Census Survey of Maricopa County Arizona Sample Design

Census Survey of Maricopa County, Arizona Sample Design October 7, 2004 Dennis Schwanz, U. S. Census Bureau 1

2005 Census Survey of Maricopa County Household Survey • Sample Sizes • Expected Accuracy • Sample Size Assumptions 2

Household Survey Sample Sizes Jurisdiction Sample Size (Housing Units) Percentage of Total Housing Units in Sample Avondale 3, 230 15% Buckeye 2, 130 53% Carefree 1, 540 73% Cave Creek 1, 470 68% Chandler 4, 640 5% 3

Jurisdiction Sample Size (Housing Units) Percentage of Total Housing Units in Sample El Mirage 4, 860 46% Fountain Hills 3, 510 30% 700 88% Gilbert 3, 350 6% Glendale 4, 650 5% Goodyear 3, 155 20% Guadalupe 900 69% 1, 170 69% Gila Bend Litchfield Park 4

Jurisdiction Sample Size (Housing Units) Percentage of Total Housing Units in Sample Mesa (Total) 13, 770 7% Mesa (each of 2 subareas) 6, 885 7% Paradise Valley 2, 265 38% Peoria 4, 665 9% Phoenix (Total) 27, 250 5% Phoenix (each of 5 subareas) 5, 450 5% Queen Creek 1, 775 39% 5

Jurisdiction Sample Size (Housing Units) Percentage of Total Housing Units in Sample Scottsdale 5, 965 5% Surprise 6, 250 16% Tempe 4, 710 7% Tolleson 1, 220 64% Wickenburg 1, 945 65% Youngtown 1, 505 68% Balance of County 7, 555 6% 6

Expected Accuracy of Household Survey Estimates • Population estimate will be subject to sampling error (random variation in sample estimates compared to the actual estimate) • Specifically, designed to have a margin of error (1 standard error) = 1% of survey estimate 7

• 95% confidence interval = survey estimate ± 2 standard errors = survey estimate ± 2% of survey estimate • This means there is a 95% chance the actual total population estimate will be in this interval 8

• In other words, this means that 19 out of 20 times, the estimated population from the sample will be within 2 percent of the actual population. • 99% confidence interval = survey estimate ± 2. 6 standard errors = survey estimate ± 2. 6% of survey estimate 9

Household Survey Sample Size Assumptions The sample sizes were based on four assumptions: 1. Variation in Household Size 2. the more the variation the larger the sample size needs to be - 2000 Census household size variation estimates used - more accurate estimates if less variation in 2005 - less accurate estimates if more variation in 2005 10

2. Vacancy Rate the higher the vacancy rate the larger the sample size needs to be - 2000 Census vacancy rate used - more accurate estimates if vacancy rate lower in 2005 - less accurate estimates if vacancy rate higher in 2005 - 11

3. Proportion of Total Housing Units in Sample - the higher the proportion in sample the smaller the sample size needs to be - the most current projected housing unit estimates used - more accurate estimates if actual housing unit total is lower than projection - less accurate estimates if actual housing unit total is higher than projection 12

4. 5. 6. 7. Household Survey Response Rate the less the higher the response rate the smaller sample size needs to be virtually 100% response rate assumed less accurate estimates if response rate than 100% 13