Methodology - Country Database
By using multiple sources of data from a range of agencies, we are able to develop a complete picture of the demographic and socioeconomic landscape and the dynamics taking place within a country.
The source for the database is primarily the demographic and socio-economic data published by the government of each country. It is ‘supported’ by data from other agencies as this helps check the consistency of the interpretation of the data. Governments typically engage in a census and by-census at regular (typically 10 and 5 year) intervals and we supplement that with information from inter year samples, annual labour force studies, annual household income expenditure surveys etc.
By using multiple sources of data from different agencies we are able to develop a more complete picture of what the demographic and socio-economic landscape is like and the dynamics taking place within it.
To check on the overall veracity of the data from a government we also examine certain headline relationships, such as water consumption per capita, to ensure the reported levels of population etc make sense. Similarly we look for consistency of relationships across countries of similar affluence and education levels and, to the extent that there are differences, the possible reasons are investigated.
The database is harmonised as much as possible to facilitate modelling and hence consistency of forecasting process.
The database is updated at least once a year, and for many countries there are interim updates with the release of labour force surveys, household income and expenditure surveys etc., throughout the year. When updating, we always check the consistency of the most recent data with the preceding series for the same variable. We acquire the source document (or online database) for all data that we include in the database thereby allowing items to be back-checked when necessary.
The database includes the following variables – although for some countries not all data is available:
Age by gender – modeled to 1 year age steps.
Births by age of mother.
Deaths by age and gender.
Number of Households; proportion by number of persons in them; number of urban and rural households and average number of persons in each.
Education profile of the adult population.
Enrolment profile of school age population covering primary, secondary, vocational and tertiary.
Number of employed persons by gender, occupation profile and industry sector employed in.
GDP and GDP per capita and deflator.
Distribution of households by income.
Expenditure of both average household and by income group.
Measuring household income is difficult and unreliable. The usual method is to do what is generally called a ‘Household Income and Expenditure Survey’. This is done by most countries with a viable government but with widely varying degrees of reliability. It suffers from the normal issues of social research – that is sample validity and size, representation, respondent error, and analysis error. This is not to say these studies are totally unreliable – that is not the case at all – but rather they are at best a good indicator only.
There is also the issue of what is ‘income’? Is it earned income, does it include social payments, capital gains (including mortgaging the increased value of a household), credit? A credit card increases the spending power of a household by the amount of the credit limit. Yes, it must be paid back, but it is an ongoing continuous cash float.
Because of this uncertainty around this measure, Global Demographics Limited has taken a different approach. A good measure of average household expenditure is provided by dividing the Private Consumption Expenditure component of total GDP by the number of households. It pays to adjust this slightly to allow for expenditure by charitable institutions. Typically, we lower the Private Consumption Expenditure amount by seven per cent to allow for charitable institutions. It is also assumed expenditure by tourists is offset by overseas expenditure of residents. Again, an assumption but there is no easy way to define the amounts so spent.
As Private Consumption Expenditure (PCE) is a component of total GDP it is subject to some rigour in method and data collection. In addition, as the total number of households is an easy variable to measure that is also relatively reliable. So total PCE divided by total households gives a relatively reliable measure of the average expenditure of households in the country which by definition is also the medium-term definition of minimum household income
The next step is to determine the likely maximum funds available to households before tax and savings as well as expenditure. Here we resort to the Household Income and Expenditure Survey, where available. Some give not only expenditure but also gross income. Global Demographics divides that by average number of workers in the household to get the average income (wage) per employed person and that is compared with the overall GDP per worker. Using the distribution of this variable across 50 countries where this data is available there is a 95% probability that the average wage per worker will be less than 70% of GDP per worker. So, 70% of GDP per worker provides a likely maximum for the average worker wage per household – and that multiplied by number of workers gives the likely maximum accessible funds for a household in a year.
The range between the minimum (Private Consumption Expenditure per household) and the maximum (70% of GDP per worker times number of workers in the household) is actually quite small. Furthermore, given most households pay tax and/or save this means the true available funds must be greater than the minimum, and similarly not all countries pay at 70% of GDP meaning the likely maximum is lower the potential range is less. So Global Demographics Ltd use the average of the two numbers meaning the likely error in available income is plus or minus 5%.
Yes, this process can be debated but it does nonetheless mean that there is some consistency between estimated available income and total Private Consumption Expenditure as well as wages. Using Household Income and expenditure surveys alone under performs on this criteria.
Our unique strength is our econometric models of the historical demographic and socio-economic profile of each country and the use of these models to forecast the demographic and socio-economic landscape of these countries. The forecasts, covering a wide range of variables including age by gender, households, labour force, education and household income and expenditure, are a useful single source for strategic planning purposes.
The key drivers of the overall model are education and birth rates. Education profile of the adult population is forecast by using the projected time series trend in enrolment profile of persons aged 5 to 17. This in turn gives good estimates of the profile of those exiting the education system each year who are then added to the education profile of the adult population of that year as well as deducting the estimated education profile of those who die each year.
An Education Index is then used to drive the projected trends in urbanisation, occupation profile, productivity per worker (together with Fixed capital per worker) and household size (together with age profile). Time series trends drives birth rates, death rates and education enrollments.
The user is expected to use these forecasts as a base point – if the past relationships and trends continue then this is what will happen. That means they are a defensible base line – all changes can be related back to a past trend or relationship (and then ultimately source data) rather than the opinion of an individual.
In working form, the key projections are run to 2060 to see the long term impact of the projected trends. However, we only publish 20 years beyond latest actual, with confidence obviously being higher for 10 year forecasts.
1) Are proprietary to Global Demographics Ltd.
2) Are based on our own comprehensive database.
3) Use recognised statistical methods and processes – mainly econometric in style.
4) Can be explained to users (not a ‘black box’).
5) Are ‘constrained’ – meaning that they continuously check that different data items fit together.
6) Are well tested and continuously improved as more data becomes available.