Model to Structuring Total Population

Transcription

1 Applications of a Cohort-Component Component Model to Structuring Total Population Estimates to Categories of Age and Sex: A Pilot Study in New Mexico Jack Baker Adelamar Alcantara Xiaomin Ruan University of New Mexico April 18, 2008 p, Population Association of America, Annual Meeting

2 Presentation Overview Housing Unit-Based Estimates and the Cohort- Component Model (a la Leslie). Methods Evaluation of Results Conclusions & Study Implications

3 Jack Baker Adelamar Alcantara Xiaomin Ruan University of New Mexico April 18, 2008 Population Association of America, Annual Meeting

4 Housing Unit-Based Population Estimates Pop(t) = [HU(t)*OR(t)*PPH(t) + GQ A flexible approach. Provides total population estimates. May be preferred to alternative methods in specific instances. Does not provide age/sex estimates population structure.

5 Cohort Component Model and the Leslie Method Bryan (2004) suggested use of N 0 = N x *l x *m x a cohort component model for estimating age/sex structure. N 1 = N 0 *l 0 In situations ti where the HU. method is preferred, this is a. possible alternative.. Not previously piloted and N t = N t-1 *l t-1 evaluated. This study used a short-term cohort-component component projection model to structure a total housing unit-based population estimate for Bernalillo County, NM (2005). N t = [N t-1 *L] + M

6 Methodological l Steps Process for Producing Age/Sex Structured Estimates From a Housing Unit Based Total Estimate Cohort Component Model HU Based Total Estimate X Allocation Factor Matrix Age/Sex Structured Estimate

7 The Use of Allocation Factors for Structuring the Total Estimate A(0 4) N (0 4) A(5 9) N (5 9).. N (Tot) X. =... A (80 84) N (80 84) A (85+) N (85+)

8 Data Sources: Population at Risk and Rates Age-specific fertility rates Average Births (99-01)/2000 Census Population By Single Years Age-specific mortality rates Average Deaths (99-01)/2000 Census Population By Single Years Age-specific migration counts 5-year Census 2000 migration count estimates by five-year intervals. Special adjustment for 0-4 years based on child/woman ratio (Alcantara) By 5-year intervals.

9 Assumptions $470, No relationship between age/sex structure and housing unit growth. OR Relationship well-captured in data (especially migration). $196,000.00

10 Relationships Between HU-Based Estimates t and Population Age/Sex Structure # in Age Interval No Relationship # in Age Interval Relationship Specified HU Growth HU Growth

11 Diagnostics of Assumptions: Roll-up vs. Distribute Down Rolled Up Distributed Down Percent Percent Typically, HU-based estimates in New Mexico are higher, leaving a residual difference with a cohort-component estimate. In the model, this difference is distributed to each age/sex category according to predicted population proportions ( distributed down ). If assumptions are met, this should yield results identical to rolling-up each age/sex category by adding a constant share of the total residual between the two sets of estimates to each age/sex category.

12 Jack Baker Adelamar Alcantara Xiaomin Ruan University of New Mexico April 18, 2008 Population Association of America, Annual Meeting

13 Study Population: Bernalillo County, 2005

14 Housing Unit Growth in Bernalillo County 6000 Figure 2. Building Permit Issuance, County of Bernalillo: 1990 to Year

15 Study Population: Age-trends Population Pyramid Bernalillo County Age Males Females Percent

16 Jack Baker Adelamar Alcantara Xiaomin Ruan University of New Mexico April 18, 2008 Population Association of America, Annual Meeting

17 Diagnostic Evaluation If the technique works well, we expect to see: Concordance with alternative estimates of age/sex structure (County estimates from Census Bureau). No significant difference between rolling-up or distributing down.

18 Concordance in Estimates of Population Age/Sex Structure Table Component Based Age/Sex Specific Percent Estimates for Bernalillo County: Comparison with Census 2005 Estimates Age 2005 Cohort 2005 Cohort Absolute 2005 Census 2005 Census Bureau Absolute Component Component Difference in Bureau Estimate Estimate Percent Difference in Estimated Estimated Percents Percent Male Female Percents Male Percent Male Percent Female Female 0 to 4 7% 7% 7% 7% to 9 7% 7% 7% 6% to 14 7% 7% 7% 6% to 19 7% 7% 7% 7% to 24 7% 7% 8% 7% to 29 7% 7% 7% 7% to 34 7% 7% 7% 6% to 39 7% 7% 7% 7% to 44 8% 8% 8% 8% to 49 8% 8% 8% 8% to 54 7% 7% 7% 7% to 59 6% 6% 6% 6% to 64 4% 4% 4% 5% to 69 3% 3% 3% 3% to 74 3% 3% 3% 3% to 79 2% 3% 2% 3% to 84 1% 2% 1% 2% % 2% 1% 2% Total 100% 100% 100% 100% Average Difference

19 Equivalence of Rolling Up and Distributing Down Table 2. Method Diagnostics: Roll-up vs Distribute Down Male Female Absolute Difference in Percents Age 2005 Distributed 2005 Distributed 2005 Rolled Up Down Down 2005 Rolled Up Male Female 0 to 4 7% 7% 7% 7% to 9 7% 7% 7% 7% to 14 7% 7% 7% 7% to 19 7% 7% 7% 6% to 24 7% 7% 7% 6% to 29 7% 7% 7% 7% to 34 7% 7% 7% 7% to 39 7% 7% 7% 7% to 44 8% 8% 8% 8% to 49 8% 8% 8% 8% to 54 7% 7% 7% 7% to 59 6% 6% 6% 6% to 64 4% 4% 4% 5% to 69 3% 3% 3% 4% to 74 3% 3% 3% 3% to 79 2% 2% 3% 3% to 84 1% 2% 2% 2% % 1% 2% 2% Total 100% 100% 100% 100% Average Difference % %

20 Jack Baker Adelamar Alcantara Xiaomin Ruan University of New Mexico April 18, 2008 Population Association of America, Annual Meeting

21 Evaluation of Results The overall congruence between the age/sex estimates and the 2005 Census Bureau estimates was extremely high. There was no significant difference between the distributed-down and rolled-up percentages of the total population in each age/sex category. Results suggested the usefulness of this method for Bernalillo County and for other Counties with adequate data.

22 Implications and Future Work Method may work well in Counties with adequate data for both methods, but for which the housing unit-based total estimate is preferred. Method performed well with a minimum i of data (no updated d migration data beyond Census 2000). DATA (ON MIGRATION ESPECIALLY) MUST BE ADEQUATE! Method inappropriate where assumptions are violated (unless migration is very small!) Study does not apply to sub-county geographies or other small-area estimates.

23 Jack Baker (505)