The IPD and Hillier Parker data was used in two main ways to calculate annual rates of depreciation for particular property types (this is discussed in more detail in section 3.4 below).
Firstly, Ordinary Least Squares (OLS) Regression was used to determine an estimated depreciation rate (EDR). This is the ‘best fit’ depreciation rate (as a percentage per annum; see sections 3.0 and 4.7 of Appendix A) obtained by pooling data within particular sectors according to their characteristics, which comprised:
• prime/non-prime;
• construction period (ie. Post-1975, 1966-74, Pre-1965), and • town type/geographical location.
• Level 1: prime, town type and construction period all included; • Level 2: prime and town type disregarding construction period; • Level 3: Towntype, disregarding prime and construction period • Level 4: Prime and construction period, disregarding town type • Level 5: Prime disregarding construction period and town type.
In addition, overall EDRs for each sector without the three variables were also calculated. Another way of viewing this is that disregarding variables ‘averages’ their effect. For example, Level 3 produces EDRs by towntype ‘averaged over prime and construction period.
This provided a powerful tool to explore the data and allow ‘slicing’ into relevant time periods or market states(ie. the end of the late 70s, early 80s downturn in 1980-83, the ‘boom’ of 1984-89, and the ‘slump’ of 1990-95). The choice of market state was influenced by the shape of the data and the need for a consistent comparative basis for the analysis. The rationale for the choice of sub-periods and the statistical methodology employed is discussed in more detail in Appendix A. It should be noted that in tables throughout the report a minus sign for a rate indicates depreciation (eg -2.5%) and no sign, appreciation (eg. 1.5%).
Secondly, an average depreciation rate (ADR) may also be calculated for a single property or a group of properties using a geometric mean (or compounding formula) to compare an ERV:HP ratio at the start of the time series with the same measure at the end of the series. Again, an annual depreciation rate in percentage terms is produced, but the method, in contrast to OLS regression, ignores movement in depreciation in the intervening years.
An ADR was calculated when data permitted. This is effectively a ‘relative’ measure of depreciation, which is consistent with the same ratio logarithmically transformed) used in the OLS regression. It should be distinguished from the alternative ‘absolute’ method of calculating ADR which the pilot study (CEM (1996)) and other studies such as Baum (1981; 1997) have used (see section 2.0 of Appendix A). To ensure consistency we have used this ‘relative’ measure of depreciation throughout the report to calculate ADR.
Both EDR and ADR have their role to play in describing rental depreciation. EDR forms the main focus of the IPD data analysis, supplemented by ADR for towns and sectors, where data permits. ADR is also used in the case studies, because of the relatively small sample size.
It is important to appreciate that for EDR to provide meaningful results, a longitudinal analysis requires data for each property for each year over the time period selected. So for a 1980-95 study, each property included in the analysis would require full data for 1980-95.
This introduces the concept of cohort, which we define as a separate group of properties studied from the same start point over the relevant time period (Figure 3.1).
37 • maximise the use of the dataset; and
• compare EDRs over time and over market state.
An increasing depreciation rate over time can also be simply a result of an ageing sample. To control for this, the mean age of each cohort is shown at the start date in the report, (ie. for the 1984 cohort, the mean age is given at 1984) and EDRs are rebased to enable comparisons between different cohorts to be made.
Construction period was used as a proxy for age, and three construction periods were adopted for all sectors (post-1975, 1966-74 and pre-1965), which are comparable with the main grades of property in the market. The size of the dataset and the number of variables in our regression also suggested three periods was a sensible number to adopt.
The key advantage of adopting construction period, rather than age, was that it logically fitted our longitudinal analysis and also made it easier to report our results in a systematic way. As properties in a cohort age over time, age and ‘market state’ have a tendency to interact. In particular, as the age profile of a cohort changes, what was a ‘young’ building in a period of ‘boom’ becomes an older building in a period of ‘slump’. Construction period, however remains fixed, so we have used this throughout our analysis, except where age could highlight particular features of our results.
We have also used construction period as a proxy for analysing the importance of age as a ‘causal’ factor in depreciation: again, this fits the logic of the OLS regression and allows us to assess the relative importance of the other variables in our regression model (i.e. prime/non-prime and town type).
Apart from the reasons associated with methodology it should also be noted that the tables in the results section, which show age in years against depreciation rate, are limited in scope because of the shortage of a sufficient age range of properties in the cohorts. Generally, a wider range of ages was found in the 1990 cohorts, so we have included these where possible.
To examine the issues of Changes in locational quality and capital expenditure we also used further statistical techniques which are explained in Appendix A. The study focuses on ‘building- based’ expenditure on refurbishment, because it relates directly to physical deprecation.
‘Tenancy-based’ expenditure (eg. the cost of restructuring leases) is not therefore included in our analysis of capital expenditure, but the following items are included:
• refurbishment costs;
• change to the fabric of building; and
• initial development and redevelopment costs.
Finally, we distinguish in the report between an original building, or one that is not refurbished, and a refurbished building, or one that is refurbished (see Appendix A).
In summary, within each sector, three groups of properties have been analysed where possible:
• original with no locational quality change; • original with locational quality change; and • refurbished with no locational quality change.
Refurbished properties with locational quality change were not analysed because they were low in number and isolating the differential impact of locational quality change and refurbishment, alongside other factors in the OLS model, was not feasible.
3.4 Case Studies
In order to enhance the understanding of the process of depreciation over time, 33 case studies were collected, which were intended to provide a greater level of detail of depreciation against different market states.
An average depreciation rate (ADR) was calculated for each individual property using the averaging technique which compares the ERV:HP ratio over different time periods. An ADR was calculated for the longest period for which data was available and also for each market state and compared with the benchmark EDR. Where a property had been refurbished, average expenditure was calculated for the two years before and after refurbishment date, together with depreciation rates.
A total of 48 properties were submitted for inclusion as case studies. However, as in the main body of the report, a number were excluded due to lack of data or inconsistencies. Data for the period 1980 - 1995 was available from IPD and sponsors were asked to provide a commentary on the property and historic data from 1970 to enhance the analysis.
Data loss and other data inconsistencies, such as refurbishment dates with no matching expenditure, lack of Zone A, or no matching HP data, resulted in a final number of 33 case studies. The breakdown of the case studies is shown in Table 3.1 below.
Table 3.1 Case Studies Sector Breakdown
PROPERTY TYPE SECTOR NUMBER
Original Buildings with no Location Quality (LQ) Change
Office 14 Retail 5 Industrial 3 Retail Warehouse 2
Original Buildings with LQ Change Office 5
Retail 1
Industrial 1
Refurbishment (No LQ Change) Office 2
TOTAL 33
No shopping centres were included in the case studies because of data loss, inconsistencies and unavailability of alternative benchmark series Where the centre was dominant in a town.
For reasons of confidentiality, only a limited number of office case studies are presented in this summary report. These are shown in Appendix E.
39 4.0 RESULTS