Introduction
Nudges are behavioral change techniques designed to help people make better decisions for themselves or society without violating their freedom of choice (Thaler & Sunstein, Reference Thaler and Sunstein2008). As a policy tool, they have influenced public administration in many countries (Whitehead et al., Reference Whitehead, Jones, Howell, Lilley and Pykett2014; OECD, 2017). There are several academic debates on governmental nudging. One concerns its alignment with citizens’ preferences and its effectiveness compared with conventional policy tools (Benartzi et al., Reference Benartzi, Beshears, Milkman, Sunstein, Thaler, Shankar, Tucker-Ray, Congdon and Galing2017; Loewenstein & Chater, Reference Loewenstein and Chater2017, but see Thaler Reference Thaler2017; Osman et al., Reference Osman, Radford, Lin, Gold, Nelson and Löfstedt2020; Weimer, Reference Weimer2020), another its transparency (e.g., Loewenstein et al., Reference Loewenstein, Bryce, Hagmann and Rajpal2015; Bang et al., Reference Bang, Shu and Weber2018), and a third its public support. One research program has surveyed the support for nudging in the USA (Jung & Mellers, Reference Jung and Mellers2016), the UK, Denmark, France, Germany, Hungary, Italy (Reisch & Sunstein, Reference Reisch and Sunstein2016; Reisch et al., Reference Reisch, Sunstein and Gwozdz2017), Australia, Brazil, Canada, China, Russia, South Africa, South Korea (Sunstein et al., Reference Sunstein, Reisch and Rauber2018), Sweden (Hagman et al., Reference Hagman, Andersson, Västfjäll and Tinghög2015, Reference Hagman, Erlandsson, Dickert, Tinghög and Västfjäll2019), and Flemish Belgium (Sunstein et al., Reference Sunstein, Reisch and Kaiser2019). Sunstein et al. (Reference Sunstein, Reisch and Rauber2018) categorize three levels of nudge support internationally: overwhelmingly pronudge nations (China and South Korea), principled pronudge nations (Brazil, South Africa, Russia, Australia, Italy, the UK, Canada, France, Germany, and the USA), and cautiously pronudge nations (Japan, Hungary, and Denmark).
In this article, we address two recent discussions in the literature. One of them concerns the relationship between trust in public institutions and the support for nudging. Previous research has theorized, ‘People who have high trust in public institutions would be more willing to accept government nudging’ (Sunstein et al., Reference Sunstein, Reisch and Kaiser2019, p. 1423) and has sought to explain international differences in the support for nudging with the same theory (Sunstein et al., Reference Sunstein, Reisch and Rauber2018).
The other discussion is the relationship between political party preferences and the support for nudging. Sunstein et al. (Reference Sunstein, Reisch and Kaiser2019) found that left-wing views implied greater support for nudging than right-wing views did, but discovered ‘no systematic correlation along approval and party affiliations’ (p. 1423).Footnote 1
With this article, we hope to deepen these discussions through an empirical investigation. It will explore the relationship between the support for nudging and citizens’ trust in public institutions and their political party preferences.
Sunstein et al. (Reference Sunstein, Reisch and Rauber2018) have commented that Denmark paradoxically stands out as a country whose citizens are trustful of public institutions yet relatively unsupportive of nudging. This article will investigate whether that is also the case in its neighboring country, Sweden, an outlier internationally regarding trust in public institutions.
The latest, sixth wave of the World Value Surveys, 2010–2014, reported Swedes’ trust in public institutions to be the highest among all countries surveyed. For example, 60% of Swedes (N = 1206) had confidence in government, and 59% had confidence in parliament. In comparison, the corresponding proportions among Americans (N = 2232) were 20% and 33%, respectively.
Sweden's liberal strategy to manage the COVID-19 pandemic has received worldwide attention because of its reliance on mutual trust between the government and the citizens, a strategy that the Swedish government has labeled ‘behaviorally informed’ and in which nudging has been involved (Almqvist & Andersson, Reference Almqvist and Andersson2020).
Under Sunstein et al.'s (Reference Sunstein, Reisch and Rauber2018) original hypothesis, a relatively high support for nudging would be expected among Swedes. Otherwise, one would be left with two thirds of a Scandinavian puzzle, indicating that high trust in public institutions does not necessarily imply greater support for nudging on a country level. Such a finding would point toward alternate explanations, or a more complex relationship between the two than suggested in the previous literature.
Nudge units can work rather independently of the government, like the Behavioural Insights Team, now a social purpose company part-owned by the UK Cabinet Office (cf., Halpern, Reference Halpern2015). Some citizens may perceive them as outside experts rather than an extension of public institutions. In this article, we, therefore, analyze the relationship between trust in expert technocracy over government and parliament and the support for nudging, using an item from the World Value Survey.
On the relationship between political party preferences and the support for nudging, its systematicity – or lack thereof – may depend on how the parties are positioned. In today's politics, there are at least two dimensions according to which parties can be positioned. One is economic: the left-right spectrum. One is cultural: the GAL-TAN spectrum (Hooghe et al., Reference Hooghe, Marks and Wilson2002), which contrasts green, alternative, and libertarian (some say postmodern) values with traditional, authoritarian, and nationalistic ones. Through a population-representative survey involving 1032 respondents, this article explores the Swedish multiparty system to consider the possible relationship between political party preferences and the support for nudging through this two-dimensional theoretical framework.
To position the parties across both of these dimensions, we use the results from the Chapel Hill Expert Survey (Polk et al., Reference Polk, Rovny, Bakker, Edwards, Hooghe, Jolly, Koedam, Kostelka, Marks, Schumacher, Steenbergen, Vachudova and Zilovic2017), which contains expert ratings of the parties’ positions in 31 countries (all European Union members, Norway, Switzerland, and Turkey). As of 2020, there are eight parties in the Swedish parliament. The present, center-left government consists of the Social Democrats (S) and the Green Party (MP), supported by the Center Party (C) and the Liberals (L). The opposition parties are the Left Party (V), the Sweden Democrats (SD), the Christian Democrats (KD), and the Moderate Party (M).
Positioning the parties using the economic left-right and cultural GAL-TAN spectra, three of the four quadrants were represented in parliament: left-GAL, right-GAL, and right-TAN parties (see Table 1).
The Swedish political parties’ positions: mean ratings from the Chapel Hill Expert Survey (Polk et al., Reference Polk, Rovny, Bakker, Edwards, Hooghe, Jolly, Koedam, Kostelka, Marks, Schumacher, Steenbergen, Vachudova and Zilovic2017).
a 0 = extreme left; 10 = extreme right.
b 0 = libertarian/postmaterialist (GAL); 10 = traditional/authoritarian (TAN).
Loibl et al. (Reference Loibl, Sunstein, Rauber and Reisch2018) have emphasized the need for more studies on the relationship between socio-demographics and differences in the support for nudging within countries, while Sunstein et al. (Reference Sunstein, Reisch and Kaiser2019) concluded that ‘little evidence – at least outside the U. S. – has yet surfaced about which population groups support nudging and which factors shape those attitudes’ (p. 1422). This article addresses those research gaps and makes an empirical contribution by adding to the relatively few studies on the public support for nudging and its relationship to socio-demographics and political party preferences (e.g., Diepeveen et al., Reference Diepeveen, Ling, Suhrcke, Roland and Marteu2013; Petrescu et al., Reference Petrescu, Hollands, Couturier, Ng and Marteau2016; Reisch & Sunstein, Reference Reisch and Sunstein2016; Loibl et al., Reference Loibl, Sunstein, Rauber and Reisch2018; Sunstein et al., Reference Sunstein, Reisch and Rauber2018; Sunstein et al., Reference Sunstein, Reisch and Kaiser2019). It also adds a 16th country – Sweden – to the catalog of countries surveyed to date using the same instrument.
Method
Respondents
An online survey was distributed by a Swedish market research company. Its target population was Swedes between 18 and 79 years of age. The data were collected in December 2017 and were analyzed using IBM SPSS Statistics, Version 25.0.Footnote 2 The sample included 1032 respondents representative of the target population. Their average age was 47.23 years (SD = 16.64). Controls included gender, age, education, job sector, household income, and living region (see Table 2).
Selected respondent characteristics (N = 1032).
* 1 SEK ≈ 0.1 EUR.
Questionnaire
The survey featured an adapted and translated version of a questionnaire developed by Reisch and Sunstein (Reference Reisch and Sunstein2016). Respondents were first asked whether they were ‘for or against the following hypothetical policy’ and then presented 13 nudges in random order (for the nudges, see Table 3).Footnote 3 Answers were provided on a binary response scale: ‘for’ or ‘against’. In addition, the respondents answered the following World Value Survey question-statement on the same response scale: ‘having experts, not government and parliament, make decisions according to what they think is best for the country’. The respondents were also asked which, if any, political party they would have voted for had it been national elections that same day, a question that 850 (82%) respondents answered by naming a party represented in parliament, while 182 (18%) respondents were undecided or preferred other parties.
Percentage support and 95% confidence intervals per nudge (N = 1032).
a Reisch et al. (Reference Reisch, Sunstein and Gwozdz2017) have suggested that higher (lower) smoking prevalence would imply lower (higher) support for nudge 12. Sweden has the lowest smoking prevalence in the European Union according to the European health interview survey, but its support for nudge 12 remains quite low.
Results
Descriptive statistics
On average, the respondents supported 7.44 (SD = 3.21, 95% CI = 7.25, 7.64) out of 13 nudges. The most popular nudge – approved of by 87%, 95% CI [0.85, 0.89] – was an information campaign against child obesity. The least popular nudge – approved of only by 18%, 95% CI [0.15, 0.20] – was a default donation to charity upon tax return (see Table 3). The average percentage support across the nudges was 57%, which, together with a relatively low number of supported nudges per respondent, places Sweden among the cautiously pronudge nations, with an average support on a par with that of Japan, Hungary, and Denmark, but lower than that of the other countries surveyed to date (cf., Sunstein et al., Reference Sunstein, Reisch and Rauber2018).
Socio-demographics
There was a gender difference in the number of supported nudges [t(1030) = 6.01, p < 0.001] with higher support among women (M = 8.03, SD = 2.93) than men (M = 6.87, SD = 3.37), 95% CI [0.79, 1.57]. There was no difference in the number of nudges approved of across annual household incomes [F(9, 1022) = 1.56, p = 0.12, ns], but a very weak positive correlation with age [r(1030) = 0.063, p = 0.042]. No differences in the number of nudges approved of were found across occupations [F(8, 1023) = 1.65, p = 0.107, ns], nor were there any between respondents living in major or large cities alternatively on the countryside [F(2, 1029) = 0.60, p = 0.548, ns] or across regions [F(7, 1024) = 0.38, p = 0.914, ns].
Political party preferences
The number of nudges approved of differed across political party preferences [F(9, 1022) = 7.649, p < 0.001]. There were differences between the supporters of left-GAL, right-GAL, and right-TAN parties in how many nudges they supported [F(2, 847) = 29.77, p < 0.001]. The support for nudging was the highest among those who supported left-GAL parties (M = 8.47, SD = 2.88, 95% CI = 8.15, 8.79), followed by right-GAL (M = 7.43, SD = 2.90, 95% CI = 6.99, 7.88), and right-TAN (M = 6.66, SD = 3.25, 95% CI = 6.33, 6.99) parties. Bonferroni post hoc tests confirmed that left-GAL sympathizers supported significantly more nudges than those who supported right-GAL, 95% CI [0.33, 1.74] or right-TAN parties, 95% CI [1.25, 2.37], while those who favored right-GAL parties approved of more nudges than did those favoring right-TAN parties, 95% CI [1.46, 0.09].
A contingency table analysis across individual nudges – comparing observed and expected observations per political party preference – revealed significant differences in support for 8 of the 13 nudges (numbers 3, 4, 5, 6, 7, 10, 11, and 13). These were calorie labels, χ 2(7, N = 850) = 15.23; sweet-free cashier zones, χ 2(7, N = 850) = 30.01; warning labels for salt, χ 2(7, N = 850) = 22.04; encouraging, χ 2(7, N = 850) = 72.68 – or requiring, χ 2(7, N = 850) = 71.09 – defaults for green energy; promoting healthy groceries, χ 2(7, N = 850) = 31.51; the carbon emission fee, χ 2(7, N = 850)= 54.11; and the default donation to the Red Cross, χ 2(7, N = 850) = 15.21.
There were no significant differences for the remaining five nudges (numbers 1, 2, 8, 9, and 12): education on child obesity, χ 2(7, N = 850) = 9.33, ns; warning pictures against distracted driving, χ 2(7, N = 850) = 2.04, ns; traffic lights for healthy food, χ 2(7, N = 850) = 12.16, ns; the active choice on organ donations, χ 2(7, N = 850) = 8.36, ns; and the antismoking and -overeating campaign, χ 2(7, N = 850) = 6.82, ns. The largest deviations from what would have been expected under the null hypothesis came from the supporters of V (left-GAL, more positive) and M (right-TAN, more negative), followed by S and MP (both left-GAL, both positive) and SD (right-TAN, negative), echoing the general differences within our two-dimensional framework (Table 4).
Percentage support (standard errors) across nudges and political party preferences (N = 850).
* χ 2(7, N = 850) > 14.07; p < 0.05.
Expert technocracy
As many as 421 (40%) respondents supported ‘having experts, not government and parliament, make decisions according to what they think is best for the country’ (for a discussion, see Foa & Mounk, Reference Foa and Mounk2016). Those who supported expert technocracy approved of more nudges (M = 7.84, SD = 3.22) than those who did not (M = 7.17, SD = 3.18), t(1030) = 3.29, p < 0.001, 95% CI [0.27, 1.07].
Multivariate analyses
We proceeded with 13 logistic regressions; one per nudge. Like Diepeveen et al. (Reference Diepeveen, Ling, Suhrcke, Roland and Marteu2013) and Sunstein et al. (Reference Sunstein, Reisch and Kaiser2019) before us, but unlike Petrescu et al. (Reference Petrescu, Hollands, Couturier, Ng and Marteau2016), a gender difference was found for several nudges. In comparison with Loibl et al. (Reference Loibl, Sunstein, Rauber and Reisch2018), a similarity between our results and theirs was the absence of a relationship between living regions and support for the nudges; a difference is that we failed to reproduce the differences across income levels found in their study.
Individual political party preferences were significant predictors of the support for nudging for 10 out of 13 nudges, driven by the left-GAL parties, contributing with 17 out of 29 such significant predictors. What also stands out in our logistic regressions is that the support for expert technocracy emerges as a significant predictor for 10 out of 13 nudges (see Table 5).
Logistic regressions: odds ratios per predictor and nudge (N = 1032).
Note: Baseline categories: undecided voters, nonresponse on annual income, other job sectors, and residents of upper Norrland.
* p < 0.05, **p < 0.01, ***p < 0.001.
Following Loibl et al. (Reference Loibl, Sunstein, Rauber and Reisch2018), we also derived a controversy proxy from the number of significant predictors per nudge. Across nudges, an average of 5.46 out of 31 (18%) independent variables were statistically significant.Footnote 4 The most controversial nudge turned out to be the sixth and seventh most popular ones: encouraging, alternatively mandating, and consumers into a green energy default. Controversy proved unrelated to the nudges’ relative popularity, ρ = −0.031, p = 0.921, ns.
We next introduced a subcategory of controversy: politicization, the number of political party preferences as significant predictors, and found that the two most controversial nudges were also the two most politicized ones. Politicization was unrelated to the nudges’ relative popularity too, ρ = 0.031, p = 0.920, ns.
As nudges 6 and 7 constituted two thirds of the green nudges (number 11 was the third), we compared the green nudges with the others in terms of controversy and politicization. They were indeed significantly more controversial χ 2(1, N = 71) = 5.90, p = 0.015 and politicized χ 2(1, N = 31) = 5.48, p = 0.019 than the others.
Conclusions
This article has investigated the public support for nudging and its relationship with socio-demographics and political party preferences, a research gap recently identified in the literature. One of its findings relates to the relationship between trust in public institutions and the support for nudging. While there is support for nudging in Sweden – a country with extraordinary trust in public institutions – it turns out to be relatively low (a cautiously pronudge nation), contrary to what previous research has theorized. A new independent variable – the support for expert technocracy over government and parliament – emerged as a significant predictor of support for most individual nudges. These findings complicate the relationship between trust in public institutions and the support for nudging. On a country level, the empirical literature now includes two Scandinavian countries (Sweden and Denmark) where trust in public institutions is very high but the support for nudging is relatively low; and on an individual level, a new predictor that formally runs counter to trust in public institutions.
Another finding relates to the relationship between political party preferences and the support for nudging. Both parametric and nonparametric univariate analyses found significant associations between the support for nudging and political party preferences across the economic left-right and cultural GAL-TAN spectra. Logistic regressions revealed that at least a couple of individual political party preferences at the time, especially for left-GAL parties, were significant predictors of the support for 10 out of 13 nudges. Going forward, we believe the two-dimensional framework used in this article could be a promising way to systemize differences in the support for nudging across political party preferences.
Like Loibl et al. (Reference Loibl, Sunstein, Rauber and Reisch2018) before us, we speculate that the domestic political debate may have contributed to the individual nudges’ controversy and politicization. Green nudges were most controversial and politicized in terms of polarization across socio-demographics and political party preferences. Consistently, environmental issues have been high on the policy agenda in Sweden for the past several years. It is also known from prior research that nudges’ support is susceptible to affiliation with the choice architect (Tannenbaum et al., Reference Tannenbaum, Fox and Rogers2017). Accordingly, nudges’ polarization, politicization, and support are not static. Surveys of their support are snapshots of moving targets. Much work remains to be done.
Open access
