APPENDIX 2

1.1        This appendix presents the industrial floor space forecasting methodology. Separate forecasting models are developed for General Industrial Use (GIU) and storage floor space requirements.

2.          FORECASTING INTERNAL FLOOR SPACE REQUIREMENTS OF GIU

2.1        The methodology used to forecast internal floor space requirements of GIU is illustrated in Figure A2.1. GIU includes the Industrial (I) and Industrial/Office (I/O) land use types. The methodology involves principally the use of linear multiple regression analysis to establish the inter-relationship between the demand for GIU floor space and a wide range of independent variables which are believed to be the major factors affecting the demand for GIU.

2.2        Two measures of floor space demand for GIU are adopted as the dependent variables. They are:

• change in take-up for GIU - take-up is a measure of the annual increase in occupied floor space (IFA). It is defined as supply less demolition less vacancy of the preceding year plus vacancy at the end of the year. Change in take-up refers to the changing level of take-up compared with the previous year; and

• total demand for GIU - a stock based measure of the total floor space (IFA) being occupied. It is defined as stock less vacancy of the year.

2.3        Time series data for the dependent and independent variables, which provide the basis for model testing, are collected over the period 1976 to 1995. Using linear multiple regression analysis, three forecasting models for GIU are constructed. Two of them are used to estimate the total demand for GIU and one to estimate the change in take-up for GIU. The specific form of the models are as follows:

Equation 1:    TD_{year t} = B₀ + B₁ * (TD_{year t-1}) + B₂ * ln(EMP_{year t}) + B₃ * ln(RI_{year t})

Equation 2:    ln(TD_{year t}) = B₄ + B₅* ln(TD_{year t-1}) + B₆ * ln(RE_{year t}) + B₇ * ln(RI_{year t})

Equation 3:    ln(TUC_{year t}) = B₈ + B₉ * ln(EMPC_{year t}) + B₁₀ * [ln(TU_{year t-1}) - ln(EMP_{year t-1})]

where    ln        =    natural logarithm

            TD       =    total demand for GIU

            TUC     =    change in take-up for GIU

            EMP    =    employment of manufacturing industries

            RI        =    retained imports for industrial machinery

            RE       =    re-exports

            EMPC  =    change in employment of manufacturing industries

            TU       =    take-up for GIU

B₀, B₁, B₂, B₃, B₄, B₅, B₆, B₇, B₈, B₉, B₁₀ are constants

2.4        The definitions of the independent variables included in the models are:

    (a)  Employment of manufacturing industries (EMP) includes working proprietors, active business partners, unpaid family workers and all employees who are at work for pay or profit in the whole of the manufacturing sector, i.e. Hong Kong Standard Industrial Classification (HSIC) No. 3.

    (b)  Retained imports for industrial machinery (RI) is used as a proxy data to indicate the level of capital investments in the manufacturing sector.

    (c)  Re-exports (RE) are goods that are imported into Hong Kong and are subsequently exported without their shape, nature, form or utility being permanently altered.

    (d)  Change in employment of manufacturing industries (EMPC) refers to the changing level of employment in the manufacturing sector compared with the previous year.

    (e)  Take-up for GIU minus employment of manufacturing industries (TU-EMP) attempts to capture the long run economic relationship with change in take-up for GIU.

2.5        The floor space forecasts for GIU (IFA) are estimated by substituting the projected values of the independent variables into the three forecasting models. The model forecasts are adopted to construct a forecasting "range" for GIU. The highest value of model forecasts from the three equations is taken as the upper bound of the forecasts while the lowest value is taken as the lower bound of the forecasts.

2.6        From within the range, a forecasting "band" for GIU is constructed to indicate the upper and lower limits of potential demand. The band should be steadily widened to about 10% of the lower band figure in the last forecast year. The band position should take into account the past trends of the property and labour market to absorb the projected level of take-up and the economic theories underpinning the various models in order to highlight the more important variables. "Outlier" equations may be excluded in forming the band.

3.          FORECASTING INTERNAL FLOOR SPACE REQUIREMENTS OF STORAGE

3.1        The same methodology for forecasting GIU is used to forecast internal floor space requirements for storage. Linear multiple regression analysis is used to establish relationships between the demand for storage floor space and a wide range of independent variables which are believed to be the major factors affecting the demand for storage. Total demand and take-up for storage is adopted as the dependent variable in model testing.

3.2        Time series data for the dependent and independent variables, which provide the basis for model testing, are collected over the period 1976 to 1995. Using linear multiple regression analysis, two forecasting models are constructed to estimate the total demand for storage. The specific form of the models are as follows:

Equation 1:     ln(TDyear t) = B0 + B1 * ln(REyear t) + B2 * ln(FPRODyear t)

Equation 2:     ln(TDyear t) = B3 + B4 * ln(REyear t) + B5 * ln(TDyear t-1)

where    ln         =    natural logarithm

            TD        =    total demand for storage

            RE        =    re-exports

            FPROD =    floor space productivity of storage premises

B0, B1, B2, B3, B4 and B5 are constants

3.3        The definitions of the independent variables included in the models are:

(a) Re-exports (RE) are goods that are imported into Hong Kong and are subsequently exported without their shape, nature, form or utility being permanently altered.

(b) Floor space productivity of storage premises (FPROD) refers to the production level of the storage sector per square metre of floor area. It is the quotient of real value added of the storage sector and the occupied storage floor space in the previous year.

3.4        The floor space forecasts for storage (IFA) are estimated by substituting the projected values of the independent variables into the two forecasting models. The model forecasts are adopted to construct a forecasting "range" for storage. The highest value of model forecasts from the two equations is taken as the upper bound of the forecasts while the lowest value is taken as the lower bound of the forecasts.

3.5        From within the range, a forecasting "band" for storage is adopted to indicate the upper and lower limits of potential demand. The considerations described in paragraph 2.6 should also be taken into account in adopting the band.

4.          CONCLUSION

4.1        The forecasting models used to estimate floor space requirements, and the forecasts that they produce, are based on the testing of 1976 to 1995 data. Model re-calibration is required on a regular basis to take into account new data for the dependent and independent variables when they become available. New independent variables should also be considered when more significant structural changes of the economy take place over time. The precise model forms and the forecasting results are therefore subject to changes. The methodology will assist the process of regular data updating and model revision in the future.

Figure A2.1 : Territorial Industrial Floor Space Requirements Forecasting Methodology

[ Text Version ]