About The ARDA | Tools | FAQs | Contact Us
Time-Series Data

One of the most challenging but potentially valuable research approaches is charting changes over time, by comparing results from older surveys with newer ones. Ideally, one would compare only identical items from surveys that collected their data following the same methods. When these criteria are not met, very good judgment is required. One would think that surveys that are repeated over time would solve these problems, but here is a case where that is not true. The 1978 Middletown Area Study FREQREL asked, “How often do you attend religious services?” Here are the responses:

1) Once a year or less = 45 cases
2) A few times a year = 61 cases
3) About once a month = 33 cases
4) Two to three times a month = 23 cases
5) Every week = 88 cases
6) Several times a week = 35 cases
7) Never = 61 cases
8) Don’t know = 2 cases
X) Missing = 5 cases

Decades later, the 2004 Middletown Area Study asked ATTEND, “How often do you attend church?” The question was very different, asking about “church” rather than “religious services,” and the response categories were very different:

1) Never = 99 cases
2) About once a year = 22 cases
3) A few times a year = 72 cases
4) About once a month = 54 cases
5) More than once a month = 56 cases
6) Once a week = 147 cases
7) More than once a week = 90 cases
X) Missing = 60 cases

Both questions contain “about once a month,” but the fact they have different responses around that response makes it invalid to compare the numbers. The “never” response was chosen more often in 2004 than in 1978, but the placement of this response at the very top in 2004 may have captured more of the ambivalent responders. One could argue this was a factual report of the respondent’s behavior, rather than a mere opinion, and thus less affected by the exact structure of the questions, but we do not know how many people in 2004 attended “religious services” that were not in “church!” Of course, if it were valid to compare, one would look at the percentages, because the two surveys did not have identical number of respondents. In 1978 17.5 percent of 348 respondents said “never,” versus 18.3 percent of 540 in 2004, which does not look like a significant difference given the small numbers of cases.

A question called DATABANK in the General Social Survey related to computer technology provides an excellent illustration of how attitudes can change over the relatively short span of a decade or two. First asked in 1985, it asks: “The federal government has a lot of different pieces of information about people which computers can bring together very quickly. Is this a very serious threat to individual privacy, a fairly serious threat, not a serious threat, or not a threat at all to individual privacy?” While some people owned personal computers in 1985, the World Wide Web did not yet exist, and people had not yet gained well-founded opinions about this new technology. The same question was asked again in 1996 and 2006, in the middle of the personal computer boom, and after it. The percent of respondents who felt the threat was “very serious” grew from 31.9 percent in 1985 to 38.8 percent in 1996, then dropped to 25.5 percent in 2006. Over the same period, the percentage considering the threat “fairly serious” grew steadily from 31.2 percent to 35.0 percent to 40.1 percent. Yes, a threat may exist, but people were especially worried about it when the technology was new and rapidly expanding, and came to a more moderate assessment of the threat after that boom time.

The General Social Survey has slightly adjusted its sampling methods over the years, but the aim when it was created was to measure social indicators, such as family conditions and public attitudes, over time. Therefore it contains many identical items that were administered in different years. The same is true for the American National Election Studies, administered prior to elections every two years since 1948. It contains a standard question, VCF0128, about the respondent’s religious affiliation, with a somewhat complex wording but consistent over the years:

If you download the dataset and compare results over the years, you find a declining fraction who are Protestant, after increases in earlier periods, and an increasing fraction who profess no religion (not limited to atheists but including a variety of non-religious orientations) or religions other than Protestant, Catholic or Jewish. In terms of decades, the percent Protestant was 69.5 in 1948, 72.8 in 1958, 70.8 in 1968, 62.5 in 1978, 64.6 in 1988, and 52.0 in 1998. The trend is downward, but not perfect, and some of the variation is bound to be meaningless sampling error. The figure for 2004 is 55.4 percent. Similarly, the fraction whose affiliation was “other or none” fluctuates, but generally rises from 4.8 percent in 1948 to 15.5 percent in 1998 and 16.3 percent in 2004.

Ordinary respondent surveys are by no means the only source of time series data. Many nations ask a question on the standard census, including Canada but not the United States, and in the US census-like data collection efforts have been carried out at different times. Unfortunately they used a variety of methods, but usually working through specific denominations, so they are better for charting the rise and fall of specific denominations, and careful use of statistical corrections may be necessary when using the data to chart the course of religion in general.

ARDA offers these data in a great variety of useful forms, allowing many kinds of analysis. For example, the changing religious mix of specific states, cities, and counties can be charted. Take Fairfield County, Connecticut, for example, the part of the state closest to New York City, which has undergone great transformation from its Protestant past. From 1980 to 2000, there was an increase of 88,588 Roman Catholics, a growth of 26.0 percent, whereas the Protestant Episcopal Church lost 8,572 members for a decline of 24.0 percent. An estimate of the Jewish population of Fairfield County shows an increase of 30,983, or fully 396.0 percent growth, during the two decades.

There are many ways to connect such changes to theories and concepts. With caution, the growth of the fraction whose religious affiliation is “other or none” can be interpreted in terms of a secularization trend, or in terms of increasing diversity. Rapid changes in the denominational mix of an area usually reflect migration trends and the changing social structure of that area, but longer-term trends also depend upon demography – different fertility rates across denominations – and conversion to especially active local churches of the denomination.

QuickSearch The Knowledge-Base

To search the knowledge-base, enter a term below:

Select a Theory below to learn more:

Select a Concept below to learn more:

TCM Contributors

Would you like to be considered for a position on the Theories, Concepts, and Measures contribution team? If so, click on the link below and complete the TCM Online Request Form.

TCM Online Request Form

If you are already a contributor for Theories, Concepts, and Measures and need assistance in managing the content, please click on the link below for instructions.

Site Administration Instructions