TABLE B-7 Deviations from the grand mean for absolute percentage error in projected population

   

Adjusted deviations

Grand mean = 5.49

Unadjusted deviations

(1)

(2)

(3)

(n)

Projection length

5 years

−2.74

−2.81

−2.58

1,256

10 years

−1.09

−1.01

−0.81

1,246

15 years

1.06

1.09

0.90

1,076

20 years

3.29

3.30

2.86

745

25 years

4.52

4.56

4.23

324

30 years

6.77

6.58

5.26

275

Year projected

1975

−2.05

0.65

−3.91

162

1980

−1.08

0.56

−2.93

437

1985

−1.38

0.16

−2.11

632

1990

−0.51

0.12

−0.66

1,189

1995

0.05

−0.02

1.40

1,133

2000

1.97

−0.40

2.90

1,369

Forecast

1972 World Bank

3.96

0.88

4.13

339

1973 U.N.

1.98

0.14

3.21

972

1980 U.N.

0.33

−0.13

1.10

810

1983 World Bank

0.37

0.69

0.34

580

1984 U.N.

−0.50

−0.26

−0.58

652

1987 U.S. Census

0.74

0.68

−1.29

246

1988 World Bank

−0.90

0.14

−1.73

480

1990 World Bank

−1.85

−0.66

−2.33

483

1994 U.N.

−3.46

−1.09

−3.67

360

Recent census

Census in 1990s

−0.48

0.07

0.07

0.07

3,174

No 1990s census

0.94

−0.13

−0.12

−0.12

1,748

Demographic quake

No quake

−0.78

−0.46

−0.46

−0.46

4,329

Quake

8.35

4.20

4.20

4.20

593

Initial population

Population <1m

3.03

1.35

1.41

1.41

1,047

Population 1m+

−0.67

−0.33

−0.34

−0.34

3,875

Initial fertility

Total fertility <4.5

−1.68

−0.14

−0.16

−0.16

2,223

Total fertility 4.5+

1.71

0.12

0.14

0.14

2,699

Region

Europe/FSU

−2.83

−1.09

−1.08

−1.08

818

U.S./Australia/Japan

−2.83

−1.46

−1.45

−1.45

152

Latin America/Carib.

0.37

0.67

0.69

0.69

945

Asia/Oceania

−0.91

−0.47

−0.46

−0.46

911

Mideast/N. Africa

3.77

1.93

1.89

1.89

671

Sub-Saharan Africa

1.41

−0.09

−0.10

−0.10

1,425

tries; and particularly, in the last few decades, for the Middle East and North Africa.

— Excluded from this regression.

Note: Adjusted deviations are given from the three analyses of covariance in the preceding table. The dependent variable was the cube root of absolute proportional error, but the grand mean and deviations are converted to absolute percentage error for convenience.

Absolute proportional error in base population estimates was analyzed separately to see if similar factors accounted for it. Some factors were not included in this analysis of variance, such as projection length, which would always be zero. Adjusted deviations for all the variables in the analysis are shown in Table B-8; all effects are significant at 0.001, except for initial population, whose significance level is 0.01. The results are not identical to those for proportional error in projected population. Which base year is estimated is somewhat more important as a determinant of error than which year is projected, in the preceding analysis. Later years are not necessarily more accurately estimated, however. Estimates for 1990 appear to be the best, but estimates for 1985 are among the worst. Initial population size still has an effect on accuracy, but a smaller one compared to its effect on projected population. Initial fertility level has a



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement