field inspectors’ reports, evaluations and comments written subse-
ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 953

quent to the earthquake. Unfortunately, largely due to the chaos subsequent to these
earthquakes there are a variety of issues with this data. One example is the loss of much
critical data collected after the Loma Prieta earthquake by the San Francisco Building
Department due to a computer problem. In general, permit data (which indicates the
type and timing of construction or demolition work) was often incomplete for three main
reasons:
1. Many owners (mostly of single-family residences) did not apply for the required
permits to repair or demolish therefore, there is inconclusive information about
the damage status and timing of work for these properties.
2. Dates reflecting work completion or date of demolition indicate that repair work
or demolition was done but may not accurately represent the timing of the work.
3. The quality control over building information varies by municipality.
The data from the field inspectors sent to evaluate buildings after the earthquakes
includes tag color information, damage assessments and other notations and comments.
These field assessments (such as damage percentage and repair cost estimates) are in-
consistent due to varied levels of experience, areas of expertise and human error among
inspectors. Furthermore, inspectors recorded similar information in different ways.
In each of the earthquake-impacted areas for which data was available, at least 1% of
the housing stock had been rendered uninhabitable. The “uninhabitable units” as de-
scribed by Perkins and Kramer (2000) are those tagged yellow or red per the post-
earthquake inspection guidelines outlined in the ATC 20: Procedures for Post-
Earthquake Safety Evaluation of Buildings series of documents (ATC 20 1989, ATC
20-2 1995, ATC 20-3 1996, Mitrani-Reiser 2007). According to these guidelines, field
inspectors indicate the building evaluation outcome with a red (unsafe for occupancy or
entry), yellow (limited entry) or green (apparently safe) placard. Green-tagged buildings
are not necessarily undamaged but were excluded from this study as the earthquake-
related information for those buildings is not tracked or maintained by authorities com-
prehensively.
The Assessor data was purchased from Metroscan, a data vendor, which matched the
addresses in the ABAG data to the relevant assessor’s database. This information, current
as of 2006, was used to confirm the ABAG data and to assist in determining the status of
property (i.e., repaired, demolished and/or rebuilt) subsequent to the earthquake by add-
ing more information to the ABAG data. The primary issue with this set of information
was missing data due to both a failure to match properties with the ABAG data and in-
complete assessor records. Despite these common building data problems, there was suf-
ficient information available to analyze the data set.

BUILDING STATUS: REPAIRED, DEMOLISHED, OR REBUILT

Informed decisions about the fates of buildings (repaired, demolished and/or rebuilt)
damaged by the Northridge and Loma Prieta earthquakes were made using a combina-
tion of ABAG data and Assessor data supplemented with anecdotal knowledge of local
conditions. The application of data in concert with independent sources results in stron-
ger conclusions than possible if these sources were used separately. In particular, the per-
954 M. C. COMERIO AND H. E. BLECHER

mitting information and field inspection data were the most useful pieces of information
from the ABAG data. The Assessor data included updated (2006) building characteristics
(such as the construction year) and financial information such as the ratio of improve-
ments to land value (the multiplier to land value which results in the value of built im-
provements to the property) and the most recent sales date and price. This financial data
is the most reliable information available from the Assessor. Individuals with extensive
inspection experience provided anecdotal knowledge based on familiarity with the char-
acteristics of the buildings, the general degree of damage sustained in the different lo-
cations and the general approach of local authorities.
The assignment to a final-outcome status of repaired, demolished and/or demolished
and rebuilt can be divided into two categories based on the quantity and quality of the
available data. The great majority (84%) of the properties were assigned a status based
on evidence available in the data. Of these, 81% were based on specific property data
and 3% which had incomplete data were assigned a status based comparable character-
istics in similar properties. The remaining 16% of the buildings were assigned a status
using the available data supplemented with anecdotal information when the available
data was inconclusive. The definitions used in assigning post-event status categories are
as follows:
Repaired. The year-built is prior to the earthquake date, there is a “work comple-
tion” date, and there is financial information from the Assessor data indicating that there
was a structure of value on the property as of 2006.
Demolished. The year-built is prior to the earthquake date, there is a “demolished
date” indicating that the owner applied for a demolition permit, and there is no financial
information in the Assessor data.
Demolished and rebuilt. The year-built is subsequent to the earthquake date and
within three years of the earthquake, and the ratio of improvements in 2006 is non-zero,
indicating that there was a structure of value on the property at that time.
Demolished and rebuilt (not necessarily due to the earthquake). The year-built is
subsequent to the earthquake date but more than three years after the event, making it
less likely that the current structure replaced one demolished because of the earthquake.
However, the ratio of improvements in 2006 is non-zero, indicating that there was a
structure of value on the property at that time.
Through this assessment, it became clear that the great majority of red- and yellow-
tagged buildings (92% in Loma Prieta and 86% in Northridge) were repaired, while only
5.4% in Loma Prieta and 10.4% in Northridge were demolished and rebuilt. A summary
of these results is in Table 1.

DATA DEFINITIONS AND ANALYSIS

Once the buildings were assigned a status (i.e., repaired, demolished and/or rebuilt)
the timing and occupancy type information was summarized as functions of tag color
(red or yellow), relevant earthquake and location. The building type incorporates con-
struction type (e.g., wood-framed, unreinforced masonry, etc.), number of stories, the
ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 955

Table 1. Estimated status of red- and yellow-tagged residential buildings damaged by the Loma
Prieta and Northridge earthquakes

Number of Buildings
Repaired Demolished Demo/Rebuilt Total

Loma Prieta, 17 October 1989

San Francisco
Red-Tagged 178 10 28 216
Yellow-Tagged 738 2 13 753

Hollister 84 0 5 89
Watsonville 212 11 25 248
Loma Prieta Total 1212 23 71 1306
Northridge, 17 January 1994
Los Angeles
Red-Tagged 1044 41 215 1300
Yellow-Tagged 1703 55 97 1855

Los Angeles County

(Unincorporated)
Red-Tagged 187 21 52 260
Yellow-Tagged 126 0 0 126

Santa Monica
Red-Tagged 13 6 8 27
Yellow-Tagged 54 3 6 63
Northridge Total 3127 126 378 3631
Total 4339 149 449 4937

construction era (before or after 1940), and whether the structure is single-family or
multifamily. This was the most consistent and complete definition of “building type”
available across the different sets of data by location. Single-family is defined as one
dwelling unit and multifamily is defined as two or more dwelling units. The number of
dwelling units is provided by the local building department or city authorities. The time
to repair is the number of months between the earthquake and the “work completion”
date provided by the local authorities for those buildings determined to have been re-
paired. Time to rebuild is the number of months between the earthquake and the “work
completion” date provided by local authorities or July 1 of the year built from the As-
sessor data (assumed for those properties without a “work completion” date) for those
buildings determined to have been demolished and subsequently rebuilt. See Table 2 for
a summary of results.
While the building characteristics were fairly complete, there were problems with the
time-to-occupancy information. In fact, while 67% of the rebuilt properties have timing
956 M. C. COMERIO AND H. E. BLECHER

Table 2. Number of buildings by occupancy type and mean months to occupancy

Number of
Buildings by
Occupancy Type Mean Months to Occupancy
Single- Repaired Rebuilt All
Family Multifamily Buildings Buildings Buildings

Loma Prieta, 17 October 1989

San Francisco
Red-Tagged 43 172 7 46 13
b b b
Yellow-Tagged 194 559

Hollister 89 0 19 29 20
Watsonville 129 119 10 22 12

Northridge, 17 January 1994

Los Angeles
Red-Tagged 781 519 27 48 31
Yellow-Tagged 0 1855 25 30 25

Los Angeles County

(Unincorporated)
Red-Tagged 181 79 20 30 21
a
Yellow-Tagged 0 126 13 13

Santa Monica
Red-Tagged 12 15 26 34 29
Yellow-Tagged 0 63 24 47 26
a
There are no rebuilt properties.
b
There is limited timing information for the San Francisco Yellow-tagged buildings.

information, only 47% of the repaired properties have timing information. In general,
the red-tagged population had a higher percentage of properties with timing information.
This is possibly because the higher levels of damage sustained by these properties re-
quired the local authorities to be more vigilant over the repair and rebuilding processes.
Furthermore, the lighter damage sustained by yellow-tagged properties may have com-
pelled building owners to proceed with repairs without applying for work permits (there-
fore, building departments would have limited or no documentation of the timing of the
work). Three likely reasons for the low percentage of buildings with timing information
are: (1) recording errors by government authorities; (2) owners who did not properly ap-
ply for work permits; and (3) bias in assigning a “repaired” status to buildings that may
not have been repaired.
The two categories which contribute most highly to the low percentages of available
timing data are the yellow-tagged Los Angeles buildings and the yellow-tagged San
Francisco buildings. These Los Angeles buildings (representing 38% of the total popu-
lation) only had timing information for 49% of the repaired or rebuilt buildings while
ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 957

these San Francisco buildings (representing 15% of the total population) only had timing
information for 2% of the repaired or rebuilt buildings (and these were only for rebuilt
buildings). If these two categories are excluded, 65% of the repaired and rebuilt proper-
ties of the other locations had information that suggest a time to occupancy. The major-
ity of these (54%) are the repaired properties. In this breakdown 73% of the rebuilt prop-
erties have timing information while 64% of the repaired properties have timing
information.

OVERALL FINDINGS
The population of buildings in the data set affected by both the earthquakes includes
4,937 residential buildings, of which 3,631 (74%) were damaged by the Northridge
earthquake. As this is almost three times the building population damaged by Loma Pri-
eta, the Northridge results heavily color the overall statistics. The vast majority (88%) of
the buildings is assumed to have been repaired and this proportion is consistent across
several arrangements of the data, such as by earthquake, tag color or location. Addition-
ally, 9% of the buildings were demolished and rebuilt and the remainder (3%) were de-
molished but not rebuilt (See Table 1). The methodology for determining a building’s
post-earthquake status (described earlier) may have been somewhat biased toward repair,
but it follows logically that most owners would prefer to repair than invest in demolition
and rebuilding, particularly for the low-rise wood-framed structures that dominate the
population of buildings studied here, where the majority of buildings (71%) are one- to
three-stories, wood-framed, multifamily. The overall mean time to occupancy was
21 months 共1.8 years兲, the mean time to occupancy for repaired buildings was 19 months
共1.6 years兲 and the mean time to occupancy for demolished and rebuilt buildings was
39 months 共3.3 years兲.

STATISTICS BY TAG COLOR

The red-tags are 43% of the total population, and yellow-tags are 57% of the total
population. As would be expected, the mean times to occupancy for the relatively more
heavily damaged red-tagged buildings exceeded those of the yellow-tagged buildings.
The overall mean time to occupancy of 26 months 共2.1 years兲, the mean time to occu-
pancy for repaired buildings of 22 months 共1.8 years兲 and for rebuilt buildings of
43 months 共3.6 years兲 were all about 1.5 times the corresponding values for yellow-tagged
buildings. For both red-tagged and yellow-tagged populations the proportion of single-family
and multifamily buildings mimics the proportion of these residential types for each earth-
quake. Almost all (93%) of the yellow-tagged buildings of the total population are multifam-
ily and all of the yellow-tagged Northridge properties are multifamily. Similarly, the more
even division between single-family (58%) and multifamily (42%) red-tagged buildings re-
flects similar proportions for the red-tagged properties for both earthquakes. Finally, as with
the overall population, one- to three-story, wood-framed buildings constructed after 1940
dominate both the red-tag and yellow-tag populations. (see Figures 1 and 2.) Other charac-
teristics, such as building size (square feet), number of units, height, and other factors, were
evaluated, but none of these showed any impact on the time to re-occupancy.
958 M. C. COMERIO AND H. E. BLECHER

LA Red Tag: Mean Months to Repair/Rebuild by Building Type (ABAG)

60 300

250

Absolute Number of Buildings

48

200
36
Months

150

24
100

12
50

0 0
Unknown Mobile Unreinforced Non-Wood Non-Wood Non-Wood 8 Non-Wood 8 Wood 4-7 Wood 4-7 Wood 1-3 Wood 1-3 Wood 1-3 Wood 1-3 Other Varies
Homes 1-3 Masonry 4-7 stories 4-7 stories and up Pre and up Post stories Pre stories Post stories Pre stories Post stories Pre stories Post Varies Varies
stories Post Varies Pre Pre 1940 Post 1940 1940 Multi- 1940 Multi- 1940 Multi- 1940 Multi- 1940 Multi- 1940 Multi- 1940 Single- 1940 Single-
1940 Single- 1940 Multi- Multi-family Multi-family family family family family family family Family Family
Family family

Repaired Demolished-Rebuilt Demolished-Rebuilt (not due to earthquake) All Total Number of Buildings

Figure 1. Example of mean months to repair by building type for red-tagged buildings in Los
Angeles.

LA Red Tag: Mean Months to Repair/Rebuild by Year Built

(Building Department)
84 250

Absolute Number of Buildings

72
200
60

150
Months

48

36
100

24
50
12

0 0
Unknown 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 1994
<= Year < <= Year < <= Year < <= Year < <= Year < <= Year < <= Year < <= Year < <= Year < <= Year < <= Year <
1910 1920 1930 1940 1950 1960 1970 1980 1990 1994 2000

Year Built (Building Department)

Repaired Demolished-Rebuilt
Demolished-Rebuilt (not due to earthquake) All
Total Number of Buildings

Figure 2. Example of mean months to repair or rebuild by year built for red-tagged buildings
in Los Angeles.
ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 959

STATISTICS BY EARTHQUAKE
The mean time to occupancy for buildings affected by the Northridge earthquake at
27 months 共2.2 years兲 was more than 2.5 times that of Loma Prieta buildings. Most of this
difference is due to the mean time to occupancy for repaired buildings, which at 25 months
共2.1 years兲 is almost 6 times that for Loma Prieta. Furthermore, for both earthquakes the
vast majority of buildings were repaired rather than demolished or rebuilt. There are several
possible reasons for this difference. There are a higher percentage of multifamily buildings in
the Northridge population which are presumably larger and more complex than single-family
homes to repair and rebuild. Also, there are a higher percentage of demolished and rebuilt
buildings (11%) in the Northridge population than the Loma Prieta population (6%) and the
generally longer times to occupancy for those buildings that are rebuilt increases the overall
mean time to occupancy. Finally, the longer times to occupancy for the Northridge-affected
buildings reflect “[t]he recession that had gripped Southern California since the late 1980s
[which] had produced a soft rental housing market…Citywide, the multifamily vacancy rate
was 8.9 percent…In the areas where the greatest concentrations of damage occurred, there
was a surplus of vacant apartments” (Comerio 1998).

DOMINANT DATA FOR NORTHRIDGE RESULTS

At 26% and 38% of the total population of buildings across both earthquakes, the
Los Angeles red-tagged and yellow-tagged data sets dominate this analysis. As such, the
statistics for these areas are consistent with the characteristics of the overall population
as described earlier. Although the relative amount of single-family and multifamily
buildings is fairly consistent with the other locations, these two sets of data tend to have
longer times to occupancy than the others (see Figures 3 and 4.)

DOMINANT DATA FOR LOMA PRIETA RESULTS

Interestingly, the San Francisco yellow-tagged data set dominates the Loma Prieta
earthquake locations with 58% of the tagged residential buildings damaged by this
earthquake. However, among the red-tagged populations (which have more significant
damage than the yellow-tagged), the San Francisco buildings and the Watsonville build-
ings have almost equal presence with 17% and 19% of the Loma Prieta population re-
spectively. While the overall time to occupancy for these locations is similar at about
1 year, the repaired San Francisco red-tagged buildings had a mean time to occupancy of
7 months 共0.6 years兲 which is a little more than half that of the corresponding Watsonville
properties while the rebuilt San Francisco red-tagged buildings took 46 months
共3.8 years兲—more than twice that for the corresponding Watsonville properties. There are a
few possible explanations for this difference given that similar proportions of buildings were
repaired and rebuilt for these locations (see Figures 5 and 6.)
First, the San Francisco housing stock consists largely of Victorian homes in older
areas of the city (such as Cow Hollow, the Western Addition, Haight Ashbury, etc.) and
multistory wood-framed buildings in more recently-settled areas (such as the Marina
district where residential lots were created on fill after the 1915 Panama-Pacific Exhibi-
tion). While the Victorians are constructed with an occupiable ground floor the newer
wood-framed buildings often have a soft story. In this earthquake, the Victorians were
960 M. C. COMERIO AND H. E. BLECHER

Northridge: Mean Months to Repair

84 900

800

Absolute Number of Buildings

72
700
60
600
Months

48 500

36 400

300
24
200
12
100

0 0
City of LA Red City of LA Yellow Unincorporated LA Unincorporated LA Santa Monica Red Santa Monica Yellow
County Red County Yellow

All Repaired Single Family Repaired Multi Family Repaired Total Number of Buildings

Figure 3. Northridge earthquake residential buildings: Mean months to repair.

Northridge: Mean Months to Rebuild

84 180

160 Absolute Number of Buildings

72
140
60
120
Months

48 100

36 80

60
24
40
12
20

0 0
City of LA Red City of LA Yellow Unincorporated LA Unincorporated LA Santa Monica Red Santa Monica Yellow
County Red County Yellow

All Demo-Rebuilt Single Family Demo-Rebuilt

Multi Family Demo-Rebuilt All Demo-Rebuilt not EQ
Single Family Demo-Rebuilt not EQ Multi Family Demo-Rebuilt not EQ
Total Number of Buildings

Figure 4. Northridge earthquake residential buildings: Mean months to rebuild.

ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 961

Loma Prieta: Mean Months to Repair

84 180

160
72

Absolute Number of Buildings

140
60
120

48
Months

100

36 80

60
24
40
12
20

0 0
City of San Francisco Red City of San Francisco Yellow Hollister Watsonville

All Repaired Single Family Repaired Multi Family Repaired Total Number of Buildings

Figure 5. Loma Prieta earthquake residential buildings: Mean months to repair.

Loma Prieta: Mean Months to Rebuild

84 30
Absolute Number of Buildings
72
25

60
20
Months

48
15
36

10
24

5
12

0 0
City of San Francisco Red City of San Francisco Yellow Hollister Watsonville

All Demo-Rebuilt Single Family Demo-Rebuilt

Multi Family Demo-Rebuilt All Demo-Rebuilt not EQ
Single Family Demo-Rebuilt not EQ Multi Family Demo-Rebuilt not EQ
Total Number of Buildings

Figure 6. Loma Prieta earthquake residential buildings: Mean months to rebuild.

962 M. C. COMERIO AND H. E. BLECHER

likely to suffer relatively less damage while the soft-story buildings were more likely to
suffer significant structural damage. It is possible that the short repair times were for
buildings like the Victorians while the longer rebuild times were for the soft-story build-
ings rendered fundamentally unsafe by the earthquake. In contrast, the housing stock in
Watsonville was dominated by small-scale buildings (Comerio 1992) and so the times to
occupancy for this location likely represent work done on almost all single-story wood-
framed buildings (see discussion below).
Another significant difference between these locations is the breakdown of single-
family versus multifamily. While only 20% of the San Francisco red-tagged buildings
are single-family, the Watsonville red-tagged population is almost evenly divided be-
tween single-family and multifamily buildings. Furthermore, more than two times the
amount of multifamily red-tagged San Francisco buildings (as a percentage of rebuilt
buildings) were rebuilt versus the corresponding Watsonville buildings. On the other
hand, more than three times as many single-family red-tagged Watsonville buildings (as
a percentage of repaired buildings) were repaired versus the corresponding San Fran-
cisco buildings. This indicates (as would be expected) that rebuilding times for multi-
family buildings are significantly longer than for single-family buildings.
Finally, the much shorter repair times in San Francisco for buildings of a similar con-
struction type to those in Watsonville may reflect the difference in ground motions ex-
perienced by San Francisco buildings which were more distant from the earthquake epi-
center than the Watsonville buildings.

LOCATION HIGHLIGHTS
The population of yellow-tagged buildings in unincorporated Los Angeles County
had the shortest overall mean time to occupancy of any of the other Northridge-affected
populations at 13 months 共1.1 years兲. The major reason for this low value is that all of the
buildings in the population were repaired (this is unique among all the data populations—for
both Northridge- and Loma Prieta-affected areas). In fact, the time to occupancy for these
buildings is less than half of both the mean time to occupancy for all Northridge-affected
buildings and the mean time to occupancy for only yellow-tagged Northridge-affected build-
ings. Furthermore, even considering only repaired buildings, the mean time to occupancy for
this population is also less than half of both the mean time to occupancy for all Northridge-
affected buildings and the mean time to occupancy for only yellow-tagged Northridge-
affected buildings. The explanation for the short times to occupancy for this location is the
distance from the epicenter in the San Fernando Valley whereby these buildings may have
suffered less damage than those in other areas. Given that this population consists solely of
multifamily buildings with a likely relatively low level of damage the short time to repair
may be because owners were motivated to complete repairs quickly on their income-
producing multifamily properties.
The overall mean time to occupancy for the Hollister properties of 20 months
共1.6 years兲 significantly exceeds the overall mean time to occupancy of 14 months
共1.1 years兲 for the properties affected by the Loma Prieta earthquake. This lengthy mean
time to occupancy reflects the heavy damage likely sustained by Hollister buildings due to
ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 963

their relative proximity to the epicenter of the Loma Prieta earthquake. Furthermore, Hollis-
ter has the highest percentage of repaired properties (94%) of any population of red-tagged
buildings by locality. This is consistent with a characteristic feature of this location that all of
the tagged buildings were single-family residences and therefore more likely to be repaired
than demolished due both to a simpler construction type and to a more immediate sense of
ownership.
The overall mean time to occupancy of 12 months 共1.0 year兲 for the Watsonville
properties was the shortest for any location in the data set for either earthquake. This rela-
tively rapid time to occupancy reflected the lenient approach by Watsonville authorities in
granting work permits and allocating emergency funds to expedite the reconstruction of their
city. The city’s building department strategy was to encourage repairs with easy permits,
followed-up by strict inspections (Comerio 1992). This was a reasonable strategy as the hous-
ing stock in Watsonville was dominated by small-scale buildings. Although the data suggests
that the properties were evenly divided between single-family and multifamily structures, al-
most half of the multifamily structures only had two units, typically these were single-family
homes with garages converted to living quarters.

CONCLUSION
The downtime data from the Northridge and Loma Prieta earthquakes presented here
focuses on time to occupancy measures by tag color, by building type and by single-
family versus multifamily. These categories were used as they are the most relevant for
residential buildings, they have consistent and recognized definitions among interested
parties (such as building authorities, construction professionals, owners and academics)
and the data was available. In addition to the building-specific factors which affect time
to occupancy mentioned in the introduction, the amount of structural damage, earth-
quake intensity, site-specific ground motions, economic conditions, and geographic lo-
cation also affect the time to occupancy for a building. Finally, the mean times to occu-
pancy for repaired buildings of almost 2 years and for demolished and rebuilt buildings of
almost 3.5 years for a population consisting almost entirely of low-rise, wood-framed struc-
tures has significant implications for the post-earthquake strategies of governments, builders
and owners. The time to occupancy for larger and more complex buildings could easily ex-
ceed these figures.
Modeling downtime requires a formula that combines three critical elements: (1) an
estimate of construction repair time for individual facilities damaged (and rendered un-
inhabitable) by a disaster; (2) an estimate of the mobilization time needed for various
building stocks; and (3) a representation of the economic conditions in the region at the
time of the event. Together, these inputs adequately represent the relationship between
prevailing economic conditions and the rate of building stock recovery. Although limited
to wood-framed structures, the data from the Loma Prieta and Northridge earthquakes
provides a detailed insight into the various issues, such as building size, degree of dam-
age and local conditions, which affect the time needed to repair and or rebuild buildings
after disasters. Estimating downtime for other building types will require construction
964 M. C. COMERIO AND H. E. BLECHER

time projections based on damage estimates combined with factors for mobilization time
and economic conditions to adequately represent the true time to repair or rebuild after
a disaster.

ACKNOWLEDGMENTS
Jeanne Perkins of the Association of Bay Area Governments (ABAG) provided ac-
cess to the data which they collected from local building departments. The work sum-
marized herein was supported by the Pacific Earthquake Engineering Research (PEER)
Center through the EERC Program of the National Science Foundation under Award
number EEC-9701568. The authors acknowledge the many contributions of PEER re-
searchers in developing the concept of performance based earthquake engineering in
general, and the concept of downtime as a decision variable in particular. Any opinions,
findings, and conclusions or recommendations expressed in this material are those of the
author and do not necessarily reflect those of the National Science Foundation or other
sponsors.

REFERENCES
Association of Bay Area Governments (ABAG), 2000. Preventing the Nightmare—Post-
Earthquake Housing Issue Papers, Oakland, CA.
Applied Technology Council (ATC 20), 1989. ATC 20: Procedures for Post-Earthquake Safety
Evaluation of Buildings, Redwood City, CA.
Applied Technology Council (ATC 20-2), 1995. ATC 20-2: Addendum to the ATC-20 Post-
earthquake Building Safety Evaluation Procedures, Redwood City, CA.
Applied Technology Council (ATC 20-3), 1996. ATC 20-3: Case Studies in Rapid Post-
earthquake Safety Evaluation of Buildings, Redwood City, CA.
Comerio, M. C., 2006. Estimating downtime in loss modeling, Earthquake Spectra 22, 349–
365.
Comerio, M. C., 1998. Disaster Hits Home: New Policy for Urban Housing Recovery, Univer-
sity of California Press, Berkeley, CA, 326 pp.
Comerio, M. C., 1992. Hazards mitigation and housing recovery: Watsonville and San Fran-
cisco one year later, in Disasters and the Small Dwelling (Yasemin Aysan and Ian Davis,
eds.), James and James, Science Publishers, London, 127–135.
Comerio, M. C., with Hamilton, Rabinovitz & Alschuler, Inc., 1996. The Impact of Housing
Losses in the Northridge Earthquake: Recovery and Reconstruction Issues, Publication num-
ber: CEDR-14-96, Appendices publication number: CEDR-15–96. Center for Environmental
Design Research, University of California, Berkeley, CA.
Comerio, M. C., 2000. The Economic Benefits of a Disaster Resistant University: Earthquake
Loss Estimation for U. C. Berkeley, IURD Working Paper 2000–02, Institute of Urban and
Regional Development, University of California, Berkeley, CA.
Comerio, M. C., and Stallmeyer, J. C., 2001. Nonstructural Loss Estimation: The UC Berkeley
Case Study, Pacific Earthquake Engineering Research Center, University of California, Ber-
keley, CA.
Deierlein, G. G., 2004. Overview of a comprehensive framework for earthquake performance
ESTIMATING DOWNTIME FROM DATA ON RESIDENTIAL BUILDINGS 965

assessment, in Performance Based Seismic Design Concepts and Implementation (P. Fajfar
and H. Krawinkler, eds.), PEER Report 2004/05, University of California, Berkeley, CA.
EQE International and The Geographic Information Systems Group of the Governor’s Office of
Emergency Services (EQE), 1995. The Northridge Earthquake of January 17, 1994: Pre-
liminary Report of Data Collection and Analysis. Part A: Damage and Inventory Data. The
Governor’s Office of Emergency Services of the State of California, Sacramento, CA.
EQE International and The Geographic Information Systems Group of the Governor’s Office of
Emergency Services (EQE), 1997. The Northridge Earthquake of January 17, 1994: Report
of Data Collection and Analysis. Part B: Analysis and Trends. The Governor’s Office of
Emergency Services of the State of California, Sacramento, CA.
Krawinkler, H., and Miranda, E., 2004. Performance-based earthquake engineering, in Earth-
quake Engineering: From Engineering Seismology to Performance-Based Engineering, CRC
Press, Boca Raton, FL, 1152 pp.
Mitrani-Reiser, J., 2007. An Ounce of Prevention: Probabilistic Loss Estimation for
Performance-Based Earthquake Engineering, Ph. D. Thesis, California Institute of Technol-
ogy, Pasadena, CA.
National Institute of Building Sciences and Federal Emergency Management Agency (NIBS/
FEMA), 1999. Earthquake Loss Estimation Methodology HAZUS99: Software and Technical
Manual, Washington, D. C.
Perkins, J. B., and Kramer, C., 2000. Post-earthquake housing issue paper B: Time needed to
repair or replace uninhabitable housing following the Loma Prieta and Northridge earth-
quakes, in Preventing the Nightmare—Post-Earthquake Housing Issue Papers, Association
of Bay Area Governments, Oakland, CA.
Porter, K. A., Kiremidjian, A. S., and LeGrue, J. S., 2001. Assembly-based vulnerability of
buildings and its use in performance evaluation, Earthquake Spectra 17, 291–312.
Taghavi, S., and Miranda, E., 2002. Seismic performance and loss assessment of nonstructural
building components, in Proceedings: 7th U. S. National Conference on Earthquake Engi-
neering, CD ROM Paper #231, Boston MA, Earthquake Engineering Research Institute.
(Received 10 November 2009; accepted 1 April 2010兲