Using machine learning to study parenting styles

How should we raise our children? Research has shown that the amount of parental time invested is not the only crucial element for the development of children’s skills (Del Boca et al. 2014, Attanasio et

How should we raise our children? Research has shown that the amount of parental time invested is not the only crucial element for the development of children’s skills (Del Boca et al. 2014, Attanasio et al. 2016); parenting style also matters (Fiorini and Keane 2014). Parenting style is a strategic choice linked to incentives (Doepke and Zilibotti 2014). In order to study the relationship between parenting style and child development, researchers rely on ad hoc perceptions or previous research to narrow down the complexities of parenting to certain key actions. For example, reading to children has been shown to be highly predictive of children’s skill development (Kalb and Jan van Ours 2013). But how do you measure parenting styles without relying on prior beliefs?

In a recent article (Rauh and Renée 2022), we take an approach that lets the data do the talking. We adopt a model from computational linguistics to determine which types of parenting styles exist and are most salient. The initial goal of the machine learning algorithm is to learn from the co-occurrence of words and form topics around them. In the context of parenting styles, the idea is that parents who engage in one activity might be more likely to engage in another given activity, so that the algorithm learns from the co-occurrence of parenting actions to define two or more parenting styles.

Data

We rely on the Quebec Longitudinal Study on Child Development (ÉLDEQ), a detailed panel of a representative sample of families from Quebec, a province of Canada, with a baby born between October 1997 and July 1998. We use information on mothers’ behaviors collected through three waves conducted when the target children were 5, 17 and 29 months old.

One of the advantages of the dataset is that the behavior of mothers towards their children is not captured by self-reported survey questions, but by the interviewer. At the end of each interview, the interviewer records ten variables classifying whether the mother engages in certain actions during the interview. This leaves the data less susceptible to bias that may result from mothers misreporting their behavior towards their children.

In Table 1, we see the list of the ten activities and the percentage of mothers who engage in them during the three interviews. Two things stand out. First, parents on the whole are more likely to be supportive in affirming progress or kissing and hugging the child rather than yelling at the child or expressing annoyance. Second, despite this imbalance appearing in all three survey waves, there is a considerable shift towards more punitive and less supportive actions as the child ages.

Table 1 Parental actions across survey waves

Remarks: The table describes the behavior of respondents and their interactions with their children during the ÉLDEQ annual interview in terms of percentages. Behaviors are assessed by the interviewer during the interview.

Parenting Styles

The model we use to classify parenting styles is the Latent Dirichlet Allocation (LDA) developed by Blei et al. (2003) for classifying text into subjects. In recent work, economists have used LDA to determine salient CEO behaviors (Bandiera et al. 2020) and political ideologies (Draca and Schwarz 2021). In our case, each parenting style is defined by the probability of each parenting action, and parents are classified into two styles. The classification is not “strict” in the sense that the parents can be a mixture of the two types: the algorithm assigns shares of each style to the parents in the data set.

In Figure 1, we show the resulting distribution of parenting actions when classifying parenting into two styles, A and B. The red bars indicate the prevalence of a given action among parenting style A and the blue bars among the parenting style B. The left panel displays the probability of each of the ten actions in a topic. A parenting style should be imagined as an urn from which a parent draws actions. A mother following parenting style A will close her eyes and draw an action from the red bars. She will most likely draw the largest bars; for example, she can hand an educational toy to her child or respond to her baby’s noises. A B parenting style mother will probably not get any action when pulling blue bars from the urn because they are so small. In other words, parenting style B is characterized by inaction. Based on these action distributions, we follow developmental psychologists McCoby and Martin (1983) and label parenting style A as “warm” and parenting style B as “cold”. In the right panel of Figure 1, we see the standardized relative prevalence of a given action within a parenting style. We see that warm mothers are relatively unlikely to scold or yell at their child. Cold parents, if any, simply check on their child.

Figure 1 Breakdown of shares by parenting style

Remarks: The left panel describes the proportion of action subjects for each of the two parental styles. The right panel shows the standardized importance of a stock in a style by setting the mean to 0 and the standard deviation to 1.

Are all parents equally likely to follow a warm parenting style? To answer this question, we examine the correlation between warm parenting and parenting and family characteristics. In Figure 2, we see the distribution of parenting styles according to maternal education. In the left panel, we see that among less educated mothers with at most a high school diploma, the cold parenting style is quite likely. This is indicated by the mass to the left of the graph. If we look at the right panel, we find the opposite for mothers with a university degree. For highly educated mothers, we observe a shift in the distribution to the right; they are more likely to follow a warm parenting style. We also find that younger mothers and those with more children are more likely to adopt a cold parenting style.

Figure 2 Distribution of styles according to maternal education averaged over the waves

Remarks: The transparent bars represent the clustered probabilities of the likelihood of engaging in warm rather than cold parenting, while the solid line is the kernel density. The sample is the pooled sample in which each parent appears three times.

Table 1 shows that styles change over time. Although there is some persistence – i.e. mothers with a hot style are more likely to continue following a hot style and those with a cold style are more likely to continue following a cold style – there are also systematic changes as the child ages. On average, parenting styles become cooler. The other remarkable change is that, while at 5 months there is no difference between parenting styles towards boys and girls, by the time the child reaches 29 months, boys are statistically more likely to to be confronted with a cold parenting style.

Do parenting styles affect skill development?

Although we cannot provide a definitive answer to this question due to a lack of exogenous variation, we can examine whether children exposed to certain parenting styles achieve higher skill levels at older ages. Specifically, we examine summary measures of cognitive test scores (such as math or logic) and non-cognitive scores (such as behavioral problems or hyperactivity) at age 6.

In Figure 3, we show the impact of warm parenting on standardized skill measures at age 6. We show the relationship for parenting styles measured at each age separately (top) and for an aggregate measure of style calculated across all three ages (bottom). We see that children exposed to totally warm parenting style rather than absolutely cold parenting style at 5 months acquire cognitive skills (left) greater than 0.3 standard deviations and non-cognitive skills (right ) higher by more than 0.2 standard deviation. The effect sizes of parenting styles at 5 months of age are larger than at 17 and 29 months even though parenting style is measured furthest from the outcome when the child is 6 years old. This supports to some extent the idea that investments in very early childhood are of particular importance.

picture 3 Regression coefficients of warm parenting on cognitive and non-cognitive skills at age 6

Remarks: The dependent variable is calculated by taking the first factor of six measures of each, cognitive and non-cognitive ability, at age 6. The score is standardized with a mean of zero and a standard deviation of one. The thin lines represent the 90% confidence interval.

The overall measure of parenting styles has an even higher correlation with the results. Warm parenting is associated with more than a half standard deviation of higher cognitive skills and a third of a standard deviation of higher non-cognitive skills.

Look forward

Given the exponential growth of available data, we should leverage machine learning to better understand what kind of parenting “works.” Letting the data speak allows us to challenge traditional assumptions and uncover patterns of success. This could help policymakers and researchers design interventions and support programs to help parents navigate the complexities of child-rearing and avoid the “parent trap” (Hilger 2022).

References

Attanasio, O, S Cattan and S Krutikova (2016), “Early Childhood Development Policy: Evidence and Research Agenda”, VoxEU.org, 9 June.

Bandiera, O, A Prat, A Hansen and R Sadun (2020), “CEO behavior and firm performance”, Journal of Political Economy 128(4): 1325–1369.

Blei, DM, AY Ng and MI Jordan (2003), “Latent Dirichlet allocation”, Machine Learning Research Journal 3: 993–1022.

Del Boca, D, C Flinn and M Wiswall (2014), ‘Household choice and child development’, Review of economic studies 81(1): 137–185.

Draca, M and C Schwarz (2021), “How polarized are citizens? Measuring ideology from the bottom up», SSRN, 11 May.

Doepke, M and F Zilibotti (2014), “Tiger mothers and helicopter parents: the economics of parenting”, VoxEU.org, 11 October.

Fiorini, M and MP Keane (2014), “How children’s time allocation affects cognitive and non-cognitive development”, Journal of Labor Economics 32(4): 787–836.

Hilger, NG (2022), The Parent Trap: How to Stop Overburdening Parents and Solve Our Inequality CrisisMIT Press.

Kalb, G and J van Ours (2013), “Reading to children: a head start in life”, VoxEU.org, 10 June.

McCoby, E and J Martin (1983), “Socialization in the Context of the Family: Parent-Child Interaction”, Handbook of child psychology 4: 1–101.

Rauh, C and L Renée (2022), “How to measure parenting styles?”, CEPR Working Paper 17326.