Acceptance of Inequality Between Children: Large-Scale Experimental Evidence from China and Norway

Abstract

In a novel large-scale experiment, we study how adults in two societies, Shanghai (China) and Norway, make real distributive decisions involving children. We find that acceptance of inequality between children increases with the ages of the children, is affected by the source of inequality and the cost of redistribution, and is lower than acceptance of inequality between adults. We document a large cross-societal difference in inequality acceptance: adults in Shanghai implement twice as much inequality between children compared with adults in Norway. Finally, we show that the willingness to accept inequality between children is predictive of attitudes to child policies.

A large literature on social preferences has studied attitudes to inequality between adults (Konow, 2000; Andreoni and Miller, 2002; Charness and Rabin, 2002; Engelmann and Strobel, 2006; Cappelen et al., 2007; 2013; 2020a; Bellemare et al., 2008; Falk et al., 2008; Balafoutas et al., 2013; Falk and Szech, 2013; Durante et al., 2014; Cappelen and Tungodden, 2019), but not attitudes to inequality between children. Understanding attitudes to inequality between children, however, is paramount, given the far-reaching consequences of child policies, influencing the distribution of resources among children and determining the trajectory of the most disadvantaged children in terms of income, education, health and life satisfaction (Heckman, 2013; Heckman et al., 2013; Conti and Heckman, 2014; García et al., 2020; Zhou et al., 2021).

There are reasons to expect that people may have different attitudes to inequality between children than to inequality between adults. Inequality between adults is often seen as justifiable if it reflects differences in merit or if there is a cost of redistribution (Almås et al., 2020), but it is arguably more difficult to justify inequality between children because children are less developed cognitively and because their circumstances are largely beyond their control. Attitudes to childhood inequality may also be shaped by a concern for the later life consequences of childhood inequality and by a desire to prepare the children for adult life. It is therefore of great interest to study whether, and in which economic environments, people accept inequality between children.

This paper aims to examine what the adult populations in two societies, Shanghai and Norway, view as acceptable inequality between children. To study this question, we implement a novel experimental design where a large sample of adults from the general populations in the two societies decide how to handle an income inequality between two children. Specifically, the adults decide whether to redistribute money between two children who have completed the same assignment, but have unequal earnings (Konow, 2000; Cappelen et al., 2013; Almås et al., 2020). In a between-subject design, we vary the ages of the children in the pair, which allows us to study the causal effect of the children’s age on the adults’ willingness to accept inequality. Furthermore, we vary two key dimensions of the economic environment: the source of inequality and the cost of redistribution. This allows us to study how the acceptance of inequality between children causally depends on the economic environment. In a follow-up study, we examine how attitudes to inequality between children (seventeen-year-olds) compare to attitudes to inequality between adults in the two societies.

The main contribution of the present study is to provide the first set of experimental evidence on how general populations consider inequality between children. In both societies, we find that acceptance of inequality between children increases with the age of the child, and is greater when the source of inequality is merit rather than luck and when there is a cost of redistribution. However, we find that the source of inequality is more important for adults in Norway than for adults in Shanghai, and that adults in Norway, but not adults in Shanghai, become increasingly sensitive to the source of inequality as the children get older. We further show that the willingness to accept inequality between children is highly predictive of support for child policies targeting low-performing children in both societies.

Strikingly, we find that adults in Shanghai implement twice as much income inequality between children as do adults in Norway, when making consequential choices in the same distributive situations. The cross-societal difference in acceptance of inequality between children is robust across subgroups, to varying the ages of the children and key features of the economic environment, and to controlling for background characteristics on education, income, age, gender and parental status. Even for children as young as five years old, adults in Shanghai are willing to accept large income inequality, while adults in Norway overwhelmingly choose to equalise incomes for this age group. We provide suggestive evidence of the large cross-societal difference in inequality acceptance partly reflecting that adults in Shanghai find it more important than adults in Norway to teach children that life is not always fair. We find much smaller differences in the acceptance of childhood inequality between subgroups within societies, which suggests that general cultural factors play a significant role in shaping attitudes towards childhood inequality.

Finally, in the follow-up study, we show that there is less acceptance of inequality between seventeen-year-olds than between adults when the source of inequality is luck, but not when the source of inequality is merit and only in Norway when there is a cost of redistribution. The implemented inequality in the Luck treatment for two adults is about 70% higher than the implemented inequality for two five-year-olds in Norway in the main study, and about 15% higher in Shanghai. Hence, we find strong evidence of inequality due to luck being more acceptable between adults than between children. Overall, we find that the cross-societal difference in inequality acceptance is greater for children than for adults. However, even for inequality between adults, we find a large cross-societal difference in inequality acceptance, which suggests that there are general cultural differences in inequality acceptance that contribute to shape attitudes towards childhood inequality.

Our study relates to several literatures. First, it provides novel evidence to the large experimental literature on social preferences (Konow, 2000; Andreoni and Miller, 2002; Charness and Rabin, 2002; Engelmann and Strobel, 2006; Cappelen et al., 2007; 2013; Bellemare et al., 2008; Falk et al., 2008; Balafoutas et al., 2013; Falk and Szech, 2013; Durante et al., 2014; Cappelen and Tungodden, 2019; Falch, 2022), by being the first study of what adults view as acceptable inequality between children and of how inequality acceptance depends on the ages of the children. We also enrich this literature by documenting how the willingness to accept inequality between children depends on the source of inequality and the cost of redistribution, how it differs from the acceptance of inequality between adults and whether it is predictive of people’s views on child policies.

The paper further contributes to the growing literature on international differences in attitudes towards inequality and redistribution, which has relied on both non-incentivised surveys (Svallfors, 1997; Edlund, 1999; Fong, 2001; Alesina and Glaeser, 2004; Osberg and Smeeding, 2006; Alesina and Giuliano, 2011; Luttmer and Singhal, 2011; Aarøe and Petersen, 2014; Ashok et al., 2015; Falk et al., 2018) and incentivised experiments (Henrich et al., 2010; Cappelen et al., 2015; Jakiela, 2015; Barrett et al., 2016; Grimalda et al., 2018; Almås et al., 2020). We add to this literature by documenting a large cross-societal difference in adults’ acceptance of inequality between children, even when the adults consider identical distributive situations. We also provide suggestive evidence of the cross-societal difference in inequality acceptance, reflecting both general cultural differences in inequality acceptance and differences in the importance assigned to children learning that life is not always fair.

Finally, our study speaks to the literature on moral development in childhood and adolescence (Piaget, 1965; Kohlberg, 1984; Fehr et al., 2008; 2013; Almås et al., 2010; Bauer et al., 2014; 2016; Sutter et al., 2019; Cappelen et al., 2020b) and the literature on the cultural transmission of preferences across generations (Henrich et al., 2010; Dohmen et al., 2011; 2012; House et al., 2013; Blake et al., 2015; 2016; Cowell and Decety, 2015; Schäfer et al., 2015; Ben-Ner et al., 2017; Falk et al., 2021; Brenøe and Epper, 2022; Chowdhury et al., 2022; Tungodden and Willén, 2023). How adults deal with distributive situations involving children is likely to be important, both for how inequality acceptance develops with age, and for the cultural transmission of inequality acceptance across generations. Previous studies have identified positive associations in economic preferences between parents and their children, including for risk, time and social preferences in Bangladesh (Chowdhury et al., 2022), risk preferences in Germany (Dohmen et al., 2012) and competitiveness in Norway (Tungodden and Willén, 2023). We extend this literature by providing novel evidence on how unrelated adults from general populations in two societies make distributive decisions involving children, which also relates to the growing literature in economics that studies how the values of the older generation is partly transmitted to the younger generation through observing the behaviour of unrelated adults, so-called oblique socialisation (Bowles, 1998; Bowles and Gintis, 2000; Bisin and Verdier, 2001; Montgomery, 2010; Bisin and Verdier, 2011; Gintis, 2011; van Leeuwen et al., 2018; Le Garrec, 2018; Roth and Wohlfart, 2018; Besley and Persson, 2019). The present paper thus highlights the importance of studying empirically different types of socialisation processes in society, both purposeful parental socialisation and more indirect oblique socialisation.

The paper proceeds as follows. Section 1 describes the experimental design and the sample, and Section 2 outlines the empirical strategy. Section 3 reports the main results, and Section 4 discusses mechanisms that may explain the cross-societal differences. Section 5 reports from the follow-up study that compares how people consider inequality between children versus inequality between adults. Section 6 concludes.

1. Experimental Design

There are two types of participants in the experiment: adults and children. The adults were recruited through two data-collection agencies, Opinion Research Shanghai in Shanghai and Norstat in Norway. Each adult was asked to make a distributive decision in situations involving two children, building on the design used in Almås et al. (2020). We recruited 3,000 adult residents in Shanghai (China) and 3,014 adult residents in Norway, with both groups constituting a general population sample (⁠|$+$|18 years old) that is representative on a set of observable characteristics. In Online Appendix Table A1 we provide an overview of the background characteristics of the adult participants, and in Online Appendix B.2 we provide a more detailed description of the samples. We specified the empirical strategy in two pre-analysis plans, one for Shanghai and one for Norway, registered in the AEA RCT Registry (AEARCTR-0002627 and AEARCTR-0003112).

To create real distributive situations for each of the adults in the experiment, we recruited 3,714 children from preschools and schools in Shanghai and Norway, who performed assignments that involved completing an age-appropriate real-effort task (instructions to the children are described in Online Appendix C.1.4). On average, each child completed three assignments and, for each assignment, they were randomly matched with another child of the same age. One child in each pair initially earned a sum of money as a bonus for completing the assignment (CNY 24 (USD 3.5) in Shanghai and NOK 48 (USD 6) in Norway), while the other child earned nothing. The earnings in the two societies were roughly similar in terms of purchasing parity. The children were only informed about the bonus and how it would be determined after they had completed the task.

As illustrated in Online Appendix Figure A1, each adult was randomly matched with a unique pair of children in their society and asked to decide whether to redistribute from the child with initial earnings to the child without initial earnings (complete instructions are provided in Online Appendices C.1.1–C.1.3). The adults were informed that the children had not been told about their initial earnings. The adult’s decision determined the payment of the two children in the matched pair of children.

The treatments varied in terms of the ages of the children and key features of the economic environment in a between-individual design. The adults were randomly assigned to decide for a pair of children who were either five years old, nine years old, thirteen years old or seventeen years old (⁠|$n=\,$|1,494, 1,499, 1,512 and 1,509). They were further randomly assigned to one of three distributive situations, referred to as Luck, Merit and Efficiency (⁠|$n=$| 2,000, 2,009 and 2,005).

In the Luck situation, a random draw determines which child gets the initial earnings and which child gets no initial earnings. Furthermore, there is no cost of redistribution, i.e., for each unit transferred from the child with initial earnings, one unit is received by the child without initial earnings. In Shanghai, the adults could decide between not redistributing or redistributing CNY 4, CNY 8, CNY 12, CNY 16, CNY 20 or everything to the child with no earnings, with a corresponding choice set in Norway.

The Merit situation is identical to the Luck situation, except that the earnings are determined by the productivity of the children on the task: the most productive child in the pair earns all the money. Hence, by comparing the Merit situation and the Luck situation, we can identify how the source of inequality matters for the acceptance of inequality between children. The Efficiency situation is identical to the Luck situation, except that redistribution is costly: for each unit redistributed, the income to the child with initial earnings is reduced by two units. The cost of redistribution implies that there is an efficiency loss of one-third of the total earnings if the adult chooses to equalise the incomes.

After the adults made their distributive choices, they were asked a set of survey questions, including questions about their policy attitudes and whether they think it is important for children to learn that life is not always fair. By linking the answers to these survey questions to the behavioural data from the experiment, we can shed light on the adults’ motivations for their distributive decision and how their acceptance of inequality between children in the experiment relates to their views on child policies. At the end, the adults completed a set of background questions. As shown in Online Appendix Tables A2 and A3, the samples in the two societies are balanced across treatments with respect to these background characteristics.

2. Empirical Strategy

The main empirical analysis consists of three parts. We study cross-societal differences in acceptance of inequality between children, the casual effects of the ages of the children and key features of the economic environment (source of inequality and cost of redistribution) on inequality acceptance, and the interaction between the ages of the children and the economic environment.

In the analysis, we use two measures of an adult’s inequality acceptance |$u_i$|⁠. First, we measure the inequality implemented by adult |$i$|⁠:

$$\begin{eqnarray} u_i= \frac{|\it{Income Child } A_i - \it{Income Child } B_i|}{\it{Total Income}}\in [0,1]. \end{eqnarray}$$

Here |$\it{Income Child } A_i$| is the income allocated to the child with initial earnings and |$\it{Income}$|  |$\it{Child } B_i$| is the income allocated to the child without initial earnings. This inequality measure is equivalent to the Gini coefficient in a two-person situation. It is equal to one if the adult decides to transfer nothing to the child without initial earnings and zero if the adult decides to equalise the incomes between the two children. Second, we measure inequality acceptance as an indicator variable for whether the adult decides to equalise the income of the two children, i.e., whether the adult is not willing to accept any inequality between them.

We start out by comparing inequality acceptance between the two societies, pooled across treatments. In this analysis, we report the regression specification

$$\begin{eqnarray} u_i=\alpha +\beta \it{Norway}_{i}+\gamma {\bf x}_{i}+\epsilon _{i} , \end{eqnarray}$$

where |$\it{Norway}_i$| is an indicator for the adult being from the Norwegian sample and |${\bf x}_i$| is a vector of background variables that includes indicators for gender, age, income, education and having children. The estimated value of |$\beta$| provides us with a test of whether there is more or less inequality acceptance in Norway than in Shanghai across all treatments in the study. We report this regression analysis with and without background variables and separately for all the pre-specified subgroups in society. Finally, we report a regression specification that compares inequality acceptance between adults in Shanghai with and without a Shanghai hukou (‘citizenship’) at birth.

In the analysis of the causal effect of the ages of the children on the acceptance of inequality, we report the regression specification

$$\begin{eqnarray} u_i=\alpha +\beta _{1} \it{Norway}_{i}+\beta _{2} 9y_{i}+\beta _{3} 13y_{i}+\beta _{4} 17y_{i}+\gamma {\bf x}_{i}+\epsilon _{i}, \end{eqnarray}$$

where |$9y_i$|⁠, |$13y_i$| and |$17y_i$| are treatment indicators for the adult distributing between nine-year-olds, thirteen-year-olds or seventeen-year-olds, with five-year-olds as the baseline treatment. The estimated values of |$\beta _{2}$|⁠, |$\beta _{3}$| and |$\beta _{4}$| thus provide us with the causal effects of varying the ages of the children.

Correspondingly, in the analysis of the causal effect of key features of the economic environment on the acceptance of inequality between children, we report the regression specification

$$\begin{eqnarray} u_i=\alpha +\beta _{1} \it{Norway}_{i}+\beta _{2} \it{Merit}_{i}+\beta _{3} \it{Efficiency}_{i}+\gamma {\bf x}_{i}+\epsilon _{i}, \end{eqnarray}$$

where |$\it{Merit}_i$| and |$\it{Efficiency}_i$| are treatment indicators for the adult being in a situation where the source of inequality is merit or in a situation with a cost of redistribution, with the Luck treatment as the baseline treatment. The estimated values of |$\beta _{2}$| and |$\beta _{3}$| provide us with the causal effects of varying the source of inequality and the cost of redistribution.

Throughout the analysis, we always also report the specifications separately for each society. To study how inequality acceptance relates to background characteristics in each society, pooled across treatments, we report estimates from regression specifications where the dependent variable is implemented inequality or an indicator variable for whether the adult decides to equalise and the independent variables are the background characteristics. For the analysis of how the ages of the children and key features of the economic environment affect inequality acceptance, we further report the estimates for regression specifications that interact the society dummy with the treatment indicators. Finally, we report results from separate regressions for Shanghai and Norway that include treatment indicators, both for the ages of the children and the economic environment, and interaction effects between these treatment indicators.

In the Online Appendix, we report p-values adjusted for multiple hypothesis testing. We calculate unadjusted p-values as bootstrap p-values following Davison and Hinkley (1997) and compute p-values adjusted for step-down multiple testing following the algorithm proposed by Romano and Wolf (2016).

3. Results

We here provide the main findings of our study, with supplementary analysis reported in the Online Appendix.

3.1. Inequality Acceptance

Figure 1 reports acceptance of inequality between children in Shanghai and Norway, pooled across all treatments.¹ From Figure 1(a), we observe that there is a large difference in the average implemented inequality in the two societies: 0.54 in Shanghai versus 0.26 in Norway (⁠|$p\lt .001$|⁠). Hence, adults in Shanghai are willing to accept twice as much inequality between children as Norwegian adults when making decisions in distributive situations that are identical in terms of the ages of the children, the source of the inequality and the cost of redistribution. Notably, as indicated in Figure 1, the difference in implemented inequality between the children is close to the difference in the disposable income inequality in the two societies.

Fig. 1.

Inequality Acceptance in Shanghai and Norway.

Note: Panel (a) shows the implemented inequality across all treatments and age groups in the experiment, by society (Shanghai (n = 3,000) and Norway (n = 3,014)). The horizontal lines indicate the disposable income Gini coefficients for China (0.51) and for Norway (0.26) (OECD, 2018). There is a lack of statistics on income inequality in Shanghai, but there is reason to believe that the income inequality in Shanghai is even larger than in China as a whole (Chen et al., 2018). Panel (b) shows the share of adults who chose to equalise the income of the two children across all treatments and age groups, by society. SEs are indicated by bars.

In Figure 1(b), we show that the societal difference in acceptance of inequality between children is large also when comparing the share of adults that are not willing to accept any inequality between the children: 18.7% in Shanghai versus 61.6% in Norway (⁠|$p\lt .001$|⁠). In Online Appendix Table A4, we show that these societal differences are robust to controlling for background characteristics on education, income, age, gender and parental status. We can thus state our first main result.

 

Result 1.

Acceptance of inequality between children differs significantly between Shanghai and Norway. Adults in Shanghai implement twice as much inequality between children as do adults in Norway. Only a small minority of adults in Shanghai are not willing to implement any inequality between the children, while a majority of adults in Norway implement an equal split.

The cross-societal differences in inequality acceptance are strikingly robust across subgroups, as shown in Online Appendix Figure A4 and Table A5. In all subgroups, we find that adults in Shanghai implement significantly more inequality between the children than Norwegian adults.

Shanghai is the top metropolitan area for migration in China, with a large fraction of first-generation residents who have moved to Shanghai because they have excelled in their education or career or simply because they seek out greater opportunities and a better life. Hence, it is interesting to study whether the observed differences between Shanghai and Norway are largely driven by these first-generation residents being particularly accepting of inequality between children. In Online Appendix Figure A5 and Table A6, we compare the subgroup of adults with a Shanghai hukou (‘citizenship’) at birth and the subgroup of adults with a non-Shanghai hukou at birth. We observe that the adults with a non-Shanghai hukou at birth indeed accept more inequality between the children than the adults with a Shanghai hukou at birth, but the difference in implemented inequality is only 12% of the difference between Shanghai and Norway. All our main results are robust to comparing the general population in Norway to each of these two subgroups in Shanghai.²

Finally, in Figure 2, we report the associations between a set of background characteristics and implemented inequality, pooled across treatments, in Shanghai (panels (a) and (c)) and Norway (panels (b) and (d)).³ In both societies, we find evidence of females implementing less inequality than males. In Shanghai, low educated people implement slightly less inequality, while low-income and younger people implement slightly more inequality. These background variables do not seem to play an important role for the acceptance of childhood inequality in Norway. Parental status has opposite effects in the two societies. Having children increases inequality acceptance in Shanghai, while it reduces inequality acceptance in Norway. Finally, political orientation, only asked in Norway, is the strongest predicator of acceptance of inequality between children. People who self-report that they would vote for the right-wing parties are much more likely to accept inequality between children.⁴ However, taken together, this analysis shows that the within-society differences in inequality acceptance are rather small compared to the cross-society difference, which suggests that general cultural factors play a significant role in shaping attitudes towards childhood inequality.

Fig. 2.

Inequality Acceptance: Within-Society Heterogeneity.

Note: The figure shows the correlates of implemented inequality in panels (a)–(b) and share equalising in panels (c)–(d), by society (Shanghai and Norway). Female is an indicator for the adult being female. Low education is an indicator for the adult having an education below the median in the sample (high school or less in both Shanghai and Norway), Low income is an indicator for the adult having an income below the median gross household income per year in the sample (USD 16,000 in Shanghai, USD 80,000 in Norway), Low age is an indicator for the adult being below the median age in the sample (forty-three years in Shanghai, fifty years in Norway), Have child(ren) is an indicator variable for whether the adult has at least one child, Right wing (only reported in Norway) is an indicator variable for whether the adult self-reports that they would vote for one of the right-wing parties. Adults with missing information on education or income are included in the above-median subgroup in the main analysis, but the results are robust to excluding these adults. See Online Appendix Table A1 for further sample descriptives. Lines indicate 95% confidence intervals.

3.2. The Causal Effect of the Children’s Ages

In Figure 3(a) and Table 1, we report how adults’ acceptance of inequality between children is causally affected by the age of the child (five, nine, thirteen or seventeen years), pooled across the different economic environments. In both Shanghai and Norway, we find evidence of adults being more accepting of inequality between children as they mature and approach adulthood. In particular, we observe that adults in both societies implement significantly more inequality between seventeen-year-olds than between five-year-olds (Shanghai, |$p=.005$|⁠; Norway, |$p\lt .001$|⁠). This finding is robust across subgroups in both societies; see Online Appendix Figure A7. We also observe that the share who do not accept any inequality between the children is lower for seventeen-year-olds than for five-year-olds, but this difference is only significant in Norway (Shanghai, |$p=.192$|⁠; Norway, |$p\lt .001$|⁠). In Table 1, we show that these findings are robust to controlling for background variables.

Fig. 3.

Inequality Acceptance in Shanghai and Norway, by Treatment.

Note: Panel (a) shows the mean implemented inequality (left panel) and the share equalising (right panel) by age group and society (Shanghai and Norway). Panel (b) shows the merit treatment effect, and the efficiency treatment effect on the mean implemented inequality (left panel) and the share equalising (right panel) by society. Panel (c) shows the interaction effect between the source of inequality/cost of redistribution and age of the child (for the comparison of inequality between seventeen-year-old children and five-year-old children), on the implemented inequality (left panel) and the share equalising (right panel). Estimates are based on the full sample with control variables. Controls are defined in Figure 2. SEs are indicated by bars.

Table 1.

Inequality Acceptance: Ages of the Children.

	Inequality			Share equalising
	Shanghai	Norway	Pooled	Shanghai	Norway	Pooled
	(1)	(2)	(3)	(4)	(5)	(6)
Norway			−0.281***			0.434***
			(0.010)			(0.011)
9 years	−0.003	0.036*	0.018	0.027	−0.093***	−0.035**
	(0.018)	(0.019)	(0.013)	(0.021)	(0.025)	(0.016)
13 years	0.019	0.073***	0.047***	0.005	−0.111***	−0.054***
	(0.018)	(0.020)	(0.013)	(0.020)	(0.024)	(0.016)
17 years	0.051***	0.071***	0.061***	−0.025	−0.113***	−0.070***
	(0.018)	(0.020)	(0.013)	(0.019)	(0.025)	(0.016)
Constant	0.489***	0.295***	0.541***	0.167***	0.599***	0.163***
	(0.025)	(0.024)	(0.019)	(0.028)	(0.029)	(0.022)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3,000	3,014	6,014	3,000	3,014	6,014
|$R^{2}$|	0.017	0.016	0.134	0.012	0.021	0.200

Note: The table reports OLS regressions with implemented inequality as the dependent variable in columns (1)–(3) and with the share that equalises as the dependent variable in columns (4)–(6). The reference category across all regressions is the Luck treatment with five-year-old children. Norway is an indicator for the adult being from the Norwegian sample. 9 years, 13 years and 17 years are indicators for the children in the pair being nine years old, thirteen years old and seventeen years old. Controls are indicators for gender, low education, low income, low age and having a child, as defined in Figure 2. Robust SEs are reported in parentheses, where * p  |$\lt $| .10, ** p  |$\lt $| .05, *** p  |$\lt $| .01; p-values adjusted for multiple hypothesis testing are provided in Online Appendix Table A11.

The causal effect of the children’s ages is small compared with the societal effect: the difference in implemented inequality between Shanghai and Norway, pooled across treatments, is about four times the difference in implemented inequality between the five-year-olds and the seventeen-year-olds in Norway and more than five times this age gradient in Shanghai.

We further observe from Figure 3(a) and from Table 1 that there is significantly more inequality acceptance between adults in Shanghai than in Norway already for the youngest children (⁠|$p\lt .001$|⁠). There is also no general convergence in the acceptance of inequality between children with age. The difference in implemented inequality between Shanghai and Norway between five-year-olds is virtually the same as between seventeen-year-olds: 0.309 (five-year-olds) versus 0.289 (seventeen-year-olds) (⁠|$p=.473$|⁠).

We summarise this analysis in the following result.

 

Result 2.

Acceptance of inequality between children increases with the age of the child in both societies. The difference in acceptance of inequality between children in Shanghai and Norway is large already in early childhood and lasts throughout childhood and adolescence.

3.3. The Causal Effect of the Economic Environment

In Figure 3(b) and Table 2, we report how adults’ acceptance of inequality between children is causally affected by key features of the economic environment. We find some evidence of the source of inequality mattering for whether adults accept inequality between children, with the effect being stronger in Norway than in Shanghai. In Norway, the estimated causal effect of the source of inequality being merit rather than luck (merit treatment effect) on implemented inequality is 0.057 (⁠|$p=.001$|⁠), which implies that adults implement almost 20% more inequality when the source of inequality between children is merit rather than luck. We further observe that the share of adults in Norway that implements some inequality between the children is 19.4 percentage points higher in the Merit treatment than in the Luck treatment (⁠|$p\lt .001$|⁠). In Shanghai, we also observe an increase in inequality acceptance when the source of inequality is merit rather than luck, but the effect is more muted: implemented inequality increases with 0.031 (⁠|$p=.054$|⁠) and the share of adults implementing some inequality increases by 5.1 percentage points (⁠|$p=.005$|⁠).

Table 2.

Inequality Acceptance: Economic Environment.

	Inequality			Share equalising
	Shanghai	Norway	Pooled	Shanghai	Norway	Pooled
	(1)	(2)	(3)	(4)	(5)	(6)
Norway			−0.281***			0.434***
			(0.010)			(0.011)
Merit	0.031*	0.057***	0.044***	−0.051***	−0.194***	−0.122***
	(0.016)	(0.016)	(0.011)	(0.018)	(0.021)	(0.014)
Efficiency	0.109***	0.066***	0.087***	−0.088***	−0.083***	−0.085***
	(0.016)	(0.018)	(0.012)	(0.017)	(0.021)	(0.014)
Constant	0.459***	0.300***	0.529***	0.216***	0.614***	0.194***
	(0.025)	(0.023)	(0.018)	(0.028)	(0.028)	(0.021)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3,000	3,014	6,014	3,000	3,014	6,014
|$R^{2}$|	0.030	0.016	0.138	0.018	0.039	0.208

Note: The table reports OLS regressions with implemented inequality as the dependent variable in columns (1)–(3) and with the share that equalises as the dependent variable in columns (4)–(6). The reference category across all regressions is the pooled Luck treatment. Norway is an indicator for the adult being from the Norwegian sample. Merit is an indicator for the adult being in the Merit situation and Efficiency is an indicator for the adult being in the situation with a cost of redistribution. Controls are indicators for gender, low education, low income, low age and having a child, as defined in Figure 2. Robust SEs are reported in parentheses, where * p  |$\lt $| .10, *** p  |$\lt $| .01; p-values adjusted for multiple hypothesis testing are provided in Online Appendix Table A12.

The introduction of a cost of redistribution significantly increases inequality acceptance in both societies, which shows that adults make a trade-off between fairness and efficiency when considering childhood inequality. In both societies, the average implemented inequality is significantly higher in the Efficiency treatment than in the Luck treatment (in both cases, |$p\lt .001$|⁠), and correspondingly, the share of adults who do not accept any inequality between the children falls by 8.8 percentage points in Shanghai and 8.3 percentage points in Norway (in both cases, |$p\lt .001$|⁠).

In Table 2, we show that the findings on the effect of the economic environment on inequality acceptance are robust to controlling for background variables, and in Online Appendix Figure A7 that they are robust across subgroups within both Shanghai and Norway.

Taken together, we state the following result.

 

Result 3.

Acceptance of inequality between children is greater when the source of inequality is merit rather than luck and when there is a cost of redistribution. The difference in acceptance of inequality between children in Shanghai and Norway is robust across different economic environments.

3.4. The Interaction between the Ages of the Children and the Economic Environment

In Figure 3(c), we report the interaction effects between the ages of the children and the economic environment, focusing on a comparison of the treatment effects for five-year-olds and seventeen-year-olds. In Online Appendix Table A7, we report the full regression analysis interacting all age groups and the different economic environments. The general impression from this analysis is that the treatment effects on inequality acceptance of varying the source of inequality and the cost of redistribution do not differ significantly across age groups.

An interesting exception is the effect of the source of inequality in Norway. As shown in Figure 3(c), we find a significantly larger estimated merit effect for seventeen-year-olds than for five-year-olds in Norway when considering the share of adults equalising (⁠|$p=.010$|⁠). However, even for five-year-olds, as shown in Online Appendix Table A7, there is a significant merit effect. Taken together, the evidence suggests that the distinction between inequality due to merit and inequality due to luck is important throughout childhood in Norway and becomes increasingly important as the children mature. In Shanghai, we do not find any difference in the merit treatment effect between five-year-olds and seventeen-year olds, and the merit treatment effect is not significant for any of the two age groups.

Finally, in Figure 3(c), we observe that the interaction effect between age and the cost of redistribution is not statistically significant in Shanghai or Norway. As shown in Online Appendix Table A7, inequality acceptance even between five-year-olds increases in both societies when there is a cost of redistribution, and the evidence suggests that adults generally make a trade-off between fairness and efficiency in their distributive decisions throughout childhood. The following result summarises this discussion.

 

Result 4.

The source of the inequality matters more for inequality acceptance between older children than between younger children in Norway, but not in Shanghai. The effect of the cost of redistribution on inequality acceptance is robust across age groups in both societies.

4. Mechanisms

We have documented a large and robust cross-societal difference in adults’ willingness to accept inequality between children, with adults in Shanghai implementing much more inequality between children than adults in Norway. It suggests that adults in the two societies have different views about what is a fair inequality between children. We here consider three alternative mechanisms that may also contribute to explain the observed cross-societal difference in inequality acceptance: a learning mechanism, a non-intervention mechanism and a randomisation mechanism.

The learning mechanism is that adults may be motivated in their decisions to prepare the children for adult life by teaching them that life is not always fair. The two societies are characterised by very different levels of income inequality, with Shanghai being one of the most unequal societies and Norway one of the most equal societies in the world. Hence, it may be that adults in Shanghai consider it more important than adults in Norway to teach children that life is not always fair. To shed light on the importance of this learning mechanism, we asked the adults to state whether they agreed with the statement ‘It is important for children to learn that life is not always fair’, on a scale from 1 to 10, where 10 indicated complete agreement. We find that adults in both societies on average agreed with this statement, but adults in Shanghai consider it more important than Norwegian adults that children learn that life is not always fair: 7.19 (Shanghai) versus 6.66 (Norway), |$p\lt .001$|⁠. In the first column in Table 3, we show that agreement with this statement is positively correlated with inequality acceptance in the experiment, which means that adults who consider it important for children to learn that life is not always fair are more inequality accepting. In the second column, we show that the learning mechanism is more important for inequality acceptance in Shanghai than in Norway (⁠|$p\lt .001$|⁠), and thus may contribute to explain some of the cross-societal difference in inequality acceptance.

Table 3.

Inequality Acceptance: Mechanisms.

	Inequality		Inequality
			|$I_B\gt 0$|	|$I_B \le I_A$|
	(1)	(2)	(3)	(4)
Norway	−0.276***	−0.105***	−0.280***	−0.288***
	(0.010)	(0.027)	(0.007)	(0.010)
Learn life is unfair	0.007***	0.021***
	(0.002)	(0.003)
Norway |$\times$| Learn		−0.025***
life is unfair		(0.004)
Constant	0.520***	0.424***	0.392***	0.578***
	(0.022)	(0.025)	(0.012)	(0.017)
Controls	Yes	Yes	Yes	Yes
Observations	6,014	6,014	4,713	5,878
|$R^{2}$|	0.132	0.140	0.266	0.136

Note: The table reports OLS regressions with the implemented inequality as the dependent variable. Norway is an indicator for the adult being from the Norwegian sample. Learn life is unfair is the adults’ responses to the question about the extent to which they agreed with the statement ‘It is important for children to learn that life is not always fair’. Responses were given on a ten-point scale, where 10 indicated complete agreement. Norway |$\times$| Learn life is unfair is an interaction between Norway and Learn life is unfair. Column (3) restricts the sample to adults who redistributed a positive amount. Column (4) restricts the sample to adults who did not redistribute more than half the money. Controls are indicators for gender, low education, low income, low age and having a child, as defined in Figure 2. Robust SEs are reported in parentheses, where *** p  |$\lt $| .01.

A second possible mechanism is that a larger share of adults in Shanghai than in Norway are reluctant to intervene in the distributive situation. Such a difference would result in more implemented inequality in Shanghai than in Norway, even if the adults in both societies had the same fairness view. However, in the third column in Table 3, we show that the estimated cross-societal difference in implemented inequality is virtually unchanged when only considering the adults who actually intervene and redistribute a positive amount, which suggests that the non-intervention mechanism is not a main driver of the observed difference in inequality acceptance between Shanghai and Norway.

A third possibility is that the share of adults who randomise when deciding for two children differ between the two societies, where randomisation would increase implemented inequality relative to an egalitarian fairness view. However, if randomisation was more prominent between the adults in Shanghai than in Norway, a larger fraction of the adults in Shanghai would give more than half of the money to the child without initial earnings. We do not find support for this randomisation mechanism. Very few adults give more to the child without initial earnings, 3.5% in Shanghai and 1.0% in Norway, and, from the fourth column in Table 3, we observe that estimated cross-societal difference in implemented inequality is unaffected by only looking at adults who do not give more to the child without initial earnings.

Taken together, the results reported in Table 3 suggest that the large cross-societal difference in implemented inequality between the two societies reflects both differences in fairness views and differences in the role of the learning mechanism: adults in Shanghai consider inequality between children to be fair to a greater extent than adults in Norway, and find it more important for children to learn that life is not always fair.

5. Follow-Up Study: Children versus Adults

People may have different attitudes to inequality between children than to inequality between adults. To study how attitudes to inequality between children compare with attitudes to inequality between adults, we conducted a follow-up study where adult populations in Shanghai and Norway make distributive choices for either children or adults. The recruitment of the adult populations was carried out by the same data collection agencies as for the main study, Opinion Research Shanghai in Shanghai (N = 1,500) and Norstat in Norway (N = 1,543). In the main study, we found that acceptance of inequality between children increases with the age of the child (Result 2). Thus, to conduct the hardest test of whether people hold different attitudes to inequality between children than to inequality between adults, we randomly allocated participants to make a decision for two seventeen-year-olds or for two adults. We used a one-to-ten matching: one out of ten participants would be randomly chosen to have their decision implemented. In all other respects, we implemented exactly the same experimental protocol for the three treatments Luck, Merit and Efficiency as in the main study; see Online Appendix C for additional details.

The main results are provided in Figure 4.⁵ In panel (a), we report the average implemented inequality between seventeen-year-olds in the main study and in the follow-up study, pooled across treatments and by society. We observe that the follow-up study closely replicates the average implemented inequality in both societies, and, consequently, also the cross-societal difference in inequality acceptance. Panel (b) reports the estimated causal effect on average implemented inequality from considering an inequality between two adults compared to the exact same inequality between two seventeen-year-old children. We find that the estimated treatment effect is positive in both societies, but only significant in Norway (⁠|$p\lt .001$|⁠). The estimated treatment effect of comparing adults and seventeen-year-olds in Norway is about the same as the estimated treatment effect when comparing five-year-olds and seventeen-year-olds in Norway in the main study (0.082 versus 0.071). The estimated difference in the treatment effect between the two societies when comparing adults and seventeen-year-olds is also significant (⁠|$p=.015$|⁠), which shows that the cross-societal difference in average implemented inequality between children is significantly greater than the cross-societal difference in average implemented inequality between adults.

Fig. 4.

Inequality Acceptance: Children versus Adults.

Note: Panel (a) shows the mean implemented inequality between seventeen-year old children, by study (main and follow-up) and society (Shanghai and Norway). Panel (b) shows the treatment effect of distributing between adults versus between seventeen-year-old children on the mean implemented inequality, by society. Panel (c) shows the treatment effect of distributing between adults versus seventeen-year-old children on the mean implemented inequality, by economic environment (Luck, Merit, Efficiency) and society. Estimates are based on the full samples with control variables. Controls are defined in Figure 2 (main study) and Online Appendix Table A8 (follow-up study). SEs are indicated by bars.

In panel (c) (Shanghai) and panel (d) (Norway) of Figure 4, we report the estimated treatment effect separately for each of the three economic environments. We observe that the greater acceptance of inequality between adults than between seventeen-year-olds is present in both societies when the source of inequality is luck (Shanghai, |$p=.001$|⁠; Norway, |$p=.040$|⁠). It is instructive to also extend this comparison to the acceptance of inequality between the youngest children in the main study. In the Luck treatment, the implemented inequality between adults in the follow-up study is about 70% higher than the implemented inequality between five-year-olds in Norway in the main study (0.298 versus 0.174), and about 15% higher in Shanghai (0.559 versus 0.484). Hence, we find strong evidence of inequality due to luck being a more acceptable source of inequality between adults than between children.

In the Merit environment, we do not find a significant difference between how people handle an inequality between two adults and an inequality between two seventeen-year-olds (Shanghai, |$p=.587$|⁠; Norway, |$p=.378$|⁠), which suggests that merit considerations have been fully integrated when considering inequality between the oldest children. Finally, we observe that the greater acceptance of inequality due to luck when considering inequality between two adults compared to between two seventeen-year-olds remains in Norway when introducing a cost of redistribution (⁠|$p\lt .001$|⁠), but vanishes in Shanghai (⁠|$p=.271$|⁠). This suggests that efficiency considerations are more acceptable when considering inequality between seventeen-year-olds in Shanghai than in Norway, and causes adults in Shanghai to handle such an inequality in the same way as they handle an inequality between adults.

Taken together, the follow-up study shows that people are indeed more accepting of inequality between adults than between children, particularly when the source of inequality is luck.

 

Result 5.

There is less acceptance of inequality between seventeen-year-olds than between adults when the source of inequality is luck, but not when the source of inequality is merit and only in Norway when there is a cost of redistribution. The societal difference in inequality acceptance is large both when considering inequality between adults and inequality between seventeen-year-olds.

6. Conclusion

Inequality considerations figure prominently in almost all spheres of society (Piketty, 2014). Childhood inequality is of particular concern since it, not only impacts the children’s current situation, but also affects their later life outcomes. Social and economic disadvantages in early life increase the risk of having lower earnings, lower standards of health and lower skills in adulthood (Heckman, 2013; Heckman et al., 2013; Conti and Heckman, 2014; García et al., 2020; Zhou et al., 2021). How inequality between children is handled and how it depends on the age of the child and the economic environment is therefore a key feature of the way we organise society. We have documented that adults, recruited from the general population in Shanghai and Norway, are more willing to accept an inequality between older children than between younger children, that they are more willing to accept an inequality between children when the inequality reflects differences in merit rather than luck and that they are more willing to accept an inequality between children when there is a cost of redistribution.

We observe a large societal difference in the willingness to accept inequality between children, with adults in Shanghai implementing twice as much inequality between children as do adults in Norway. We provide evidence suggesting that this partly reflects that adults in Shanghai find it more important than adults in Norway that children learn that life is not always fair. However, the cross-societal difference in acceptance of inequality between children also appears to reflect a general cultural difference in inequality acceptance, where, as shown in the follow-up study, people in Shanghai are more accepting of inequality both between adults and between children. Interestingly, we find in the follow-up study that the cross-societal difference in inequality acceptance is even greater when considering inequality between children than inequality between adults, which is consistent with the learning mechanism reinforcing the acceptance of inequality between children in Shanghai, but not in Norway.

Societies differ significantly in their child policies, in particular in the extent to which the most disadvantaged children are allowed to fall behind in terms of income, education, health and life satisfaction (UNICEF, 2016; 2020; OECD, 2022). The present study sheds light on these differences by documenting that the acceptance of inequality between children is very different in Shanghai and Norway, consistent with existing differences in childhood policies in the two societies. In Shanghai, there is intense focus on differences in school performance from an early age (Zhao, 2015; Zhang, 2018), and academic rankings of classmates are extensively used as a motivational tool (Liang et al., 2016). In Norway, in contrast, very little attention is paid to differences in school performance of young children: children are not graded until eighth grade (regulations pursuant to The Norwegian Education Act section 3-4), and there are no visible hierarchies of academic success. Another example is the organisation of children’s sports activities. In Shanghai, children as young as six years old are selected to attend elite sports schools with six to eight hours of daily training and a competitive environment, which contributes to highly visible inequality in athletic performance (Hong, 2010; XueDong and ShuShu, 2016). In Norway, there are policies designed to avoid large inequality in children’s sports. There are strict guidelines against the use of rankings for younger children, children cannot compete in championships before the year they turn thirteen years old and coaches are encouraged to co-operate to facilitate even matches for children below this age (The Football Association of Norway, 2017).

To study the association between the elicited willingness to accept inequality between children and policy views, we asked the adults to state the extent to which they agreed that ‘A society should have a particular focus on helping low-performing children in school’. In Table 4, we report separately for each society the estimated regression coefficient from a regression of the public policy response on the implemented inequality in the experiment.

Table 4.

Policy Preferences: Helping Low-Performing Children.

	Shanghai		Norway
	(1)	(2)	(3)	(4)
Inequality	−0.501***	−0.360***	−0.602***	−0.664***
	(0.125)	(0.124)	(0.154)	(0.152)
Constant	7.197***	7.028***	7.025***	7.974***
	(0.077)	(0.162)	(0.074)	(0.170)
Controls	No	Yes	No	Yes
Observations	3,000	3,000	3,014	3,014
|$R^{2}$|	0.005	0.055	0.005	0.041

Note: The table reports OLS regressions where the dependent variable is the answer to the public policy question about the extent to which they agree with the statement ‘A society should have a particular focus on helping low-performing children in school’. Responses to the question were given on a ten-point scale, where 10 indicated complete agreement. The sample are the Shanghai adults in columns (1) and (2), and the Norwegian adults in columns (3) and (4). Inequality is the inequality implemented by the adult. Controls are indicators for gender, low education, low income, low age and having a child, as defined in Figure 2. Robust SEs are reported in parentheses, where *** p  |$\lt $| .01.

We observe that implemented inequality in the experiment is highly predictive of policy attitudes: in both societies, the degree to which adults agree with having a particular focus on helping the low-performing children in school is strongly positively associated with the extent to which they equalise in the experiment (⁠|$p\lt .001$|⁠). These findings suggest that differences in policies towards children in the two societies partly reflect differences in what is considered a fair inequality between children.

The present study also provides evidence of interest for understanding how inequality acceptance is transmitted across generations in society (Bisin and Verdier, 2001; 2011; 2024; Besley and Persson, 2019). Gene-culture co-evolution has facilitated the propensity in humans for cultural learning already at a very young age (Boyd and Richerson, 1985; Henrich and Boyd, 1998; Gintis, 2011), and oblique socialisation has been argued to be important in governing the formation of values. Oblique socialisation is not necessarily purposeful, in contrast to much of parental socialisation, but rather reflects that society at large inevitably transmit values to the younger generation through their actions. This study may be seen as providing initial evidence of the oblique socialisation process of inequality acceptance being very different in Norway and Shanghai, where children in Norway are socialised into a low level of inequality acceptance and children in Shanghai into a high level of inequality acceptance. Models of oblique socialisation thus suggest that different levels of inequality acceptance sustain across generations in these two societies.

It is further interesting to compare our findings with the results from experimental studies, starting with the seminal work by Fehr et al. (2008), that use versions of the standard dictator game to examine the moral development of children (Fehr et al., 2013; Bauer et al., 2014; Blake et al., 2015; Sutter et al., 2019; Samek et al., 2020). A stylised fact from these experiments is that simple inequality aversion seems to peak in early childhood (7–10 years), with adolescents internalising more complex notions of fairness that are sensitive to the source of inequality and the cost of redistribution (Almås et al., 2010). These findings are broadly in line with oblique socialisation following the pattern observed in Norway in our study—adults are the least inequality accepting between the youngest children, but become more inequality accepting during the adolescent years of childhood, especially when the source of inequality is due to differences in merit.

How inequality between children should be handled is a fundamental question for any society, but we lack evidence of people’s attitudes to childhood inequality. We provide novel large-scale evidence showing that there are large societal differences in the acceptance of inequality between children, and we hope that future research will enrich these findings through other comparative studies and by exploring different dimensions of childhood inequality. Our experimental paradigm can be extended in many directions to uncover various types of fairness concerns raised when allocating resources between children, and thus opens up a new research avenue for understanding societal differences in the acceptance of childhood inequality.

Additional Supporting Information may be found in the online version of this article:

Online Appendix

Replication Package

Notes

The data and codes for this paper are available on the Journal repository. They were checked for their ability to reproduce the results presented in the paper. The replication package for this paper is available at the following address: https://doi.org/10.5281/zenodo.13981874.

This project was organised by the Centre for Experimental Research on Fairness, Inequality and Rationality (FAIR) and implemented in collaboration with East China Normal University. We thank Editor Steffen Huck and three anonymous referees for valuable comments and suggestions. We are grateful to Michal Bauer, James Heckman, Sam Bowles, Julie Chytilová, Matthias Sutter and Kathleen Vohs for many valuable comments and suggestions, and to Xiao Chi, Wu Jie, Jin Jinglin, Iselin Jordan, Shi Lan, Yuanhao Li, Fan Mengting, Wang Qian, Wang Qiuhua, Zhang Rui, Gu Yiyun, Sun Xiaoxue, Xiaogeng Xu and Lena Øyberg for their research assistance. The study was reviewed and approved by East China Normal University IRB (HR 065) and NHH Norwegian School of Economics IRB, and reviewed by the NSD Norwegian Centre for Research Data (59488/3/HJT). The data that support the findings of this study are openly available on Harvard Dataverse (Cappelen et al., 2024).

The project was supported by the Institute for Applied International Studies (Fafo) Project No. 225132, NORFACE (New Opportunities for Research Funding Agency Co-operation in Europe) Network, Project No. 462-14-030, the European Research Council Project FAIR No. 788433 and the Research Council of Norway through its Centres of Excellence Scheme, FAIR Project No. 262675.

Footnotes

1

Histograms of the distributive choices by treatment are reported in Online Appendix Figure A2 (Shanghai) and Figure A3 (Norway).

2

In Online Appendix Figure A6, we also show that there are no large differences in how adults with Shanghai hukou at birth and non-Shanghai hukou at birth respond to the ages of the children or to variations in the economic environment.

3

See Online Appendix Table A4 for the corresponding regression analysis.

4

The associations with the background variables for Norway are very much in line with what is observed for the acceptance of inequality between adults in Norway in Almås et al. (2020).

5

Further details are provided in Online Appendix Figures A8 and A9 and Table A8.