Weighting Household Surveys

1
Weighting Household Surveys

• By David F. Pearson, Ph.D., P.E.
• April 2007

2
Household Surveys in Texas

• Small sample
• Stratified by household size and income
• Telephone recruitment and data retrieval

3
Sampling Plan

• Samples not proportional to population household
  distribution
• Plan uses population household distribution and
  expected trip making
• Survey samples weighted for results to represent
  population

4
Household Weight

• Based on surveyed households and estimated
  population of households
• Computed for each cell in stratification matrix

5
Household Weight Example
2003 Smith County Household Weight Factors

• 4,297 is household population of Size 1 with
  annual income
• between 0 and 9,999
• 48 is number of households surveyed of Size 1
  with annual
• income between 0 and 9,999
• 89.52 is weight given to each household surveyed

6
Use of Weights - Examples

• Estimate population
• Estimate number of trips by purpose
• Estimate TLFDs by trip purpose
• Estimate vehicle availability distribution

7
Use of Weights - Examples

• Estimate employment, including self-employed
  individuals
• Develop household trip rates for use in TDMs
• Estimate persons by age and sex, including person
  trip rates

8
Results Dilemma

• Population estimates did not agree with expected
  values
• Population estimates by age and sex did not agree
  with expected values
• Employed persons were significantly different
  than estimates from other sources
• Concern Potential bias

9
Revised Approach

• Develop additional weight based on results from
  household weights
• Apply additional weight only to person variables

10
Revised Approach Steps

• Estimate number of persons by age cohort and
  gender using household weights
• Compute person factor based on population of
  persons by age cohort and gender

11
Calculating Person Weight Factor
i - age cohort, e.g., 40-to-44 years of
age j - gender, e.g., male or female PW - person
weight factor POP - population, i.e., number of
persons EST - estimate of number of persons from
applying household factor
12
How It Works

• Household variables weighted using household
  factor
• Examples of these variables include
• - Number of vehicles available
• - Household income
• - Number of bicycles available
• - Household length of residence

13
How It Works

• Person variables weighted using household AND
  person factor
• Examples of these variables include
• - Person trips
• - Persons licensed to drive
• - Persons employed (including
• occupation and industry)
• - Student status
• - Days worked at home

14
Area Characteristics
15
Distribution of Persons by Age
16
Distribution of Persons by Age
17
Person Trips per Person by Age
18
Distribution of Persons by Age
19
Distribution of Persons by Age
20
Person Trips per Person by Age
21
Comparison of EstimatesSmith County
22
Comparison of EstimatesHidalgo County
23
Comparison of Trip EstimatesSmith County
24
Comparison of Trip EstimatesHidalgo County
25
Employment Estimate Comparisons
26
Findings

• Use of two-stage weighting procedure
• Improves estimates of HBW travel
• Improves estimates of employed persons in area

27
Questions