[Archive] Consultancy on best overseas manpower forecasting
- This report summarises the findings of a consultancy study commissioned by the Education and Manpower Bureau (EMB) of the Hong Kong SAR Government entitled "Consultancy on the Best Overseas Manpower Forecasting Practice."
- The study was commissioned in response to greater demands for information on labour market issues by businesses, career seekers and the Government in Hong Kong. In particular, EMB was keen to identify and review:
- existing manpower forecasting approaches in Hong Kong;
- the approach to manpower forecasting adopted in a selection of leading overseas economies;
- the perceived effectiveness of these approaches; and
- the scope for drawing from best practice overseas to improve the approach to manpower forecasting for Hong Kong.
- Hong Kong has in fact prepared manpower forecasts for many years. These forecasts have been prepared by Government and the Vocational Training Council (VTC) to meet specific needs. EMB's forecasts, for example, have been tailored towards planning educational provision for the whole economy, whilst VTC have focused on short term forecasts of employment requirements for specified industries in order to plan and revise the provision of courses. In addition, sectoral forecasts have been prepared by Government. For example, Business and Services Promotion Unit sponsored research on the future manpower and training needs of the information technology sector.
- The approach adopted by EMB, which also includes inputs from Census and Statistics Department (C&SD) is based on statistical projections. The basic approach is that labour force forecasts are produced using projections of population and labour force participation (based on extrapolation of past trends). Assumptions about future unemployment rates are then used to generate forecasts of employment, and these are broken down into occupations, again on the basis of past trends. From these, forecasts of future qualifications requirements are produced. The work that has been done to date has been diligent and thorough. However, there are some basic limitations to the approach - in particular the lack of modelling of underlying behavioural relationships in the economy (employment levels and inflation for example). This means that the approach is less responsive to external shocks, such as the 1997 Asian economic crisis, as models used in many overseas countries.
- The VTC work focuses on 22 industry groups for which it provides training. The forecasts are produced on a rolling two year programme and are survey based. Employer reporting of current and anticipated employment and qualification levels forms the basis for projections a year ahead. In addition, time series methods are used to generate forecasts for up to four years ahead (i.e. a trend is identified and extrapolated forward).
The Need for Improved Labour Market Forecasts
- This study has reviewed the past manpower forecasts and found them to be prepared in a thorough and diligent manner. However, a number of points have emerged from the review:
- Because the forecasts have been prepared with the specific objectives in mind, there is an understandable lack of integration between the forecasts, which tend to focus on a relatively narrow set of labour market features.
- The approaches adopted are often not best practice, partly because the systems were designed a number of years ago when the latest methodologies were not available and, in some cases, because certain data now collected were not available.
- The nature of Hong Kong's economy is changing. In particular there is an increasing focus on highly skilled occupations, where specific skills and qualifications are increasingly required. Such occupations clearly need more careful manpower planning than low skilled occupations, and this trend will be exacerbated by Government initiatives such as the proposed Cyberport.
- Changing patterns of migration may also place new pressures on the labour market. It is therefore important that Government is able to assess the implications of immigration and emigration on unemployment, wage rates and the adequacy of the workforce in Hong Kong. The major migration patterns in Hong Kong, including migration from the Mainland of China and the outflow of Hong Kong residents to countries such as Canada and Australia, are often determined by factors that are not economic in nature (for example, Mainland and Hong Kong policies towards migration into the SAR by Mainland residents). A flexible approach that enables the Administration to assume differing migration scenarios, and then, model possible impacts of those scenarios, would therefore be useful (1).
- There is therefore a need for an enhanced system of labour market forecasts which, by taking account of specific local circumstances and drawing on best overseas practice, can develop a customised model to suit the needs of potential end users of forecasts in Hong Kong.
Requirements for Future Manpower Forecasting
- Before embarking upon overseas research into manpower forecasting models, a detailed assessment was made of the requirements of potential end users of manpower forecasts in Hong Kong. This was undertaken through a series of consultation meetings with a range of organisations within and outside the Government. Analysis of the recommendations highlighted eighteen suggested requirements for future manpower forecasting. These suggestions related to both the scope and content of the forecasts, and the manner in which they were prepared and disseminated. The suggestions are summarised below.
SUGGESTED REQUIREMENTS FOR MANPOWER FORECASTING
- Able to respond to changing economic conditions and trends
- Predict aggregate labour force, employment and unemployment
- Provide sectoral breakdown of employment
- Forecast future skills needs
- Forecast future skills availability
- Assist in planning of education and training
- Assist businesses and individuals to plan for the future
- Provision of short, medium and long term forecasts
- Assist with importation of skilled and unskilled labour
- Predict vacancies
- Able to take account of Mainland economy
Forecasting Process Issues
- Provide assumptions and framework for other forecasts
- Tap private sector expertise
- Capable of providing frequent updates
- Easier to manage data and undertake model runs
- Innovative / proactive dissemination of results
- Moderate demands on parties providing data
- Ability to provide a range of forecasts
- This is an ambitious set of requirements for any labour market forecasting system. In developing a system to meet these needs, no one model would suffice. Therefore, developing a manpower forecasting model should be viewed as designing and integrating a framework of activities (some of which are already undertaken to a high standard in Hong Kong). In addition, although overseas approaches can be used as guides to improve forecasts for Hong Kong, it is important to recognise that manpower forecasting models must be tailor-made for an economy. There is no "off the shelf" model which can be purchased and then "filled" with Hong Kong data.
International Case Studies
- In order to investigate best practice overseas, a number of case studies were undertaken. The case studies were selected using a number of selection criteria, including:
- existing knowledge about the sophistication, uniqueness of approach or level of innovation in aspects of labour market forecasting;
- a desire to select a broad group of comparator economies, with a particular focus on developed countries; and
- a desire to select countries whose economic or demographic circumstances are similar (as far as is possible) to Hong Kong, such as being a city economy, being affected by the Asian economic crisis or experience of large migration flows.
- Based on the above criteria, the following case studies were selected:
The United Kingdom
The United States
International Labour Organisation
- The types of manpower forecasting practices revealed by the international research can broadly be grouped into four types:
- Those based upon an econometric approach, which produce GDP forecasts as well as breakdowns of employment and/or output by sector . These approaches are used in Australia, the Netherlands and the UK, and some of the work done in London, the US and New York. Singapore is now working with an Australian company to develop a detailed econometric model.
- Those based upon projections, in which forecasts are based on past behaviour. This approach is also used in some of the work done in Australia, the Netherlands, the UK and the US. The Hong Kong Government's current practice and the Hong Kong VTC's forecasts are using this approach.
- Those based upon surveys, with a focus on predicting short term employment vacancies. Belgium is an exponent of this approach, but the focus is on vacancies rather than employment. The Hong Kong VTC's own forecasts of employment are similar in nature to this approach.
- Those based on expert opinions, in which economic forecasts and survey data are reviewed by a panel of specialists to predict future trends. This is the approach adopted in Japan, and also for some of the work undertaken in London. The US also uses this approach to interpret and revise the outputs of econometric forecasts, and the UK is moving in this direction too.
A brief summary of the results of the overseas case studies is presented as an Annex (48 KB) to this Executive Summary.
- No one country or city's approach meets all of the requirements for manpower forecasting that have been suggested for Hong Kong. However, all of the requirements are addressed to some extent by at least one of the case studies. A table that shows which overseas countries have practices that would meet the various suggested requirements for Hong Kong is presented as Table 1 of this Executive Summary. However, it should be noted that many of the suggested requirements for Hong Kong are ambitious, and would draw upon the latest international approaches if adopted.
Options for Developing Hong Kong's Manpower Forecasting Framework
- In order to highlight the possible components of an enhanced manpower forecasting model, a series of options have been developed. One of these options (Option 8 - enhancement of the existing statistical projection model) represents a package of measures that could be implemented in a reasonably short time frame and make maximum use of existing expertise within the Administration. The other options would each improve upon one or more aspects of current practice and, if implemented as a package, would function as a coherent whole. However, not all of the options need to be implemented at once - each option has merits when considered on its own. Indeed, a phased implementation approach would be more manageable. In recognition of this,
- Ten options have been proposed. The options relate to both the methodology and scope of manpower forecasts, and also to process and management issues. They can be summarised as:
- Option 1. Refocus VTC surveys onto short term survey interpretation (one to two years), but broaden the scope to cover the whole economy. In addition to a more comprehensive coverage of the economy, it would also be advantageous to invest in the more rapid release of the results, and to present key findings in a summary format that would be easily accessible by end users, such as businesses and non-VTC education and training institutions. This option would mean that VTC no longer attempted to produce medium term forecasts through the use of time-series techniques, but instead highlighted trends in the near-future, based upon survey results. Data collected by VTC would, however, be of use in preparing the medium term forecasts described in Option 2, below.
- Option 2. Develop a detailed econometric model of the Hong Kong economy, to produce medium term manpower forecasts. Such a model establishes functional relationships between employment, productivity and other economic variables. It generates forecasts for output and productivity by sector which match total spending in the economy as measured by GDP. The forecasts for output and productivity are then used to generate forecasts of employment. This approach is adopted in the majority of the case studies. Econometric models offer the greatest potential for the timely production of forecasts, and for the running of "what if" test scenarios, such as assessing the impacts of major development projects or migration flows on employment and economic output. Models also have the advantage of being based upon economic theory, which ensures that outputs are consistent with what is known about the fundamental workings of an economy.
- Option 3. Develop a long term model and incorporate structural factors. The model would consider factors such as increasing labour productivity and the impact of technological change, in generating long term (ten year plus) projections. This option would require the development of a simplified version of the type of econometric model described above in Option 2. Having a separate, simplified model would allow relationships that are known to hold true over the long term (such as increasing labour productivity) to be specified, and the implications for the future demand of labour to be explored. This overcomes a weakness with short term models, which are dominated by demand side factors (relating to the demand for goods and services) in the short to medium term. These demand side factors can be volatile in the short term, and they also become less important in the long term, when it is the economy's structural development which influences its performance.
- Option 4. Improvements to the forecasting techniques for occupations. This option entails making use of existing, regularly updated survey data to enhance the understanding of changes in employment by occupation. This approach would supplement data derived from the census and by-census with data from the Business Establishment Survey and the General Household Survey. VTC survey data may also be of use. The present approach in Hong Kong for generating projections of employment by occupation is only to use occupations data from the census and by-census. That practice means that very few observations are available, and some of these are outdated and potentially misleading. The problems with the more up to date information sources are that they do not provide the wealth of occupational detail contained in the census and by-census data, and that they are derived from samples which inevitably limits their accuracy. However, there are techniques available for taking account of systematic differences between differing data sets. These include adjusting the coarser data to fit with detailed census and by-census data before introducing it to the model by comparing results in years when both surveys are conducted (3), the use of expert judgement or the use of moving averages rather than simple raw data.
- Option 5. Improvements to the forecasting techniques for qualifications. Qualifications forecasts are primarily of value for providing indications of any mismatches between employers' demand for people with differing levels of qualifications and the number who are likely to be available. The existence of mismatches then signify possible needs to alter educational, training or retraining plans, migration policies or policies for the alleviation of unemployment. It is important that the demand and supply parts of the methodologies be kept distinct, so that mismatches can be seen clearly. The current Hong Kong practice takes the projected supply of labour (based on net migration, demographics and a participation rate assumption), assumes a projection for unemployment, and thus derives an employment forecast as the difference between the two figures. This is treated as a forecast for the demand for labour, and may give misleading messages.
- Option 6. Improvements to the analysis of skills needs. This would entail amendments to VTC's survey design with questions on specific future skills needs. Both generic (team work, numeracy, language skills) and specific skills would need to be covered by the surveys. This area of work is relatively new, even amongst countries that are recognised as leaders in the field of manpower forecasting, so a certain amount of experimentation might be required. It is therefore recommended that it be considered once the major elements of a new forecasting system are in place.
- Option 7. Expert review of forecasts. This would apply the VTC's approach of employing sector experts to comment on forecasts, to the medium term economy-wide forecasts. The aim would be to provide opinion on trends, and early warning of major shifts in business or economic conditions. Genuine, senior experts from business and academia would be required for this to be effective.
- Option 8. Enhancement of existing statistical projection model. This option would be less ambitious than developing a new econometric model of the Hong Kong economy, but could still offer many improvements on current practice. The option would comprise collecting "soft data" on present and future trends through interviews with senior Government, business and academic specialists, making greater use of the results of the VTC's surveys, more frequent revision of manpower forecasts (perhaps biannually) and incorporating the results of forecasts produced by Options 4 (improvements to the forecasting techniques for occupations), 5 (improvements to the forecasting of qualifications), 6 (improvements to the analysis of skills needs) and 7 (expert review of forecasts).
- Option 9. Improved data warehousing. Under this option, forecast data and other labour market data would be collected together to form a database that would provide a comprehensive picture of the Hong Kong economy. This is resource intensive and therefore may not be a priority at present.
- Option 10. Improved publicity and dissemination of results. Past forecasts have generally been produced for internal Government consumption. To enable businesses, school leavers and other job seekers to take advantage of the work rapidly released, user friendly bulletins should be prepared for different target groups, making use of the print media, the internet and existing business and careers newsletters.
The resource implications presented by each of these options are summarised in Table 2 (64 KB) of this Executive Summary.
- Whilst a range of options have been identified, it is not recommended that all be implemented at once. This would be difficult in resource terms, and would probably detract from the ability to introduce the most important revisions in the short term.
- The first approach relates to priorities for the short to medium term and would be to adopt the course outlined in Option 8 (enhancement of the existing statistical projection model). Option 8 would mean that, in the short term at least, a macro-economic model (Option 2) would not be developed. This would limit the potential outputs of the forecasts, but could be up and running in a shorter time frame.
- In the longer term, the most important and effective improvement is to pursue Option 1 (refocus VTC surveys) and 2 (detailed econometric model). We would suggest that the development of the model (Option 2) should also incorporate structural factors into the long-term projections (Option 3) as part of the process. This would provide a variety of end users in Hong Kong with improved short term occupations data and a macro-economic model for better medium and long term forecasts.
- Publicity and dissemination should also be improved to coincide with the release of the first medium term projections produced under the new framework (Option 10). Furthermore, all options would benefit from an expert review of the forecasts (Option 7).
- As a part of improved forecasting, maximum use should be made of the most up to date data, and not just census and by-census results which are only available on a five yearly cycle. This would enable forecasts to provide substantially improved data for occupations. VTC should also include questions on skills (as opposed to qualifications) in future VTC surveys (Option 6), although it would probably be over-ambitious to incorporate this in the near future, particularly as such survey techniques are in their infancy elsewhere. Finally, data warehousing (Option 9) is an attractive option, but is an expensive and time consuming process and is therefore better addressed after the revised manpower forecasting framework has been developed and tested over the course of a year or two.
- Finally, there are alternative ways of implementing the recommended options. The key issue concerns the development of a macro-economic model (Option 2). Should Government wish to adopt Option 2 and the related recommendations, it is realistic to assume that outside assistance will be required. This should be provided by an organisation with previous experience of preparing econometric models for the purposes of manpower forecasting. As the overseas Government departments reviewed in this report all contract out model development and operation to consultancies or specialist university research centres, this would be in line with practice elsewhere. In time, however, it would be possible for the Administration to train an individual to operate the model, and retain the operation in-house. However, many of the other options such as Option 1(refocus VTC surveys) and Option 8 (enhancement of the existing statistical projection model) would be able to be developed by the Administration itself, as staff already have the required skills to do this.
- Overseas approaches may not be appropriate in the SAR. For example, migration between countries is often based on past trends. Similarly, methods of addressing internal migration flows, which examine factors such as job vacancies and house prices, are unlikely to be of direct relevance to the SAR given the small geographical size and good transport linkages of Hong Kong.
- ILO guidelines were found to concern coverage and process rather than techniques, and were very much aimed at developing countries.
- For example if the General Household Survey results for a particular data series are 10 percent lower than the census results in a given year, General Household Survey reulsts are uprated accordingly in that year and each subsequent year.