Impact Consumer Survey Methodology

The Readership Institute's consumer survey data is from a multi-stage probability sample of the general U.S. population. The data were collected as part of the Impact study conducted by the Readership Institute at Northwestern University. The sample was designed to be both representative of the population and of newspapers. Technical details of the sampling procedures are given below.

The first step of the sampling process was to select a representative sample of daily newspapers in the United States. We compiled a sampling frame using lists of newspapers from the Newspaper Association of America (NAA), the Audit Bureau of Circulation (ABC) and Editor and Publisher. We excluded newspapers with the following characteristics: (1) average daily circulation under 10,000; (2) non-English language; (3) specialty newspapers such as Investor's Business Daily; (4) national newspapers (i.e., New York Times, Wall Street Journal, or USA Today). In total, the sampling frame consisted of 846 newspapers.

We stratified the sampling frame into six strata by applying k-means clustering to structure data from ABC, household counts from the US Postal Service, and demographic data from Claritas and the US Census. In defining the strata we needed to identify the "market" for each newspaper. We defined home counties as those counties that make up 80% of total circulation. The strata were defined using the average daily circulation, number of households in the home counties, Claritas' measure of urbanicity averaged over the home counties, number of competitive daily newspapers in the DMA, and a measure of market penetration in the home counties. Characteristics of six strata are summarized in Table 1.

We drew simple random samples without replacement from each stratum so that we would have approximately the same number of newspapers from each stratum. In total, 101 out of 116 newspapers agreed to participate in the study. The final list of participating newspapers included 18 from small town, 20 from small town/city+, 14 from small city local, 17 from city local, 15 from city regional, and 17 from big city.

Stratum	N	Circ	HHs	Zips	County	Urban	Pene	Comp
Small town	278	15,464	36,529	11.9	1.3	2.0	1.3	6.2
Small town/city+	162	36,500	68,897	30.6	3.6	1.6	1.3	3.7
Small city local	184	29,763	131,281	21.8	1.3	2.9	0.8	12.0
City local	81	96,864	212,684	34.4	1.5	3.0	1.2	9.2
City regional	64	111,397	219,378	59.2	6.1	2.0	1.2	3.4
Big city	77	366,887	956,606	112.7	3.3	3.6	0.9	10.2

The second step of the sampling procedure was to draw a random sample of consumers from each of the 101 newspaper markets. We drew names randomly from the zip codes accounting for 80% of circulation within each newspaper's home market. The sampling frame was lists of names compiled from a direct marketing list provider. We mailed 115,890 surveys between June 1, 2000 and July 15, 2000. The number of surveys mailed to each market was selected to produce approximately the same number of respondents. Surveys were allocated to zip codes within a market in proportion to a number of people living in the zip code. The individual in the household 18 years or older with the most recent birthday was asked to complete the survey. An incentive of $3 was attached to each survey, and responders were entered into drawings for 15 cash prizes. In total 37,036 responded, giving a response rate of 37%. The distribution of the number of responses in each market was normal shaped with a mean of 337, standard deviation of 46, minimum of 271, and a maximum of 472. Response rates in individual markets varied between 25% and 50% with a standard deviation of 6%.

The last step in the sampling procedure was to do a telephone survey of nonresponders. This was done to determine if nonresponders were systematically different from responders. Over the phone, we administered an abridged version of the mail survey to a random sample of 2000 nonresponders to the mail survey, approximately 20 from each market. We found that nonresponders were more likely to be nonreaders. The results of the phone survey were accordingly used to compute sampling weights to correct for this in the main survey. It turned out that 74% of the nonresponders were "readers," meaning they look at a newspaper during a typical 7-day week, while 93% of responders were readers.

Respondents to the mail survey were also weighted based on age and sex to make the sample more representative. Weights were computed to reflect a random sample from the United States using data from phone survey, Claritas, and the 1990 Census.

Impact Consumer Survey Questionnaire - PDF
Download a copy of the Impact Consumer Survey Questionnaire.