02.02.2015 Views

East Bay Jewish Community Study - Jewish Federation of the ...

East Bay Jewish Community Study - Jewish Federation of the ...

East Bay Jewish Community Study - Jewish Federation of the ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Sampling Design<br />

The survey employed three different types <strong>of</strong> sample, each with <strong>the</strong>ir own sampling method:<br />

random-digit-dial (RDD) landline sample (n=620 completed interviews), listed cell phone sample<br />

(n=100 completed interviews), and listed landline phone sample (n=100 completed interviews).<br />

1) RDD Landline Sample: The representative landline sample employed an RDD procedure,<br />

<strong>the</strong> gold standard in survey methodology, to obtain access to a representative sample <strong>of</strong> listed<br />

and unlisted landline telephone numbers. The sample was ordered separately by <strong>the</strong> initial<br />

regional definitions. The sample was pulled in replicates, a method for structuring sample that<br />

allows for more a representative sample pull across a diverse population, initially using Type A,<br />

a broader sample pull that accounts for <strong>the</strong> increasing size <strong>of</strong> sample exchanges. This method<br />

is better at reaching a wider population area, but can lead to more disconnects and non-working<br />

phone numbers. This type <strong>of</strong> replicate proved to be inefficient, so we switched to Type B, a<br />

narrower sample pull that <strong>of</strong>ten results in a greater number <strong>of</strong> working phone numbers, after <strong>the</strong><br />

initial sample pull in order to better target productive telephone exchanges. Productivity <strong>of</strong> <strong>the</strong><br />

sample increased markedly. Due to <strong>the</strong> way California assigns telephone numbers and<br />

exchanges, nei<strong>the</strong>r <strong>the</strong> switch from Type A to Type B nor <strong>the</strong> targeting <strong>of</strong> productive exchanges<br />

should impact our ability to make accurate population estimates.<br />

The RDD portion <strong>of</strong> <strong>the</strong> survey also employed an adaptive sampling procedure. We compiled<br />

rough estimates <strong>of</strong> <strong>the</strong> <strong>Jewish</strong> population by county from a variety <strong>of</strong> sources, including previous<br />

demographic studies <strong>of</strong> <strong>the</strong> <strong>East</strong> <strong>Bay</strong> <strong>Jewish</strong> population, and used <strong>the</strong>se estimates to formulate<br />

an initial target number <strong>of</strong> completes for each region and to pull <strong>the</strong> Type A sample. As we<br />

began to collect data, we used <strong>the</strong> incoming data to re-calculate <strong>the</strong> incidence information and<br />

formulate a more accurate number <strong>of</strong> completes. In all, we recalculated regional targets and<br />

revised our incidence estimates four times during <strong>the</strong> fielding period to achieve <strong>the</strong> most<br />

productive and efficient sampling plan possible.<br />

Finally, we modified <strong>the</strong> sampling plan to slightly oversample regions with smaller <strong>Jewish</strong><br />

populations, in particular, Napa and Solano counties. Had we sampled <strong>the</strong>se counties in a<br />

purely representative way, we would have had too few interviews in <strong>the</strong>se counties and been<br />

unable to conduct a separate regional analysis. We oversampled <strong>the</strong>se counties just enough to<br />

obtain enough interviews to formulate population estimates in Napa and Solano without<br />

impacting <strong>the</strong> overall representativeness <strong>of</strong> <strong>the</strong> sample as a whole. Following <strong>the</strong> survey, <strong>the</strong><br />

data were weighted so that <strong>the</strong> share <strong>of</strong> <strong>the</strong> <strong>Jewish</strong> population from each region reflected to <strong>the</strong><br />

expected, representative proportions <strong>of</strong> <strong>the</strong> <strong>Jewish</strong> population by each region and county. Thus<br />

<strong>the</strong> final sampling plan sought to balance <strong>the</strong> need to obtain enough interviews in Napa and<br />

Solano counties with our need to truly representative data for analysis.<br />

The response rate for this portion <strong>of</strong> <strong>the</strong> sample was 2.4 percent overall (AAPOR response rate<br />

4), meaning that 2.4 percent <strong>of</strong> numbers dialed completed <strong>the</strong> entire screening portion <strong>of</strong> <strong>the</strong><br />

survey. Approximately 6 percent <strong>of</strong> households reached using <strong>the</strong> RDD were <strong>Jewish</strong><br />

households, using <strong>the</strong> definition outlined above.<br />

2) Listed Cell Phone Sample. Phone numbers for <strong>the</strong> cell phone sample were obtained from a<br />

consumer list supplied by Info USA Corp, a reputable sample vendor. The total list contained<br />

8,635 records for <strong>the</strong> greater <strong>East</strong> <strong>Bay</strong> region that had been deemed likely to be <strong>Jewish</strong> by <strong>the</strong><br />

vendor. This determination is made using consumer data, such as product purchases and<br />

subscriptions, as well as membership lists, surname lists, and o<strong>the</strong>r sources. As a result, listed<br />

cell phone sample tends to over-represent older Jews, more affiliated <strong>Jewish</strong> families, and those<br />

with traditionally <strong>Jewish</strong> surnames. However, it provided <strong>the</strong> only feasible, cost-effective<br />

101

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!