Statistics in science and in society: From a state-of-the-art to a new research agenda

Abstract

Statistics occupies a prominent role in science and citizens’ daily life. This article provides a state-of-the-art of the problems associated with statistics in science and in society, structured along the three paradigms defined by Bauer, Allum and Miller (2007). It explores in more detail medicine and public understanding of science on the one hand, and risks and surveys on the other. Statistics has received a good deal of attention; however, very often handled in terms of deficit – either of scientists or of citizens. Many tools have been proposed to improve statistical literacy, the image of and trust in statistics, but with little understanding of their roots, with little coordination among stakeholders and with few assessments of impacts. These deficiencies represent as many new and promising directions in which the PUS research agenda could be expanded.

Keywords

public understanding of science science communication science education scientific literacy social responsibility of scientists

1. Introduction

Nowadays, it is rare to hear a medical or economic report, to find a policy issue or to open a newspaper without being deluged by statistics. Statistics are frequently used as authoritative information: numbers are used to promote a point of view and challenge opponents. National statistics offices monitor economic, social and cultural changes in society and constitute a key factor in social and political debates. In politics, mostly linked with the New Public Management, objectives must be quantifiable to allow assessment and statistics are used as a tool for governance. As Robert Rodriguez (2013: 1) underlined, “statistics serves society” because “our society is data rich and data dependent”.

In science, mathematics has played an increasingly fundamental role over the past two centuries, in particular with the testing of hypotheses; the proximity between mathematics and physics has spread to natural and medical sciences, then to economic and social sciences. For Keller (2010: 914), statistics can be called “the quintessential interdisciplinary science”. In this context, techniques have been developed that generate subdisciplines (biostatistics, econometrics, chemometrics, etc.).

The importance taken by statistics in daily life has had consequences and scholars have emphasised underlying problems.

2. A series of problems framed differently

First, a call has emerged for greater statistical literacy of scientists, journalists, politicians and the public, and for a more critical evaluation of statistical results (Best, 2001; Goldin, 2009). The statistics community has long bemoaned the poor understanding of statistics on the part of the public; e.g., the 1952 report of the Royal Statistical Society on the teaching of statistics focused on the need for citizens to understand statistics used in the mass media or in survey results. Few data are available to document the level of statistical knowledge, but rhetoric about a growing statistically illiterate public has spread. Likewise, the poor statistical literacy of scientists, research ethics committees, and referees in journals publishing quantitative results has been deplored, but mostly without proof. As a consequence of scientists’ poor skills in statistics, society may encounter false statistical results. Statistical fallacies lead to erroneous conclusions, likely to be reported in the news and therefore to distort the public’s understanding of science (Siegfried, 2010a).

Debates about the construction of statistics have emerged, in particular for official statistics such as crime (Maguire, 2007) or health statistics (Harkness, 2012). How are indicators constructed? What affects these constructions? The issue of the purpose of such statistics has also been addressed: Do statistics aim to inform? To put an issue on the agenda? To achieve a hidden agenda, e.g., to induce guilt with indicators related to health (Harkness, 2012)? This discussion has cast doubt in the minds of many people who would feel justified by A. Levenstein’s quote: “Statistics are like a bikini. What they reveal is interesting. But what they hide is vital”. Spread in the media, this debate may have affected society’s confidence in statistics. As Hand (2009: 298) pointed out “There are many quips and quotations ridiculing the subject.”

Finally, as a result of the revolution in information technology over the last 40 years, we witness an expansion of the amount of data generated, data-processing power and new modes of data dissemination. Besides promises, such as open access via the Internet, concerns have emerged, such as the risk of feeling overwhelmed and of producing meaningless statistics, the issue of data quality control and of the interpretative monopoly over data. How do we help Internet users judge the quality of data and preserve them from malice or misinterpretation? When does the availability of data equate to increased knowledge?¹ The necessity for a regulatory system arose and various instruments were discussed.

Three paradigms

The way problems with statistics have been framed has similarities with the public understanding of science research paradigms, as codified by Bauer, Allum and Miller (2007). They defined three paradigms. Each “frames the problem differently, poses characteristic questions, offers preferred solutions, and displays a rhetoric of ‘progress’ over the previous one” (Bauer et al., 2007: 79). Their main points are:

The scientific literacy paradigm sets the problem in terms of a public deficit of scientific knowledge incompatible with public participation in political decisions. This deficit calls for more knowledge. The obvious solution is public education, paying attention to school curricula and encouraging continuing education.

For the public understanding of science paradigm, the problem lies in the tendency of society to become negative towards science. The axiom of this paradigm is “the more you know, the more you love it”. The solutions in this paradigm are to either seduce or educate the public.

The science and society paradigm sets the problem in terms of a crisis of the public’s trust due to a deficiency of the technical experts. Solutions are to change institutions and policy to rebuild trust. This paradigm recommends public participation, deliberation, and upstream engagement, including society in the early stages of the research process by balancing scientific evidence with public evaluations.

According to their authors (Bauer et al., 2007) these paradigms do not supersede each other, but continue to influence the setting of problems and the search for solutions.

The use of Bauer et al.’s (2007) paradigms aims to provide an original perspective on some current problems related to statistics, putting the evolution, growing influence, and concerns of statistics and science side by side, and proposing solutions to fix those problems derived from the paradigms.

In this 2013 Year of Statistics, we want to explore the relationship between statistics and science (Section 3) and between statistics and society (Section 4) by showing the relative place of the three previous paradigms. For each section, two case studies will be developed. We will select articles from various disciplines – statistics, history or philosophy of science, sociology, science education, science communication, etc. – to emphasise synergies between disciplines and dispel common misunderstandings (e.g., the statistics community is uninterested in tackling the issue through public engagement, social scientists are reluctant or naïve about the use of statistics). Finally, we discuss the limitations of the analogy with the three paradigms and propose an expansion of the research agenda.

3. Statistics in science

Two arguments – somewhat caricatural – compete within science: either statistics are at the core of the scientific method, mainly with null hypothesis significance testing, or statistics are the problematic products of a social construction. In any case, statistics are associated with the problem of being over- (Sorokin, 1956) or mis-used.

Authors lament scientists’ poor statistical knowledge, which falls into the scientific literacy paradigm, a concern expressed for example by the British Academy in October 2012 (BA, 2012). Few surveys have been completed to assess the statistical knowledge of scientists, most frequently in disciplines such as medicine (“Statistics and medicine” subsection below).

Instead of demonstrating it, authors underlined the frequent statistical misuses of scientists (Crettaz von Roten, 2006; Siegfried, 2010b; Bakker and Wicherts, 2011). The statistics community took action to better educate scientists, following the preferred solution of this paradigm. Among these actions, it is worth mentioning the modification of curricula (Meng, 2009), the creation of dedicated reviews (e.g., The American Statistician from the American Statistical Association (ASA) since 1947, Liaison from the Statistical Society of Canada since 1986, or Significance since 2004), the publication of guidelines for scientists (e.g., ESOMAR, 2009; United Nations, 2009), and statistics literacy campaigns such as Getstats² from the Royal Statistical Society.

Other authors highlighted a problem of attitudes (second paradigm). Keller (2010: 914) underlined negative attitudes towards statisticians: “many scientists still do not properly understand or appreciate the role of statisticians”; in contrast Brown and Kass (2009: 106) stressed the “cavalier attitude” of some scientists who attack statistical problems hastily, flouting their limited knowledge. Hand (2009: 299) asked “statisticians to be more outspoken when such incidents occur”, while Meng (2009) called upon statisticians to play a useful role in society by helping scientists provide correct results. The “publish or perish” rule has sometimes led scientists on an obsessive quest for significant results (Button et al., 2013).

Authors adopting a science and society paradigm are scarce. They believe quantitative scientists must regain the confidence of society, particularly by improving their communication skills (Fellegi, 1988). Lachenbruch (2009) explained communication issues by mutual distrust (statisticians did not trust the public or the press, and on the other hand the public shared the vision that statistics can lie by producing desired conclusions). Following the favourite solutions of this paradigm, Geller (2011) suggested moving from statistical consultation to statistical collaboration: statisticians should not simply offer data manipulation, but be “fully engaged in the project” (2011: 1226) at the early stages. The 2009 President of the ASA suggested that “statistics should be connected to society” (Morton, 2010: 2): statisticians should contribute to society’s needs to “transform the image of statistics” (2010: 2). And, to rebuild trust, the Swiss Statistical Association for example created an Ethics Board for Public Statistics to examine the quality of public statistics.³

The discussion on the social construction of statistics was developed in general knowledge books (e.g., Best, 2001) and specialised studies (Desrosières, 2002; Henning, 2010). MacKenzie (1978) showed how the dispute between Karl Pearson and George Yule on the type of indicator for statistical association among nominal variables could be linked with their different attitudes. Pointing out that attitudes cause problems, his study belongs under the public understanding of science paradigm. For MacKenzie, Pearson adopted a pragmatic position – maximising the nominal/interval analogy – in order to achieve his commitment to eugenics: the closeness of coefficients for nominal and interval variables allowed him to “prove the dominance of nature over nurture” (1978: 56) fitting his work into a social and political project. Yule, instead, adopted a looser approach by proposing a more general theory of association: he was interested in social issues (pauperism, vaccination statistics, etc.) and his approach was impelled by the broader needs of applied statistics.

To better understand the way statistical problems in science have been framed, we next discuss two scientific disciplines: medicine and the public understanding of science.

Statistics and medicine

Statistics have long been widely used in medicine, but, in the 1990s, evidence-based medicine reinforced the importance of empirical information for physicians and therefore the question of their understanding of probability and statistics, i.e., their ability to critically appraise the design, conduct and analysis of a study as well as the interpretation of results.

Many surveys were realised to assess the statistical knowledge of medical scientists (scientific literacy paradigm). To cite but a few, Wulf et al. (1987) reported a median number of two correct answers out of nine questions in a random sample of Danish doctors. Altman and Bland (1991) underlined doctors’ need of statistical knowledge by reporting their poor results in surveys. A series of studies on physicians showed deficiencies in basic numeracy skills and in the interpretation of medical tests and of relative risk reduction (Gigerenzer et al., 2007). Windish, Huot and Green (2007) compared statistical content in the medical literature and the understanding of statistical concepts of American medicine residents: the overall mean knowledge score was 41 percent, therefore most residents lack the knowledge needed to interpret most published clinical research.

The statistical literacy deficit can also be analysed indirectly at the level of reviews. The percentage of statistical errors in journals from several medical fields was documented (Altman, 2002; Ercan et al., 2007). John Ioannidis has taken a leading role in highlighting how the medical field of research is flawed, due either to statistical illiteracy or to conflicts of interests. He showed contradictions and problems in the most highly regarded research findings (Ioannidis, 2005). Button et al. (2013) documented consequences and ethical dimensions of the power failure of studies in neuroscience. In addition, many authors have just commented upon statistical illiteracy, for example Strasak et al. (2007) discussed 47 potential statistical errors and shortcomings in medical research. Switzer and Horton (2007) described the evolution of the statistical content of the New England Journal of Medicine to demonstrate that statistical knowledge of doctors should increase. Tressoldi et al. (2013) emphasised that a journal with a high impact factor does not guarantee high statistical standards.

The influence of medical scientists’ attitudes on statistical quality has been underlined, in line with the public understanding of science paradigm. Kaptchuk (2003) showed that doctors’ attitudes when assessing empirical research, for example, expectation of results or plausibility of results, can engender interpretation biases (e.g., confirmation, rescue or mechanism biases). Some studies found discrepancies between doctors’ positive attitudes towards statistics and low self-assessment of statistical knowledge (West and Ficalora, 2007; Swift et al., 2009). Conversely, Reynolds (2011) showed that emergency medicine residents’ negative attitudes toward statistics – varying from dislike to anxiety – drove them to increase their proficiency in statistics.

Analysing the problem in terms of trust (science and society paradigm), Gigerenzer et al. (2007) attributed statistical illiteracy mainly to doctor–patient relationship patterns based on paternalism and determinism on the part of the physician and trust in authority and illusion of certainty on that of the patient. For the authors, “Medicine in fact has held a long-standing antagonism towards statistics” (Gigerenzer et al., 2007: 74), because medical tact and personal trust have been replaced by quantitative facts.

Statistics and public understanding of science

Much has been said on the role of statistics within the public understanding of science (PUS) field. In the first Handbook of Science and Technology Studies (1995), Brian Wynne identified large-scale quantitative surveys as one main approach in this field. How did it evolve? To answer this question, we performed an exploratory analysis of Public Understanding of Science,⁴ the main journal in the field. We found that the quantitative approach reaches its peak in 2000 with roughly 40 percent of articles relying mainly on quantitative material, but without a clear pattern (Figure 1). The proportion of articles with no quantitative material varies from year to year, roughly 50 percent of the total, but the proportion of mixed methods (either quantification of qualitative material or quantitative and qualitative materials) increased slightly – between 17 percent and 24 percent until 2000, between 27 percent and 29 percent afterwards.

Figure 1.

Proportion of articles in Public Understanding of Science journal according to the type of material in 1995, 2000, 2005 and 2010.

Some authors reported problems with PUS surveys. Firstly, they stressed the influence of PUS scientists’ attitudes during the development of the survey, which refers in a way to the public understanding of science paradigm. Wynne (1995: 370), who played an important role in the association of survey research with the deficit model, summarised:

Large-scale social surveys of public attitudes toward and understanding of science inevitably build in certain normative assumptions about the public, about what is meant by science and scientific knowledge, and about understanding. They may often therefore reinforce the syndrome … in which only the public, and not science or scientific culture and institutions, are problematized in the PUS issue.

In the same vein, Davison, Barns and Schibeci (1997: 318) criticised surveys of attitudes towards biotechnology because they “serve to construct public opinion in a way that legitimates the commercialisation of biotechnology without necessarily enabling effective public debate”. Bensaude-Vincent (2001) argued that surveys aimed at enhancing the social legitimacy of science by depicting the public as ignorant and irrational.

On the other hand, Bauer (2008: 124) stressed that “[t]he intrinsic association of sample surveys and elite anxieties and the control of public opinion is a logical fallacy … The abuse of survey data is clearly possible and documented by the polemic, which is an achievement to be appreciated. But the misuse of an instrument does not exhaust its potential.” Kallerud and Ramberg (2002: 221) “explored a possible rationale for a ‘constructivist’ use of surveys of public understanding of science” and suggested the inclusion of new items related to civic perspectives on science.

Beyond the question of the validity and primacy of PUS surveys, some authors have underlined classical statistical problems in the survey design following the scientific literacy paradigm. At the questionnaire design stage, PUS survey developers were blamed for:

problems with measures: the influence of item wording, e.g. the term “science and technology” when no one knows if people differentiate between them and what the labels include (Pion and Lipsey, 1981; Ziman, 1991); doubt about attitudes towards European science (Bauer et al., 1994); the effect of the “neither/nor” option in the measure of attitudes towards science (Pardo and Calvo, 2002) or of “don’t know” responses in the measure of scientific knowledge (Bauer, 1996) and the danger of reifying knowledge measures (Ziman, 1991);

problems with bias: order bias and context effects (Beveridge and Rudell, 1988; Gaskell, Wright and O’Muircheartaigh, 1993), acquiescence response bias (Bauer et al., 1994) and cultural bias due to different conceptualisations of science among countries⁵ (Peters, 2000);

problems with the construction of indicators: absence of dimensionality analysis in the measure of scientific knowledge (Miller, 1998); difficulty in assessing precisely the reliability and validity of attitudes toward science and therefore in constructing reliable scales (Evans and Durant, 1995; Pardo and Calvo, 2002); problems with the construction of the attentiveness criteria (Beveridge and Rudell, 1988);

other shortcomings: the lack of background variables in surveys, such as media consumption, general value orientation, religious orientation, and the lack of open-ended items, for example on scientific method.

At the analysis stage, various issues were mentioned, for example:

frequent absence of margin of errors, limited information available on the data, instruments and sampling design (Beveridge and Rudell, 1988);

influence of designers’ implicit theories (deficit model), which leads to the assumption of a monotonic relationship between knowledge and attitudes towards science (Peters, 2000);

question of statistical power to detect differences of attitudes (Sturgis, Brunton-Smith and Fife-Schaw, 2010);

inefficiency of the use of a single demographic variable (e.g., age) to categorise science’s publics when composite descriptors are needed (Pion and Lipsey, 1981);

introduction of categorical predictors in regression model without interaction terms (Crettaz von Roten, 2004).

These statistical problems are in no way specific to the PUS field; however, methodological critical work has been sadly spread over reviews and books more or less close to the PUS field,⁶ which might have gone unnoticed by many readers.

4. Statistics in society

H. G. Wells’ prediction “Statistical thinking will one day be as necessary for efficient citizenship as the ability to read and write” seems now to be so, bringing about a series of problems. In accord with the scientific literacy paradigm, many have complained about the poor public understanding of statistics. Holmes (2003) traced the extent to which statistics penetrated English schools over the past 50 years, but identified a lack of statistical literacy. Two broad lines in this paradigm may be distinguished in the literature: one focused on the definition of public statistical literacy, the other concentrated on the measurement of the statistical literacy deficit in society or among its subgroups (students, patients, journalists, etc.).

Along the first line, many statisticians and statistics educators have defined and developed the concept of statistical literacy (Gal, 2002; Garfield, 2003). Wallman (1993: 1) characterised statistical literacy as “the ability to understand and critically evaluate statistical results that permeate our daily lives – coupled with the ability to appreciate the contributions that statistical thinking can make in public and private, professional and personal decisions”. To foster research and share information on statistical literacy, various networks were created,⁷ and national statistical offices took several initiatives.⁸

Along the second line, surveys were used to assess elements of public statistical knowledge. Some international surveys on science have included an item on probabilistic reasoning about a couple whose children have a one-in-four chance of suffering from an inherited disease: 66 percent of Americans and 69 percent of Europeans correctly identified the independence of risks among the children (NSF, 2010; EB, 2001). Longitudinal American data indicate that understanding has been fairly stable over time, the percentage answering correctly being around 65 percent since 1999. Another survey showed that 77 percent of Americans can read a simple chart correctly and 66 percent understand the concept of “mean” in statistics; the correct answer on these two concepts was associated with formal education, number of maths and science courses taken, income, and verbal ability (NSF, 2010). Unfortunately, surveys from international organisations, such as ALL, TIMSS or PISA, rank countries on mathematical or numeracy scales but offer few results specifically on statistical literacy. Gigerenzer et al. (2007) reported a series of results on the quantitative skills of various populations, e.g., 76 percent of US adults but 56 percent of female veterans in New England were able to estimate how many times a coin would come up heads in 1000 flips. More, a majority of German citizens incorrectly believed that DNA tests (78 percent of citizens) and HIV tests (63 percent) were absolutely certain. A survey on the understanding of probabilistic weather forecasts in five metropolises found that the correct answer was given by a majority of respondents only in New York (two-thirds, against roughly one-third in the European cities), a city where people have been exposed for a longer period to probabilistic weather forecasts (Gigerenzer et al., 2005). Schield (2011) found that 68 percent of college students misread a stacked bar chart, 44 percent misread the description of a single percentage, 19 percent a pie chart, 25 percent a scattergram, and 44 percent the percentage in a contingency table.

Unfortunately each survey covers a limited part of statistical literacy, and/or segments of the population. Therefore, public statistical illiteracy is more often discussed than demonstrated. Utts (2003) defined what educated citizens should know about statistics but frequently misunderstand, while other authors track the problem from statistical misuses in the media that may have an impact on the public (Best, 2001; Goldin, 2009).

Some research focused on attitudes (second paradigm) and therefore defined instruments to measure students’ attitudes towards statistics (Roberts and Bilderback, 1980). Then a positive link between students’ attitudes towards statistics and achievement in statistics was shown (Schau, 2003; Williams et al., 2008). A majority of sociology students in England and Wales expressed anxiety about learning statistics (52 percent) but saw a need for statistics in sociology (66 percent). For psychology students, low competence in mathematics as well as negative attitudes and anxiety towards statistics yielded low performance (Chiesi and Primi, 2010). More generally, Hand (2009) considered that a narrow perception of the role of statistics in modern life could reduce willingness to learn statistics.

Finally, some authors were concerned about low public trust in statistics (science and society paradigm). As reported in a 1998 meeting of the Royal Statistical Society, “statisticians are greatly concerned about the low public esteem for statistics. The discipline is often viewed as difficult and unnecessary, or at best as a necessary evil” (Lindsey, 1999: 1). In the UK, a 1999 White Paper initiated a discussion on trust in statistics and the government announced a series of arrangements to enhance trust in official figures (UK Government, 1999; Holt, 2008), in particular a system of “kitemarking” to certify quality. Monitoring of public trust in official statistics was discussed by the OECD in 2008. A 2009 British survey found an increase in disagreement over the accuracy of official statistics since 2004, a majority believing in manipulation or misrepresentation via political, governmental or media interference; people’s opinions discriminate among public statistics, and some statistical series were less trusted, in particular domestic burglary and unemployment figures (Bailey, Rofique and Humphrey, 2010). The UK Statistics Authority (2012) recommended improving the relationship between statistics and society by allowing users to provide feedback (from a one-way model of communication to a two-way model) and with recognition of the importance and value of government statisticians’ engagement.

To deepen the exploration of the relation between statistics and society, we now analyse two case studies: risks and surveys.

Risks and the public

One way in which people encounter statistics is in reading about or seeking guidance on risks. The issue of risk became prominent after memorable technoscientific crises such as the Chernobyl disaster or genetically modified organisms. For Beck (1992: 19), we have entered a “risk society” because “potential threats have been unleashed to an extent previously unknown”. Although there are many conceptions of risk, this section is limited to the traditional view of event probabilities and consequences, what Zinn and Taylor-Gooby (2006) called the statistical-probabilistic approach to risk. This approach is related to various problems under the umbrella of the scientific literacy paradigm. First, “They [people] rebel against being given statements of probabilities, rather than fact; they want to know exactly what will happen” (Slovic, 1986: 405). Then, quantitative estimates of risks are sometimes translated into qualitative statements (negligible, minimal, low, small, etc.) which confuse the public (Roche and Muskavitch, 2003; Amberg and Hall, 2010). Poor numeracy and misunderstanding of decimals, fractions or ratios may have consequences for grasping the notion of risk (Gigerenzer and Edwards, 2003; Galesic and Garcia-Retamero, 2011). The presentation of a risk in terms such as “10^-x per year” is not understood by most people (Slovic, 1986). More generally, the presentation of risk influences its interpretation. For example, people rated cancer as riskier when it was described as “kills 1,286 out of 10,000 people” than when it was put as “kills 24.14 out of 100 people”, illustrating how anchoring and base-rate neglect affect assessment of riskiness (Yamagishi, 1997). Information on consequences of risky behaviours was better recalled when presented in terms of life expectancy than of risk of disease (Galesic and Garcia-Retamero, 2011). Absolute and relative risks are comprehended in different ways (Edwards et al., 1999): if the benefit of a test was presented in the form of relative risk reduction, acceptance of the test was higher than when the same information was presented in terms of absolute risk reduction (Gigerenzer et al., 2007). Studies showed the generally lower understanding of numerical expressions of risks compared to graphs (Ancker et al., 2006; Goodyear-Smith et al., 2008). Risk management is often misunderstood: a study found that 81 percent of women in Italy believed that regular mammography would reduce or prevent the risk of getting breast cancer, 69 percent in the UK, 65 percent in Switzerland, and 57 percent in the US (Domenighetti et al., 2003). Finally, if the media have a key role in informing the public about risks, studies have reported misinformation and distortion (Roche and Muskavitch, 2003; Amberg and Hall, 2010).

Underlining the limitations in the public understanding of quantitative risks, Slovic (1986) stressed the need to analyse conceptual and cultural differences between the public and risk managers before communicating because “risk communication efforts are destined to fail unless they are structured as a two-way process. Each side, expert and public … must respect the insights and intelligence of the other” (p. 410). Consequently, many authors summarised the ways in which communication of risk can be improved (Calman, 1996; Paling, 2003; Goodyear-Smith et al., 2008). A significant example of failure in risk communication is the case of the L’Aquila earthquake that occurred in Italy in April 2009: a one-year trial followed where scientists were sentenced to six years for manslaughter not for failing to predict the earthquake but for incorrect risk communication to the population.⁹ That the presentation of risks influences its understanding and related behaviour may also raise dangers of manipulation for Thornton (2003) and ethical problems for Slovic (1986).

Risk management has become increasingly conflictual, due in part to a changed consideration of experts’ judgements. Therefore risks are associated with trust, in particular with scientists involved in risk assessment and regulators involved in risk management (Eiser, Miles and Frewer, 2002), following the third paradigm. Individual trust in risk management is influenced by gender, race, world views and affects (Slovic, 1999). “Trust is especially important in the absence of knowledge” (Siegrist, Gutscher and Earle, 2005: 146). In the case of biotechnology, trust varied between 70 and 80 percent of Europeans for doctors, university scientists and consumer organisations, between 60 and 69 percent for environmental groups and newspapers and magazines, and between 54 and 59 percent for the European Union, industry, government and shops (Gaskell et al., 2010). However, trust in actors fluctuated over time (e.g., doctors, university scientists and consumer organisations increased their trust level between 1999 and 2010) and among European countries (e.g., the biotechnology industry increased its level of trust in the UK, France and Sweden, but decreased it in Spain and Greece).

Surveys and the public in Switzerland

Following the science and society paradigm, a Swiss study¹⁰ indicated a majority of respondents trusted the Swiss Federal Statistical Office (Swiss Statistics) (58 percent). However, this level of trust was lower than for universities (83 percent) or the Swiss legal system (64 percent). Compared to respondents in a similar European study (EB, 2007), the Swiss tended to trust official statistics less than the Danish (73 percent) or the Dutch (77 percent) but more than the French (35 percent) or the British (33 percent). Swiss respondents agreed more readily to participate in surveys originating from a university than from Swiss Statistics and even more than from a newspaper.

Moreover the study allowed the checking of a potential problem of negative attitudes towards surveys (second paradigm). Representations of surveys were generally positive: useful as ways to gather information (75 percent), important for science (68 percent), participation may be interesting (66 percent). A relative majority thought that surveys are faithful to reality (49 percent) and essential for helping political decision-makers to take essential steps (40 percent). On the possible flaws of surveys, a relative majority considered them an intrusion into private life (39 percent agreed, but 35 percent were ambivalent).

Crettaz von Roten (2006) suggested that lack of public statistical literacy might obstruct the public understanding of science, and more generally the relationship between science and society. As a proxy to this hypothesis, we studied the relationship between attitudes towards statistics (two factors from a factor analysis: “benefits of surveys” and “concerns related to surveys”) and attitudes towards science. We found a positive significant correlation between attitudes towards science and the first factor (r = .13) – the more positive one is towards science, the more one perceives the benefits of surveys and vice versa – and a negative significant correlation between attitudes towards science and the second factor (r = −.18) – the more positive one is towards science, the less one expresses concerns about surveys and vice versa. These results indicate a link between attitudes towards statistics and attitudes towards science.

5. Discussion

This review of statistics in science and in society has presented relevant articles emerging from a wide range of disciplines. Problems emerging have similarities with the paradigms of the PUS field. In accordance with the pervasive scientific literacy paradigm, the problem with statistics is predominantly handled in terms of deficit – of either the scientists or the public. Results indicate a low level of statistical literacy but more data are needed to fully document the common assumption of a widening statistical knowledge gap. One reason why collective statistical illiteracy is not prioritised is that it is simply taken for granted (Gigerenzer et al., 2007). This brings a first research question to the agenda of PUS: What is the level of statistical knowledge? How does it vary among segments of the population? Across countries? How is it related to attitudes towards statistics and towards science? How do the public assess statistics when met for example in the media? How do the public use it in daily life? Eurobarometer surveys, which played an important role in PUS research, no longer scrutinise the public’s scientific knowledge. The PUS field might instead be informed by international surveys such as ALL or TIMSS. More than snapshot studies, there is a need for studies that fuel integrated databases for longitudinal analysis.

Many tools have been implemented to improve statistical literacy (Sanchez et al., 2011), similar to those devoted to improving scientific literacy. Regarding education, examples include modifications of curricula, support to teachers, new teaching tools, material proposed by national statistical offices, etc. Regarding informal teaching, activities were proposed such as science cafés, festivals, the creation of special events such as October 20, the UN’s World Statistics Day, contests such as CensusAtSchool or international statistical literacy competitions of the IASE/ISLP, support to journalists, media output such as the BBC programme The Joy of Stats, etc. (Isham, 2012). This variety of resources belies the belief that the statistics community is not interested in tackling the issue. Statistical approaches that build on public personal interest in a subject could be more actively used: for example increasing use of visualisation technologies, in particular in sports or weather forecasts, has the potential to enhance public understanding of measurement errors and confidence intervals, or the financial crisis has the potential to illustrate the consequences of poor understanding of variance or correlation.¹¹ The impact of these activities has not been fully assessed yet (Sanchez et al., 2011), maybe because as Townsend (2011) claimed, this is a long-term investment. Based on PUS experience of evaluation of science communication activities, a second research question could follow. What are the successes or failures in communication about statistics? Is there any specificity compared to science and technology communication? What are the best practices of statistics in the media, what are their impacts on the public and how can they be expanded? What is the best synergy among the actors related to statistics communication?

If one can welcome the multiplicity of tools proposed and of actors devoted to this mission (national statistical offices, statistics teachers, statistical associations, statisticians, etc.), one may wonder about the level of coordination among them. Leaving aside the possible difference in nature of this mission (mandatory or voluntary), it would nevertheless be worthwhile to explore drivers and barriers to such a coordination. To what extent is there a lack of commitment/trust? A competition for power distribution? A communication problem, for example a lack of conceptual clarity or cultural differences? A difference in orientation (long- versus short-term orientation)? A variety of specific goals (to increase the surveys’ response rate for national statistical offices, foster vocations for statistical associations, improve corporate image for statisticians, etc.)? On the other hand, what might be an efficient impetus for coordination?

However, thinking in terms of public deficit has shown its limitations, even if in this article, it is the element where the reference to the PUS paradigms is the most comprehensive. Statistics, as “an all-encompassing discipline” (Geller, 2011), has ramifications with attitudes or trust in the same ways as science, even if these aspects have not yet been studied extensively. We observe the emergence of contestations related to the constructions of some statistical indicators (unemployment or crime) and of discussion of their possible implications for political and social goals. This raises the question of the social responsibility of the statisticians building indicators, also of the eventual social responsibility of statisticians who develop statistical methods that are misused by others.

Moving from public deficit also means moving to the other actor of the relationship, bringing a third direction for research. Studies on statisticians could be developed to assess a series of aspects: the number of students educated in statistics, the number of statisticians involved in research projects, the evolution of the role played by statisticians in research teams, the impact of the production of knowledge (Gibbons et al., 1994) and of the pressure to publish on statisticians involved in research projects. More generally how has the image of the profession evolved? Has it preserved or lost its identity due to too much collaboration? Do statisticians feel obliged to engage in dialogue and debate? If we move towards a systemic vision of engagement (the statistician commits to explaining statistics to other team members and a spokesman engages with society), there is no guarantee the spokesman can engage effectively. The question of who must take the responsibility of public engagement has been extensively studied in the PUS field; results have shown that people want to meet the direct producers of the results but the media tend to interview people with the highest status (professors, directors of research units, etc.), therefore the status of a statistician in a research team may be a barrier to public engagement. For statisticians involved in higher education, the issue of the valorisation of the teaching task is important: in practice, introductory statistics courses are too often taught by lecturers with low degrees in statistics and little statistical practice, or by a lecturer without tenure and seniority, therefore preoccupied by their career (Soler, 2010). Is this the consequence of a shortage of young graduates in that area or of institutional inconsistency? Considerable pedagogical efforts do not, alas, improve a CV as much as collaborative or consulting work.¹² What are the problems encountered by academics in statistics in their career? All these questions deserve attention.

The common belief is that the public image of statistics is poor; if it’s the case in some countries (see De Santos, 2009 for Argentina), it’s not the case everywhere. Swiss respondents indicated an average trust for the Federal Statistical Office and positive representations of surveys. More than that, we found a significant positive relationship between attitudes towards statistics and attitudes towards science, which indicates the place of statistics in representations of science. If the motivation to learn is rooted in attitudes, this could signal that the improvement of statistical literacy and that of scientific literacy mutually reinforce one another. This gives an all the more important place to statistics in the PUS field.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

Author biographies

Fabienne Crettaz von Roten holds a doctorate in statistics from the Ecole Polytechnique Fédérale of Lausanne. She is now a senior lecturer and researcher and director of the Observatoire Science, Politique et Société, University of Lausanne, where she teaches statistics and is in charge of the research area “Relations Science–Society” of the Observatoire. Her research interests include public perceptions of science and technology, and scientists’ engagement toward society.

Yves de Roten, PhD, senior lecturer and researcher, is the director of the Center for Psychotherapy Research, Institute for Psychotherapy, University Hospital Center and University of Lausanne. He is also adjunct professor at McGill University, Montreal, Canada. He teaches research methodology for psychologists and psychiatrists at the University of Lausanne.

References

Altman

(2002) Poor-quality medical research: What can journals do? Journal of the American Medical Association 287(2): 2765–2767.

Altman

Bland

(1991) Improving doctors’ understanding of statistics. Journal of the Royal Statistical Society A 154(2): 223–267.

Amberg

Hall

(2010) Precision and rhetoric in media reporting about contamination in farmed salmon. Science Communication 32(4): 489–513.

Ancker

Senathirajah

Kukafka

Starren

(2006) Design features of graphs in health risk communication: A systematic review. Journal of the American Informatics Association 13(6): 608–618.

Bailey

Rofique

Humphrey

(2010) Public Confidence in Official Statistics 2009. Report for UK Statistics Authority, National Centre for Social Research.

Bakker

Wicherts

(2011) The (mis)reporting of statistical results in psychology journals. Behavioral Research 43: 666–678.

Bauer

(1996) Socio-demographic correlates of DK-responses in knowledge surveys: Self attributed ignorance of science. Social Science Information 35(1): 39–68.

Bauer

(2008) Survey research and the public understanding of science. In: Bucchi

Trench

(eds) Handbook of Public Communication of Science and Technology. New York: Routledge, pp. 111–129.

Bauer

Durant

Evans

(1994) European public perceptions of science. International Journal of Public Opinion Research 6(2): 163–186.

10.

Bauer

Allum

Miller

(2007) What can we learn from 25 years of PUS survey research? Liberating and expanding the agenda. Public Understanding of Science 16(1): 79–95.

11.

Beck

(1992) Risk Society: Towards a New Modernity. London: SAGE.

12.

Bensaude-Vincent

(2001) A genealogy of the increasing gap between science and the public. Public Understanding of Science 10(1): 99–113.

13.

Best

(2001) Damned Lies and Statistics: Untangling Numbers from the Media, Politicians and Activists. Berkeley: University of California Press.

14.

Beveridge

Rudell

(1988) The polls – a review: An evaluation of “public attitudes toward science and technology”. Public Opinion Quarterly 52: 374–385.

15.

British Academy (BA) (2012) Society counts: Quantitative skills in the social sciences and humanities. Available at: http://www.britac.ac.uk/policy/Society_Counts.cfm

16.

Brown

Kass

(2009) What is statistics? The American Statistician 63(2): 105–110.

17.

Button

Ioannidis

Mokrysz

Nosek

Flint

Robinson

Munafò

(2013) Power failure: Why small sample size undermines the reliability of neuroscience. Nature Reviews Neuroscience 14(5): 365–376.

18.

Calman

(1996) Cancer: Science and society and the communication of risk. British Medical Journal 313: 790–802.

19.

Chiesi

Primi

(2010) Cognitive and non-cognitive factors related to students’ statistics achievement. Statistics Education Research Journal 9(1): 6–26.

20.

Crettaz von Roten

(2004) Gender differences in attitudes toward science in Switzerland. Public Understanding of Science 13(2): 191–199.

21.

Crettaz von Roten

(2006) Do we need a public understanding of statistics? Public Understanding of Science 15(2): 243–249.

22.

Davison

Barns

Schibeci

(1997) Problematic publics: A critical review of surveys of public attitudes to biotechnology. Science, Technology, & Human Values 22(3): 317–348.

23.

De Santos

(2009) Fact-totems and the statistical imagination: The public life of a statistic in Argentina 2001. Sociological Theory 27(4): 466–489.

24.

Desrosières

(2002) The Politics of Large Numbers: A History of Statistical Reasoning. Cambridge, MA: Harvard University Press.

25.

Domenighetti

D’Avanzo

Egger

Berrino

Perneger

Mosconi

Zwahlen

(2003) Women’s perception of the benefits of mammography screening: Population-based survey in four countries. International Journal of Epidemiology 32(5): 816–821.

26.

EB (2001) Europeans, Science and Technology. Special Eurobarometer 55.2. Brussels: European Commission.

27.

EB (2007) Europeans’ Knowledge on Economical Indicators. Standard Eurobarometer 67. Brussels: European Commission.

28.

Edwards

Elwyn

Stott

(1999) Communicating risk reductions: Researchers should present results with both relative and absolute risks. British Medical Journal 318(7183): 603.

29.

Eiser

Miles

Frewer

(2002) Trust, perceived risk, and attitudes toward food technologies. Journal of Applied Social Psychology 32(11): 2423–2433.

30.

Ercan

Yazıcı

Yang

Özkaya

Cangur

Ediz

Kan

(2007) Misusage of statistics in medical research. European Journal of General Medicine 4(3): 128–134.

31.

ESOMAR (2009) ESOMAR/WAPOR Guide to Opinion Polls and Published Surveys. Available at: www.esomar.org

32.

Evans

Durant

(1995) The relationship between knowledge and attitudes in the public understanding of science. Public Understanding of Science 4(1): 57–74.

33.

Fellegi

(1988) The importance of good communication for statisticians. Liaison 2(2): 43–45.

34.

Gal

(2002) Adults’ statistical literacy: Meanings, components, responsibilities. International Statistical Review 70(1): 1–25.

35.

Galesic

Garcia-Retamero

(2011) Consequences of risky behaviours: Life expectancy versus risk of disease. Patient Education and Counseling 82: 30–35.

36.

Garfield

(2003) Assessing statistical reasoning. Statistics Education Research Journal 2(1): 22–38.

37.

Gaskell

Wright

O’Muircheartaigh

(1993) Measuring scientific interest: The effect of knowledge questions on interest ratings. Public Understanding of Science 2(1): 39–57.

38.

Gaskell

Stares

Allansdottir

Allum

Castro

P, et al.

(2010) Europeans and Biotechnology in 2010: Winds of Change? Report to the European Commission’s Directorate-General for Research. Brussels: European Commission.

39.

Geller

(2011) Statistics: An all-encompassing discipline. Journal of the American Statistical Association 106(496): 1225–1229.

40.

Gibbons

. (1994) The New Production of Knowledge: The Dynamics of Science and Research in Contemporary Societies. London: SAGE.

41.

Gigerenzer

Edwards

(2003) Simple tools for understanding risks: From innumeracy to insight. British Medical Journal 327(27 September): 741–744.

42.

Gigerenzer

Hertwig

Van den Broek

Fasolo

Katsikopoulos

(2005) A 30% chance of rain tomorrow: How does the public understand probabilistic weather forecasts? Risk Analysis 25(3): 623–629.

43.

Gigerenzer

Gaissmaier

Kurz-Milcke

Schwartz

Woloshin

(2007) Helping doctors and patients make sense of health statistics. Psychological Science in the Public Interest 8(2): 53–96.

44.

Goldin

(2009) Spinning heads and spinning news: How a lack of statistical proficiency affects media coverage. JSM – Section on Statistical Education 3890–3896.

45.

Goodyear-Smith

Arroll

Chan

Jackson

Wells

Kenealy

(2008) Patients prefer pictures to numbers to express cardiovascular benefit from treatment. Annals of Family Medicine 6(3): 213–217.

46.

Hand

(2009) Modern statistics: The myth and the magic. Journal of the Royal Statistical Society, Series A 172(2): 287–306.

47.

Harkness

(2012) Seduced by stats? Significance 9(1): 46–48.

48.

Henning

(2010) Mathematical models and reality: A constructivist perspective. Foundations of Science 15: 29–48.

49.

Holmes

(2003) 50 years of statistics teaching in English schools: Some milestones. The Statistician 52(4): 439–474.

50.

Holt

(2008) Official statistics, public policy and public trust. Journal of the Royal Statistical Society, Series A 171(2): 323–346.

51.

Hurwitz

Gardenier

(2012) Ethical guidelines for statistical practice: The first 60 years and beyond. The American Statistician 66(2): 99–103.

52.

Ioannidis

(2005) Contradicted and initially stronger effects in highly cited clinical research. Journal of the American Medical Association 294(2): 218–228.

53.

Isham

(2012) The evolving society: United we stand. Journal of the Royal Statistical Society, Series A 175(2): 315–335.

54.

Kallerud

Ramberg

(2002) The order of discourse in surveys of public understanding of science. Public Understanding of Science 11(2): 213–224.

55.

Kaptchuk

(2003) Effect of interpretive bias on research evidence. British Medical Journal 326(7404): 1453–1455.

56.

Keller

(2010) Vital statistics. Nature 467: 914.

57.

Lachenbruch

(2009) Communicating statistics and developing professionals. Journal of the American Statistical Association 104(485): 1–4.

58.

Lindsey

(1999) Some statistical heresies. The Statistician 48(1): 1–40.

59.

MacKenzie

(1978) Statistical theory and social interests: A case-study. Social Studies of Science 8(1): 35–83.

60.

Maguire

(2007) Crime data and statistics. In: Maguire

Morgan

Reiner

(eds) The Oxford Handbook of Criminology. Oxford: Oxford University Press, pp. 241–301.

61.

Meng

(2009) Desired and feared – What do we do now and over the next 50 years? The American Statistician 63(3): 202–210.

62.

Miller

(1998) The measurement of civic scientific literacy. Public Understanding of Science 7(3): 203–223.

63.

Morton

(2010) Statistics: From evidence to policy. Journal of the American Statistical Association 105(489): 1–5.

64.

NSF (2010) Science and Engineering Indicators 2010. Arlington: National Science Foundation.

65.

OECD (2008) How to Monitor Public Trust in Official Statistics. 5th Meeting of the Committee on Statistics, Paris. Available at: www.oecd.org/std/50167990.pdf

66.

Paling

(2003) Strategies to help patients understand risks. British Medical Journal 327: 745–748.

67.

Pardo

Calvo

(2002) Attitudes toward science among the European public: A methodological analysis. Public Understanding of Science 11(2): 155–195.

68.

Peters

(2000) From information to attitudes? Thoughts on the relationship between knowledge about science and technology and attitudes toward technology. In: Dierkes

von Grote

(eds) Between Understanding and Trust: The Public, Science and Technology. Reading: Harwood Academic Publishers, pp. 265–285.

69.

Pion

Lipsey

(1981) Public attitudes towards science and technology: What have surveys told us? Public Opinion Quarterly 45: 303–316.

70.

Reynolds

(2011) How do we promote clinical statistics literacy of emergency medicine residents: Is clicker technology the answer? JSM – Section on Teaching of Statistics in the Health Sciences 454–461.

71.

Roberts

Bilderback

(1980) Reliability and validity of a statistics attitude survey. Educational and Psychological Measurement 40: 235–238.

72.

Roche

Muskavitch

(2003) Limited precision in print media communication of West Nile Virus risks. Science Communication 24: 353–365.

73.

Rodriguez

(2013) Building the big tent for statistics. Journal of the American Statistical Association 108(501): 1–6.

74.

Royal Statistical Society (1952) The teaching of statistics in schools. Journal of the Royal Statistical Society A 115: 126–137.

75.

Salmon

(2012) The formula that killed Wall Street. Significance 9(1): 16–20.

76.

Sanchez

. (2011) The Millennium Development Goals, national statistical offices, the International Statistical Literacy Project and statistical literacy in schools. Statistical Journal of the IAOS 27: 157–171.

77.

Schau

(2003) Students’ attitudes: The “other” important outcome in statistics education. Proceedings of Joint Statistical Meetings, San Francisco. Available at: http://evaluationandstatistics.com/JSM2003.pdf

78.

Schield

(2011) Statistical literacy: A new mission for data producers. Statistical Journal of the International Association of Official Statistics 27: 173–183.

79.

Siegfried

(2010a) Flaws of statistics afflict science news reporting. Science News 177(27 March): 2.

80.

Siegfried

(2010b) Odds are, it’s wrong: Science fails to face the shortcomings of statistics. Science News 177(27 March): 26–29.

81.

Siegrist

Gutscher

Earle

(2005) Perception of risk: The influence of general trust, and general confidence. Journal of Risk Research 8(2): 145–156.

82.

Slovic

(1986) Informing and educating the public about risk. Risk Analysis 6(4): 403–415.

83.

Slovic

(1999) Trust, emotion, sex, politics, and science: Surveying the risk-assessment battlefield. Risk Analysis 19(4): 689–701.

84.

Soler

(2010) Who is teaching introductory statistics? The American Statistician 64(1): 19–20.

85.

Sorokin

(1956) Fads and Foibles in Modern Sociology and Related Sciences. Chicago: Henry Regnery Company.

86.

Strasak

Zaman

Pfeiffer

Göbel

Ulmer

(2007) Statistical errors in medical research – A review of common pitfalls. Swiss Medical Weekly 137: 44–49.

87.

Sturgis

Brunton-Smith

Fife-Schaw

(2010) Public attitudes to genomic science: An experiment in information provision. Public Understanding of Science 19(2): 166–180.

88.

Swift

Miles

Price

Shepstone

Leinster

(2009) Do doctors need statistics? Doctors’ use of and attitudes to probability and statistics. Statistics in Medicine 28(15): 1969–1981.

89.

Switzer

Horton

(2007) What your doctor should know about statistics (but perhaps doesn’t …). Chance 29(1): 17–21.

90.

Thornton

(2003) Patients’ understanding of risk. British Medical Journal 327(27 September): 693–694.

91.

Townsend

(2011) The national statistical agency as educator. Statistical Journal of the IAOS 27(3–4): 129–136.

92.

Tressoldi

Giofré

Sella

Cumming

(2013) High impact = high statistical standards? Not necessarily so. PLOS ONE 8(2): e56180.

93.

UK Government (1999) Building Trust in Statistics. Cm. 4412. White Paper. Available at: http://www.statisticsauthority.gov.uk/about-the-authority/uk-statistical-system/history/key-documents-about-the-uk-system/index.html

94.

UK Statistics Authority (2012) Strengthening User Engagement. Monitoring Report 7. Available at: http://www.statisticsauthority.gov.uk/reports–correspondence/reports/index.html

95.

United Nations (2009) Making Data Meaningful: Part 1 and Part 2. Available at: http://www.unece.org/stats/documents/writing.html

96.

Utts

(2003) What educated citizens should know about statistics and probability. The American Statistician 57(2): 74–79.

97.

Wallman

(1993) Enhancing statistical literacy: Enriching our society. Journal of the American Statistical Association 88(421): 1–8.

98.

West

Ficalora

(2007) Clinician attitudes toward biostatistics. Mayo Clinic Proceedings 82(8): 939–943.

99.

Williams

Payne

Hodgkinson

Poade

(2008) Does British sociology count? Sociology students’ attitudes toward quantitative methods. Sociology 42(5): 1003–1021.

100.

Windish

Huot

Green

(2007) Medicine residents’ understanding of the biostatistics and results in the medical literature. Journal of the American Medical Association 298(9): 1010–1022.

101.

Wulf

Andersen

Brandenhoff

Guttler

(1987) What do doctors know about statistics? Statistics in Medicine 6(1): 3–10.

102.

Wynne

(1995) Public understanding of science. In: Jasanoff

Markle

Petersen

Pinch

(eds) Handbook of Science and Technology Studies. Thousand Oaks: SAGE, pp. 361–388.

103.

Yamagishi

(1997) When a 12.86% mortality is more dangerous than 24.14%: Implications for risk communication. Applied Cognitive Psychology 11: 495–506.

104.

Ziman

(1991) Public understanding of science. Science, Technology, & Human Values 16(1): 99–105.

105.

Zinn

Taylor-Gooby

(2006) Risk as an interdisciplinary research area. In: Taylor-Gooby

Zinn

(eds) Risk in Social Science. Oxford: Oxford University Press, pp. 20–53.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.28 MB