FIVFIV 98d Release Notes

FIVFIV 98d Copyright 1998 Frederic C. Shorter and HPN Technologies Inc.

FIVFIV 98d is a major upgrade from FIVFIV 11.1. The authors would like to thank The Population Reference Bureau and the David and Lucile Packard Foundation for the funding necessary to make this upgrade possible. We would also like to thank Paula Hollerbach and Virginia Coon of the Association for Educational Development for the Spanish translations required for the upgrade and Linx Connexions for the French translations required for the upgrade.

Table of contents

• Extrapolation of the regional model life-tables to very high expectations of life at birth.

• Aggregating clusters

• Aggregations with differing initial years

• Known problems

• Technical notes

Extrapolation of the regional model life-tables to very high expectations of life at birth.

A basis for extrapolating the four regional model life-tables used by FIVFIV to expectations of life not so far found in recorded mortality experience is given by Coale and Guo. (Ansley Coale and Guang Guo. "Revised regional model life tables at very low levels of mortality," Population Index 55(4): 613-643. Winter 1989. With an addendum published in the next issue, 56(1): 26-41. Spring 1990.) They extrapolate to a female e(0) of 85.0 years. Their life tables have been included in the FIVFIV software for population projections ever since the year in which they were published (see note 1). The purpose here is to present additional extrapolations up to an e(0) for females of 95.0 years.

Recorded mortality shows a convergence of the country by country empirical basis of the tables in all the four regions (West, South, North, and East) to a single age pattern as e(0) rises. At high e(0)s, therefore, a single model is appropriate. It is named the "New West" model by Coale and Guo. It fits the age patterns experienced by advanced countries very well. In an impressive chart on the face of the 1989 issue of Population Index, Coale and Guo show how closely Japanese mortality experience fits this model.

Above an e(0) of 80 years (females), there is no recorded experience on the basis of which to build models for higher e(0)s. Nevertheless, Coale and Guo extrapolated an additional 5 years to an e(0) for females of 85 years. While one cannot be sure that any such extrapolations will be confirmed by future mortality experience, they are based on past trends and may be useful for making population projections.

We carry the extrapolations an additional 10 years to an e(0) for females of 95 years. There are so many possibilities for different developments in the future that these models for exceptionally high e(0)s should only be used with the understanding that they incorporate a single future scenario. Nevertheless, the demand for such models to illustrate future possibilities of population growth, and to illustrate theorems about the effects on population structure of exceptionally low death rates, motivates us to offer them as part of the FIVFIV population projection software (similar extrapolations are not made for the United Nations model life-tables. They remain unchanged in FIVFIV 98).

Since the four regional models all become the same at high e(0)s, the models are the same as "West" for e(0)s above a certain level in each regional model, starting at 77.5 years (females) for "East" and at 80.0 years for "South" and "North". Therefore, the extrapolations to higher e(0)s than 80 years for females are the same in all the regional models. Males are handled in the same way. At each level of the models, males have higher mortality and hence lower e(0)s.

The assumptions incorporated in the new extrapolations are reflected in the four levels (28-31) added to, "Table 14-1. Correspondence between selected indices of life tables by sex and both sexes combined: Sex ratio = 1.05: West," which originally appeared on page 92 of the Manual. The table shows, for example, that a joint (two sexes) expectation of life of approximately 90 years could be selected as a mortality assumption by choosing level 30. The table also shows the assumption of a decrease in the female-male difference of mortality rates as e(0) increases to very high values. A user of FIVFIV may accept this assumption, or could override it by choosing lower level numbers for males than females, or by directly choosing lower e(0)s and higher IMRs for males to fit one's own assumptions on this point.

An example of the life tables that we have constructed is appended. It is Level 28, which follows the highest level published by Coale and Guo, Level 27. Each life table is based on assumed age-specific mortality rates, (n)m(x) values. It is these rates which are extrapolated to produce higher and higher values of e(0). We also show next to the example the survival rates, P(x) values, which are derived from the life table for use in population projections. On the first line, P(birth) is shown, and then the regular survival rates up to the last one which closes the series, T(80)/T(75). More information is found in Chapter 14 of the Manual.

(Return to FIVFIV 98 Release Notes Table of Contents)

Aggregating clusters

FIVFIV 98d greatly extends the flexibility of the AGGREGATE subject. You may now aggregate individual projections into different clusters in addition to an overall aggregate. This is accomplished by assigning each projection to be aggregated to a cluster number. For example, suppose you have a FIVFIV run that looks like this:

       Projection 1: Nairobi
       Projection 2: Senegal
       Projection 3: Mali
       Projection 4: Thailand
       Projection 5: Malaysia
       Projection 6: Japan
       Projection 7: France
       Projection 8: Spain
  

and you wanted to aggregate into the following clusters: Africa, Asia, Europe, Developed, Developing, and World.

You would do this as follows:

• In FIVBLD select the new AGG.TITLE subject and create cluster titles as follows:

  
            Cluster 0: World
            Cluster 1: Africa
            Cluster 2: Asia
            Cluster 3: Europe
            Cluster 10: Developed
            Cluster 11: Developing
          

• For each projection select the AGGREGATE subject and assign the projection to the appropriate clusters. For example, you would assign the Nairobi projection to clusters 0, 1, and 11; the Thailand projection to clusters 0, 2, and 11; the France projection to clusters 0, 3, and 10; etc.

Notes:

• Cluster 0 is ALWAYS a complete aggregate containing all projections from the first projection that contains the AGGREGATE subject onwards. This is for compatibility with previous versions of FIVFIV.

• Each projection may belong to at most 4 clusters (cluster 0 plus up to 3 more).

• You may define up to 32 clusters (numbered 0 through 31; 0 always being predefined within FIVFIV 98 as a complete aggregate). Titles may be up to 24 characters long.

• You may place AGG.TITLE subjects anywhere in the FIVFIV run and the titles will accumulate across projections. Multiple titles assigned to the same cluster number will all be ignored except for the last. To avoid confusion we strongly recommend placing all AGG.TITLE titles only in the first projection of a FIVFIV run.

• You may not use SUBDIVIDE or AGESEXSPEC when using AGGREGATE.

• If you use PRINT PX in any projection, it will also appear in all the aggregates.

• If you want single year output for aggregate projections, you must have a REGROUP subject in the last projection of your run, and all single year aggregate output will use the grouping values specified in that REGROUP subject.

• If pyramids are requested in any projection, pyramids will also be produced in the aggregations. Aggregations will include pyramids for all periods that were requested in any individual projection. For example if projections 1 and 2 are aggreaged and projeciton 1 requests pyramids for periods 1 and 8 and projection 2 requests pyramids for periods 7 and 8, the aggregate will produce pyramids for periods 1, 7, and 8.

(Return to FIVFIV 98 Release Notes Table of Contents)

Aggregations with differing initial years

FIVFIV 98d will also aggregate based on initial year. For example, aggregations, if requested, will take place for stable extends, projections with old pops, and projections with different initial years (previous versions of FIVFIV would only aggregate based on a single initial year).

(Return to FIVFIV 98 Release Notes Table of Contents)

Known problems

Some FIVFIV runs created with FIVBLD prior to version 98 may abort with an error F6700. If this occurs, simply read the file into the FIVBLD 98 and read and save. This should eliminate the error. (If you are familiar with editing FIVFIV command files with a text editor, placing a HALT subject as the very last line in the file should also solve the problem.)

(Return to FIVFIV 98 Release Notes Table of Contents)

Technical notes

For those who like to edit FIVFIV command files directly, following is the syntax for the new features:

       AGG.TITLE
       TITLE.nn.aaaaaaaaaaaaaaaaaaaaaaaa
                 :
       TITLE.nn.aaaaaaaaaaaaaaaaaaaaaaaa
       AGGEND
  

where "nn" is a cluster number between 0 and 31, and "aaaaaaaaaaaaaaaaaaaaaaaa" is a title up to 24 characters long.

       AGGREGATE
       CLUSTER.nn.nn.nn.
  

where "nn" indicates up to 3 cluster numbers (0 to 31) into which this projection will be aggregated. Note that putting cluster 0 on the CLUSTER information line is redundant since all aggregated projections will be aggregated into cluster 0 whether so indicated on the CLUSTER information line or not. Therefore, each projection may be aggregated into at most 4 different clusters: 0 and 3 more indicated on the CLUSTER information card. The CLUSTER information card is optional.