Summary: Narrative presents a data visualisation (such as a bar, line, or pie chart) which misrepresents data in support of a claim.
This Technique was developed with reference to work by Nathan Yau of FlowingData, who has developed a resource containing interactive charts which help demonstrate how they can be used to mislead.
Data visualisations can be produced in such a way that they mislead the viewer on what conclusions can be drawn from the data.
For example, a chart which showed a change in voter support on a policy from 30% to 32% would look relatively insignificant on a scale which started at 0% and incremented 20% between labels, but would look significant on a scale which started at 25% and incremented 0.5% between labels.
Examples of misleading data visualisations as described by Nathan Yau include:
Axes’ scales overstating or dampening significance: As described above, charts can misrepresent what can be concluded from data by modifying the scale of an axis.
(Dual-axis charts with misleading scales:)[https://flowingdata.com/projects/dishonest-charts/#:~:text=Probable-,Cause,-It%E2%80%99s%20challenging%20to] A variant of the above, dual-axes charts can be used to imply correlation between two different datasets, facilitated by multiple y-axes. One or both can have their scales modified to misrepresent the presence (or lack of) correlation.
Cherry picking data: Data visualisations can be created with data selected for the purpose of supporting a claim, rather than for the purpose of providing an accurate representation of reality (T0164.002: Narrative Uses Selective Statistics to Support Claim). This can include focusing on a selection of data points which show an opposite trend demonstrated by the whole set, or selecting data from timeframes which support a given narrative.
Over-smoothing variance: Overapplication of statistical tools which filter out noise or outliers can mislead viewers into believing a trend is more reliable than it actually is.
Over-grouping data: Similar to above, data can be selectively grouped into buckets in a way which misrepresents conclusions which can be drawn from them. Users may be familiar with age range buckets (e.g. 21-30 y/o). A large bucket (e.g. 11-40 y/o) could produce results which don’t accurately represent reality.
Non-zero base scales on bar charts: Starting a bar chart at a non-zero baseline enables overexaggeration of a difference between two datapoints.
Visualisation produced using fabricated data: Visualisations produced using data which doesn’t exist (T0164.001: Narrative Presents Fabricated Statistics as Genuine Data).
Claim associated with visualisation which does not support it: A visualisation can be accompanied by a claim which is not actually supported by the visualisation. In today’s fast-paced information space, people commonly don’t have the time or expertise required to fully audit the validity of a data visualisation, and how it supports a given claim.
Tactic: TA14 Develop Narratives
Parent Technique: T0164 Issues with Presented Statistical Evidence
| Associated Technique | Description |
|---|
| Incident | Descriptions given for this incident |
|---|---|
| I00186 Fact check: No evidence of miscarriage surge since vaccine rollout | While vaccines are generally considered safe for pregnant women and new mothers, this group's exclusion from COVID-19 vaccine trials has left health care professionals with no clear data to guide their patients. But a new study released by the American Journal of Obstetrics and Gynecology in late March found that the vaccines based on messenger RNA, or mRNA, conferred good protection against the virus to both pregnant and lactating women, and likely their newborns. These encouraging results come on the heels of social media buzz claiming that, instead of protection, the COVID-19 vaccines are causing pregnant women to miscarry. "Miscarriages skyrocket 350% in six weeks due to k*vid vacsines (sic)," writes one Facebook user in a March 29 post. "Hmm... & y'all said it was safe. y'all said it was just 'conspiracy theories' when talking about the effect it would have on women's fertility," writes another user in a March 30 post that includes a screenshot of a headline from Natural News, a known conspiracy theory site, asserting the same claim as the March 29 post but with a figure of "366%." [...] So where does this 350% or 366% figure cited in the Facebook posts come from? The source appears to be London-based alternative news site The Daily Expose, which, according to its website, markets its mission as "to report the facts that the mainstream refuse to." In late March, The Daily Expose claimed that data from the U.K.'s Medicines and Healthcare products Regulatory Agency's Yellow Card Scheme (the British equivalent to the U.S. Vaccine Adverse Event Reporting System) showed an increase in miscarriages over a six-week period. "Using data inputted to the MHRA Yellow Card Scheme (from Dec. 9, 2020) up to 24th January 2021 a total of 4 women had suffered a miscarriage as a result of having the Pfizer/BioNTech vaccine," that article claims, including two more after vaccination with AstraZeneca's vaccine. After Jan. 24 to March 7, the total count goes up to 28 miscarriages for both vaccines which, accordingly, is a 366% increase. USA TODAY was able to verify the exact numbers provided in the January and March reports. While the Daily Expose's arithmetic isn't wrong, its conclusion isn't exactly right. "There is no pattern to suggest an elevated risk of miscarriage related to exposure to the COVID-19 vaccines in pregnancy," the MHRA said in a statement to Reuters. The agency explained the number of women vaccinated between December to March had to be considered alongside the expected frequency of miscarriage in a population. "The numbers of people who have received a 1st dose COVID-19 vaccination increased from 1,340,043 to 4,322,791 for the same time frame. At least half of these would be expected to be women, so the number of women of child-bearing age (taking the vaccine) is estimated to have increased from 665,424 to 2,146,866 for the same time frame," the MHRA said. It is estimated that as many as 26% of all pregnancies end in miscarriage, with nearly 80% of early miscarriages occurring in the first 12 weeks, or first trimester, according to the American College of Obstetricians and Gynecologists. Given this, the MHRA acknowledged "some miscarriages would be expected to occur following vaccination purely by chance." The Daily Expose article also claims two separate incidences of a premature birth and a stillbirth following vaccination. The MHRA explained that in the U.K., around 8 in 100 births are premature, so some would be expected to happen after vaccination also "purely by chance." The MHRA, however, disputed the claim of the stillbirth, telling Reuters some events can be incorrectly reported. It confirmed "no actual stillbirths" were reported to the agency at the time of its statement to Reuters in late March. [...] We rate this claim MISSING CONTEXT because without additional context it might be misleading (T0162: Reframe Context, T0160.004: Information is Misleading). Claims of miscarriages increasing by 366% over a six-week period originate from a U.K.-based alternative media site, The Daily Expose, citing data from the U.K.'s regulatory body, Medicines and Healthcare products Regulatory Agency. While the total counts of miscarriages are accurate, The Daily Expose article fails to take into account the increased number of women being vaccinated over the six-week time period (from December 2020 to March) alongside the expected frequency of miscarriages in the general population, which is around 26% (T0164.003: Narrative Uses Misinterpreted Statistics to Support Claim). The MHRA has stated there is no cause-and-effect association between COVID-19 vaccinations and the incidence of miscarriage. |
| Counters | Response types |
|---|
DO NOT EDIT ABOVE THIS LINE - PLEASE ADD NOTES BELOW