I. Introduction
Although
planning and policy analysis are perhaps ten percent retrospection and ninety
percent prospection, the amount of computer software
available to assist these functions is vastly weighted in favor of analyses of
historical data, rather than on forecasting the future implications of
alternative policies or demographic developments. The software for statistical analysis is well-known
and ubiquitous -- SAS, SPSS and many others -- with powerful versions available
for desk-top microcomputers. While
recent years have seen the advent of much-improved time series analysis
forecasting packages, the software available for supporting simulation or
projections under alternative assumptions about demographic or programmatic
changes is, by comparison, limited and relatively little-used.
A major factor
is cost -- a single application of a major microsimulation
forecasting program, for example, can cost thousands of dollars to set up, and
weeks of time may pass before the final results are available. The exigencies of many planning situations do
not allow for the luxury of such slow, high-cost, data-intensive and
labor-intensive techniques. Faced with
only a few hours to obtain, for example, an estimate of the budgetary
implications of a proposed new program regulation or policy, the analyst often
has to resort to "back-of-the-envelope guesstimates."
Times have
changed! The DESTINY Estimation,
Forecasting and Analysis System has been developed to provide the planner with
a low-cost, easy-to-use means for making detailed forecasts of target
populations, the need for services, the requirement for facilities, equipment,
and personnel, and the cost under varying programmatic and demographic
assumptions. The time required to
implement a DESTINY run ranges from a few minutes to a few hours, and the
computer cost is negligible -- a few minutes on a desktop microcomputer. The DESTINY system is a powerful new tool for
the planner who needs detailed, state-of-the-art forecasts on a
quick-turnaround, low-cost basis.
The essence of
planning and policy analysis is the ability to identify potential demographic
and economic developments, to synthesize alternative responses to those developments,
and to evaluate the impact of those responses.
The DESTINY planning and forecasting system has been developed to play a
crucial role in this process -- it offers the user the ability to make fast,
detailed population projections, and to forecast quantities linked to growth or
structural changes in the population.
DESTINY can
offer the planner or policy analyst valuable help in making a wide range of
forecasts...
In
Population Planning, DESTINY can be used to forecast the population by
age, sex, and race, or by geographic region and race.
In
Education Planning, DESTINY can forecast local-area school enrolments,
based on recent trends in birth rates and regional migration.
In
Health Systems Planning, DESTINY can project the requirement for health services
personnel (nurses, physicians), supplies and equipment (beds, CT scanners), by
geographic region.
In
Social Services Planning, DESTINY can be used to forecast the levels of
need for various social services (counseling, day care, protective services,
chore services), the requirement for agency personnel (counselors) and for
contract services, and the associated budget.
These forecasts can be specified by geographic region, or broken down by
age, sex, and race. The budget estimates
may be disaggregated by service, by resource (e.g., counselor), or by
demographic characteristics of the served population (e.g., age, sex, race, or
region).
In
Market Research, DESTINY can be used to estimate demand for products or
services whose demand is related to demographic changes.
In
Criminal Justice Planning, DESTINY can make projections of the prison
inmate population, and thereby be used to assist planning for prison
construction requirements under different sentencing policies.
The preceding
are but a few examples of DESTINY applications.
With DESTINY, the user can make forecasts such as the above quickly and
easily. The system can be used to make
forecasts at national, state, or local levels.
The analyst can use DESTINY to answer a wide range of "what-if"
type questions. DESTINY eliminates the
analyst's dependence on "standard" populations projections that may
be too highly aggregated, or may correspond to assumptions that are no longer
reasonable. Detailed forecasts,
corresponding to alternative demographic or programmatic assumptions, may be
developed in a matter of minutes or hours, in most cases using information that
is readily available from standard statistical sources.
The DESTINY
Planning and Forecasting System is an essential technical resource for anyone
engaged in planning who needs to be able to produce detailed demographic-based
forecasts quickly and at reasonable cost.
This brochure
describes the DESTINY system and illustrates its capabilities by means of
several examples.
This “international” version has been tailored for use in
developing countries, where the availability of detailed demographic, health
and social-system data and highly trained demographers, statisticians and
programmers is limited. The standard
version of DESTINY allows the user to make population projections under a wide
range of assumptions about the future demographic characteristics of a
population, by race and regional breakdowns.
This version is a simplified version that makes use of World Bank population
projections for 193 countries.
II. What DESTINY Does
DESTINY is a
computer-program system that is used to forecast the general population,
special "target" or "served" populations of interest, demand
for services to these populations, and the amounts of resources (personnel,
facilities, equipment, supplies) and cost required to provide these services to
the target population. It is designed to
provide fast, detailed forecasts in applications in which the items of interest
(target populations, served populations, services, resources, and cost) vary
roughly in proportion to the sizes of various segments of the general
population, such as age/sex/race/geographic-region categories. (For
the International Version, the population segments are restricted to age and
sex.)
A major use of
the DESTINY system is to support policy analysis and program evaluation in
service-oriented fields, such as health care, social services, education, and
public safety. DESTINY can be used to
develop service forecasts and budget estimates under a variety of demographic
and programmatic assumptions, and to generate detailed breakdowns (tables and crosstabulations) of these forecasts by age, sex, race, and
geographic-region categories.
Forecasting
models based on population projections are not new. In general, however, the amount of data and
computation required to develop a detailed population-based forecast is
substantial, since to achieve an acceptable level of accuracy the method
requires estimates of the future population broken down by age, sex, race, and
region. While a few total-population
projections may be readily available from national or state-level agencies, it
is usually not possible to obtain detailed projections (disaggregated by age,
sex, race, or region) quickly, corresponding to arbitrary demographic
assumptions, particularly at the local level.
One of the major features of the DESTINY system is its ability to generate
detailed population projections rapidly, corresponding to a wide range of
demographic assumptions (concerning fertility, mortality, and migration).
Except in the
field of population planning, most forecasting applications are not directly
concerned with projections of the general population, but instead with special
"target" subpopulations, such as the physically or mentally ill, the
disabled, the student population, or the prison population. The DESTINY system can be used to provide
forecasts of these special subpopulations, and forecasts of the resources and
cost required to provide specified services to them. Furthermore, these forecasts can be broken
down, or disaggregated, by detailed demographic characteristics -- by
age, sex, race, and geographic region.
At the local
level, planners are often confronted with the problem of estimating numbers of
persons with certain characteristics (e.g., acute illnesses, chronic health
conditions, disabilities, or social problems) or the number of
population-related events (e.g., school or prison admissions), but there is
usually no usable Census or sample survey information on incidences or prevalences of these characteristics at the local
level. What is often available, however,
are national or regional sample survey data that indicate the incidence or
prevalence, broken down by detailed demographic categories, such as age, sex,
and race. In the absence of local-level
incidence/prevalence data, the planner must use this national or regional
information to develop local-area estimates.
In this situation, a recommended procedure for estimating the numbers of
persons having the specified characteristics in the local area is synthetic
estimation. With this procedure, the
national incidences or prevalences for various
age-by-sex-by-race categories are multiplied by the total numbers of persons of
the general local-area population in these categories, to obtain estimates of
the numbers having the specified characteristics in the local area. DESTINY has the capability for rapidly
computing synthetic estimates. With this
feature, the local planner can quickly construct reasonable target population
estimates which make full use of available regional or national
incidence/prevalence data.
While the
DESTINY system can be quickly used to construct target-population estimates
based on available incidence/prevalence rates (specific to age x sex x race
categories), it can also be used in conjunction with other forecasting
procedures to produce high-precision forecasts.
For example, an analyst may use a multiple regression model, or an
autoregressive integrated moving average (ARIMA or Box-Jenkins) time-series
model, or a dynamic systems model to develop a precise forecast of a future
incidence rate for a specific demographic category (e.g., arrest rates for males
aged 20-35). This estimated arrest rate
can then be input to the DESTINY system to forecast total number of arrests for
a local area, taking into account not only trends in the arrest rate, but also
anticipated trends in birth rates, death rates, migration, and population
aging.
DESTINY has
been designed to enable the planner to obtain forecasts of the general
population, target populations, served populations, service needs, and the
resources and cost required to provide these services. These projections can be made under a
variety of "what-if" assumptions concerning demographic and
programmatic conditions. The analyst can
make alternative assumptions regarding population growth, target population incidences
or prevalences, program service ratios, and
costs. The corresponding projections can
be computed quickly, and can be disaggregated by age, sex, race, and geographic
region. DESTINY offers the user the
ability to make detailed forecasts quickly and easily, at substantial savings
over manual or partially-automated procedures.
III. How to Use the DESTINY System
The DESTINY
system was designed to provide a substantial analytical capability to the user,
but it was also designed with ease-of-use in mind. The DESTINY system works by setting up a
mathematical representation, or "model," of the population, and using
this model to project the future. To use
the system, the user needs to set up a "parameter" file containing
the following information:
Demographic
Information
o Total Fertility Rate (World Bank (“WB”) data
used in International version)
o Fertility Age Distribution (WB data used in
International version)
o Infant Mortality Rate or Expectation of Life
at Birth (WB data used in International version)
o Base-year Population, by age and sex (race
optional) (not included in International version)
o External Migration Rates or Amounts (WB data
used in International version)
o Regional Populations (optional) (not included
in International version)
o Regional Migration
Rates or Amounts (if regions are included in the model) (not included in
International version)
Target
Population Information (Optional)
o Incidences (rates of occurrence of events) or
Prevalences (proportions of the general population
belonging to subpopulations of interest), either overall or by age and/or sex
and/or race and/or region (overall or by age and/or sex in International
version)
Served
Population Information (Optional)
o Service Ratios (proportions of the target
populations that are served), either overall or by age and/or sex and/or race
and/or region (overall or by and and/or sex in international version
Service
Information (Optional)
o Average number of service units per year per
case served (a “case” is a member of the served population)
Resource
Information (Optional)
o Average number of resource units per year
required per service unit expended
Cost
Information (Optional)
o Average cost per resource unit
Names
o The names of the races, regions, target populations,
served populations, services, resources, and cost categories (races and regions
omitted from International version).
If only
general population projections are desired, (i.e., no target population or
service projections are desired), the user need specify only the demographic
parameters. If a detailed program cost
estimate is desired, all of the information must be specified. The demographic data are available from
standard statistical publications, such as national and state vital statistics
annual reports or statistical abstracts.
The target population incidence/prevalence data are available from
Census publications, national sample survey reports, agency publications, and
statistical abstracts. The service,
resource, and cost data are generally available from program administrative
records.
The process of
setting up the parameter file requires a little effort, but once the data are
input to the computer, they are stored on disk and are easily updated.
While running
the program, the user needs also to specify the following parameters:
o how many five-year periods to project
(limited to three for synthetic estimates, in the International)
o what tabular output is desired for the
various population, service, resource, and cost projections
With regard to
output, the program computes projections for the following items:
o the general population
o the specified target populations
o the specified served populations
o service units
o resource units
o cost (by cost category (element) or total)
The program
can generate aggregate projections for each of the above quantities, and will
(at the user's option) provide disaggregated projections, by age and/or sex
and/or race and/or region (by age and/or sex for the International version).
The standard
version of the DESTINY system is set up to accommodate up to three races,
fourteen regions, four target populations, ten services, seven resources, and
four cost categories. The International
version of the DESTINY system is set up to accommodate one target population,
one served population, up to ten services, ten resources, and ten cost
categories.
IV. Special Features of the DESTINY System
The DESTINY
system is designed to accommodate a wide range of detail in both the input and
output. In setting up the parameter
file, the user may specify that the same value of a demographic or
service-system parameter for the entire population for all future time. On the other hand, quite detailed demographic
or programmatic specifications may be made, to obtain highly
"conditioned" projections.
With regard to
specifying incidence/prevalence rates for the target and served populations,
DESTINY offers the user the choice of a number of different demographic
"stratifications." The user
may specify that the same rate applies to the entire general population, or
that the rates vary according to various demographic classifications, or
stratifications, of the general population.
Specifically, the user may specify rate stratifications by age, sex,
race, age x sex, age x race, sex x race, age x sex x race, region, or race x
region. With this flexibility, the
analyst may make use of a wide variety of available crosstabulation
data from national surveys, to construct synthetic estimates of target
populations for the local area. (For the International version, the
incidence/prevalence rates may be specified by eight different stratifications:
none, age, sex, age by sex, year, year by age, year by sex, year by age and
sex.)
In technical
terms, the DESTINY system uses the "cohort-component" method for
making its population projections. This
method is the most widely-used analytical method for preparing regional
population projections. (Shryock, Siegel, and Associates, The Methods and Materials of Demography, US Government Printing
Office, Washington, DC, 1980, presents a detailed description of this
method.) For the target population
estimates, the program uses the method of synthetic estimation (also known as
cohort-component participation rates). (As mentioned, the International version
uses the cohort-component population projections provided on the World Bank
World Development Indicators CD.)
The system is
set up to enable parameter input and updating interactively through the
computer screen. The specified
parameters are stored on disk for future use.
The projections may be viewed on the screen or printed.
The user may
specify either brief aggregated projections, or detailed disaggregated
projections, in the form of tables by age, sex, race, or region. In addition to demographic tables (such as a
breakdown of service utilization by race or region), the user may request
tables of services by target or served population; resources by service, served
population or target population; and cost by resource, service, served
population or target population.
V. Examples
This section
presents a number of examples of DESTINY applications. For the International version of DESTINY, the
examples apply to the country of Zambia.
These examples include:
o Projection of the General Population
o Projection of Various Target or Served
Populations
o
HIV/AIDS
o Basic school-age children and school
enrolments (grades 1-9)
o Projection of Service Needs
o HIV/AIDS services (VCT, ART, CSM)
o Formal schooling services (grades 1-9)
o Projection of Resource Requirements
o Personnel, drugs, other supplies, vehicles,
other equipment, facilities, outside services
o Teachers, classrooms, desks, textbooks
(grades 1-9)
o Projection of Program Costs
o HIV/AIDS
o Basic schooling (grades 1-9)
The following
examples illustrate the wide range of applications of the DESTINY system. They also illustrate the varying levels of
detail which are possible. On the one
hand, the user may use the program to construct a single estimate of the total
population; on the other hand, the user may request a detailed breakdown of a
target population (e.g., the HIV/AIDS population) by age, sex and year, or a
breakdown of costs for basic-school services by service and resource type, by
year.
While the
examples presented here illustrate the possible levels of detail of DESTINY
projections, they do not represent examples of policy analysis exercises. In a real planning or policy analysis
situation, the user would very likely make a set of DESTINY runs, under a wide
range of alternative demographic and programmatic assumptions (“what if”
studies). DESTINY is ideally suited for
such use, since it stores all of the input data in "parameter"
tables, which the user may easily modify with the touch of a few buttons on the
computer keyboard. Hence, while the
first DESTINY run requires some effort (to assemble the needed data and set up
the parameter tables), successive runs are easily and quickly accomplished.
The examples
and the principal features of each are listed below. They are presented in order of increasing
complexity. As mentioned, the examples
use Zambia data. The examples are
illustrated using actual system output, although some of the input data are
hypothetical. The projections are made
from a "base year" of 2000 to the year 2015 (for synthetic estimates
– longer for World Bank population projections). The user may structure the system output in a
variety of ways, requesting projections for various years and requesting different
tables for different years (e.g., detailed breakdowns for 2005, and summary
aggregations for 2010 and 2015). These
examples illustrate the wide variety of the program output.
Example
1. National Population Projection
Demographic
representation: single population (i.e., the total population), single
geographic region (i.e., the entire country)
Service-system
representation: none
Example
2. Health: Projection of HIV/AIDS
Services
Demographic
representation: single population, single region
Service-system
representation: one target populations (the general population), one served
population (the population receiving program services), two services (VCT,
CSM), seven resources (personnel, drugs, other supplies, vehicles, other
equipment, facilities, and outside services), and seven cost categories
(corresponding to each of the resource types)
Example
2. Health: Projection of HIV/AIDS
Services
Demographic
representation: single population, single region
Service-system
representation: one target populations (HIV/AIDS infected), one served
population (HIV/AIDS infected ART treatment), one service (ART), seven
resources (personnel, drugs, other supplies, vehicles, other equipment,
facilities, and outside services), and seven cost categories (corresponding to
each of the resource types)
Example
4. Education, Projection of
Basic-School Enrolment
Demographic
representation: single population, single region
Service-system
representation: one target population (citizens of basic-school age), one
served population (basic-school students), one service (formal schooling), four
resources (teachers, classrooms, desks, textbooks), and four cost categories
(salaries, facilities, equipment, supplies)
Example
1. National Population Projection by
Age and Sex
This example
illustrates the most basic application of DESTINY -- projection of a general
population by age and sex. In this
example, the resident population of
The system can
produce output in default (“standard”) graphical or tabular form, on
demand. DESTINY users who know how to
use the Microsoft Access database system can use the DESTINY output tables to
produce a wide variety of other tables (e.g., crosstabulations)
and reports. As users of MS Access well
know, the design of a custom-tailored report can consume a large amount of
time. The DESTINY output data are stored
in a small number of highly “normalized” tables from which reports may be
quickly produced using the MS Access report “wizard.”
The data on
which this projection is based were taken from the World Bank World Development Indicators CD for the
year 2005.
Projection of
Zambia Population by year, 2000-2015 (data from World Bank 2005 WDI CD)
|
Year |
Value |
|
2000 |
9842000 |
|
2001 |
10057600 |
|
2002 |
10273200 |
|
2003 |
10488800 |
|
2004 |
10704400 |
|
2005 |
10920000 |
|
2006 |
10920000 |
|
2007 |
11114400 |
|
2008 |
11308800 |
|
2009 |
11503200 |
|
2010 |
11697600 |
|
2011 |
11892000 |
|
2012 |
12095600 |
|
2013 |
12299200 |
|
2014 |
12502800 |
|
2015 |
12706400 |
Example 2. Projection
of HIV/AIDS Services to the General Population
This example
and the next one illustrate a typical use of the DESTINY system – the
estimation of a target population based on available incidence or prevalence
data, and the estimation of services, resources and costs associated with the
provision of service to a segment of the target population (i.e., to a “served”
population). This example illustrates
the application of the technique of "synthetic estimation," by which
incidence/prevalence rates by demographic category are used to construct
estimates of future caseloads, and specified service rates, resource
utilization rates, and unit costs are used to estimate future service
requirements, resource requirements and costs.
This example
is based mainly on hypothetical service-system data (target and served
population incidence/prevalences; service, resource
and cost parameters). It is intended to
illustrate the functionality of the DESTINY system. In a real setting, it would be necessary to
expend time to identify units of service, resource and cost, and to obtain
reasonable estimates or “what if” ranges for the various model parameters.
This example
is concerned with the provision of HIV/AIDS-related services to the general
population: voluntary counseling and treatment (VCT) and community
sensitization and mobilization (CSM).
The next example will address provision of antiretroviral drug therapy
(ART) treatment to HIV-positive persons.
The provision of HIV/AIDS-related services has been split into two
examples because the target populations for the services are different. (The standard version of DESTINY can handle
two or more target populations in the same model, but in the International
version a model is restricted to a single target population.) Much of the VCT and CSM services are provided
to the general population, whereas the ART services are provided only to the
HIV-positive population. A second reason
for splitting the example is that the outreach services are likely to be
“supply driven” (i.e., the budget specified by budget policy or availability of
funds) whereas the treatment services are likely to be “demand driven” (e.g.,
if government policy is to provide ART treatment to all HIV-positive persons).
To reduce the
complexity of the DESTINY system, service units, resource units and cost units
are not labeled – the units used in the examples may be as desired by the
user. For example, program costs may be
denominated either in Zambia Kwacha or in an
international currency, such as the US dollar.
This is the approach used in many statistical analysis programs. The units (of time, service, resource, and
cost) may be very specific, well defined entities, or they may represent very
general groupings of diverse elements.
In this first
example dealing with HIV/AIDS, the served-population unit (or “case”) is a
member of the general population (to whom services are provided) and the time
unit of analysis is one year. The
service rates are amounts of service (of each type) provided to each served
person per year. For example, the
service rate for voluntary counseling and treatment (VCT) is the average amount
of VCT service (however defined) provided per served person per year. An exact specification of the physical
composition of these services is not required – the service may be a single,
well-defined service (e.g., a single Community Sensitization and Mobilization
(CSM) training session conducted for 75 teachers over three days) or a “basket”
of expected amounts of many different types of service, provided by many
different service providers, per case, over the course of a year. The resource utilization rates are the
average amounts of resource (of each type) provided per unit of service (for
each type of service specified). Cost is
specified by cost category (element of cost), and may be specified in local or
international currency. In this example,
a separate cost category is specified for each resource type. In this example, the “Personnel Cost”
category might represent all personnel costs of all service providers and
government monitoring personnel for a year, or it might represent simply
government program personnel, with contract personnel costs included in
“Outside Services.”
In this
example and the one that follows, the service, resource and cost types are
purely illustrative – they may be defined and specified and measured as desired
by the user. This example will focus on
estimation of program cost, with no attention given to estimation of specific
service or resource elements. To keep
this example simple, it is assumed that each case (member of the served
population) utilizes one unit of each service, that the various services
utilize different resources, that each unit of service utilizes exactly one
unit of the resources that it does utilize, and that a separate cost category
is defined for each resource type. Costs
are assumed to be denominated in
In this
example, the target population is the entire Zambian resident population. As mentioned, the entire population is used
as the target population since the focus is on services are provided to the
general population (e.g., CSM), not just the HIV-positive or active-AIDS
segments. (If the focus of this example
were exclusively on services provided to the HIV-positive population, then the
target population would have been defined as all HIV-positive persons. Similarly, if the focus were exclusively on
Ministry of Education employees, then the target population would have been
defined as such.) The served population
is all persons receiving government-funded HIV/AIDS-related services
(preventative or treatment). In this
case, the prevalence rate is specified as 1.0 for all demographic groups.
For this
example, we shall assume that services are provided to 20% of the target
population in year 2000, to 50% of the target population in 2005, and to 90% of
the population in 2010 and 2015 (with linear interpolation in intermediate
years). We shall assume that each CSM
case costs an average of one US dollar per year ($0.8 for personnel, $0.1 for
vehicles, $.07 for other equipment, $.03 for facilities); that CSM services
cost $0.1 per case per year; and that all of the CSM services are provided as
“outside services” (e.g., services provided by contractor personnel). In a real setting, these parameter values
would be obtained from policy decisions (e.g., about the proportion of the
population to be served each year, and the level of service or service modality
to be employed) and historical unit-cost data (e.g., the average cost of
providing CSM services per person per year).
Normally, the costs of durable assets (vehicles, facilities) would be
the annualized expense, not the total purchase price. (The resources specified in this example
include “drugs” and “other supplies,” both of which occur at zero levels of
utilization. Those zero-level resources
are included in this example so that the resource sets are the same for this
example and the next one.)
The
Model
Input Data
Target Population:
The entire resident
Served
Population as a fraction of the target
population:
Year Fraction
2000
.2
2005
.5
2010
.9