Self-Reporting News Use in Situ and in Retrospect
Self-Reporting News Use in Situ and in Retrospect (DOI: https://doi.org/10.1093/ijpor/edaf027) was written by Danit Shalev, Teresa K Naab, and Yariv Tsfati in 2025. It was published in the International Journal of Public Opinion Research (vol. 37, no. 2)
There are two strategies for measuring news exposure: retrospective self-report and in-situ reporting. The former is challenged by memory and aggregation errors, social desirability bias, and proliferation of new media in new venues (i.e. social media). The latter is typified by ESM, in which respondents are repeatedly queried over time for their "recent behaviors, feelings, and cognitions". This eases the cognitive load, but is overall more burdensome for respondents.
The authors compare these measures. Over 4 days in August 2021, respondents participated in an ESM style study. On the 5th day, they were surveyed about their news consumption in the prior 4 days. The experiment is conducted in Israel, where prior research has established that television news is the most prevalent form of news (Limor, Tiargan-Orr, & Moshe, 2021, p. 90).
Because the key metric is televised news consumption, and because that platform within Israel is in Hebrew, the sample is effectively only the Jewish adult population of Israel. Additional screening is used to select respondents with at least one television.
The web panel Panel4all was used to recruit 6,396 sample members. The final data contains 1,414 complete responses. Note that attrition was almost entirely in the in-situ phase. It is difficult to compare the eligible sample to true population counts due to the unusual requirement of owning a television. In particular, ultra-orthodox Jews are severely under-represented. More generally, the eligible sample is more female, more educated, and more secular than the overall population.
Respondents significantly under-report televised news consumption in retrospect and over-report social media consumption in retrospect. Importantly, there isn't a consistent trend, so there may be important differences between news platforms.
Correlations of the measures are also reported: 0.507 for television, 0.680 for website/app, and 0.567 for social media.
The authors then regress on the difference between measures, to see how predictable the gaps are.
Frequent consumption (i.e. the count of in-situ reports that indicate consumption) is a significant predictor with a negative effect for all three platforms measured: television, social media, and apps/websites. There are two interpretations:
- frequent news consumers are more likely to under-report news consumption for all platforms
- there is a cieling effect in that frequent users cannot over-report given that their true value approaches the logical maximum
Habitual consumption (i.e., the odds that a respondent indicated consumption at a particular time of day; morning, afternoon, or night) is also a signficant predictor in several cases. Habitual consumption in any of the three times is significant for televised news consumption, with a positive effect. Habitual consumption in morning and afternoon is significant for social media news consumption, also with a positive effect. Altogether, there is evidence that habitual news consumers over-report.
Age and education are significant predictors for social media news consumption. Older respondents under-report, while educated respondents over-report.
Reading Notes
I find the way they display statistical information to be very confusing. Table 1 in particular. The retrospective consumption is reported in column 1, the aggregated in-situ consumption is reported in column 2, and the differences of the two are reported in column 3. But statistically significant differences between 1 and 2 are marked in 3 for some reason.
I cannot find the source (i.e. Limor, Tiargan-Orr, & Moshe, 2021) for the authors' assertion that television news is so prevalent in Israel. I suppose it's not so hard to believe that some sources are hard to come by, but this one is really obscure.