An update on the scientific evidence for and against the legal banning of disciplinary spanking

Abstract

In 2004, the Canadian Supreme Court upheld the right of Canadian parents to use open-handed swats to a child’s buttocks to correct child misbehavior, but only as developmentally appropriate (i.e., between the ages of 2 and 12). Some social scientists believe that the Canadian Supreme Court did not go far enough. These researchers support total bans on physical discipline. Other social scientists support Canada’s existing age-specific legislation. This commentary provides a concise overview of physical discipline research since 2004, emphasizing the methodological rigor of the research used to argue for and against total spanking bans. Advocates of total bans primarily cite reviews based on bivariate correlations and non-randomized methods known to be inherently biased against disciplinary actions (i.e., methods known to make all disciplinary responses to defiance appear harmful). In contrast, those who support Canada’s existing legislation have systematically compared methods known to be inherently biased in opposite directions (i.e., harmful- and beneficial-looking), to demonstrate that the true average effect size of customary spanking on child outcomes is likely very near zero. These researchers also emphasize four randomized clinical trials in which spanking increased compliance in defiant preschoolers. Other issues discussed in this commentary are: the developmental trajectories for children who do not learn to comply with parental directives while they are young; and children’s risk of assault in countries with and without total bans. We conclude that the most rigorous empirical studies and available crime statistics validate the appropriateness of Canada’s existing legislation on disciplinary spanking.

A primary goal of parenting is to ensure that children become well-adjusted, responsible adults. Most parents and professionals recognize that good parenting requires some discipline and thus welcome guidance about the most appropriate ways to administer discipline. Toward this end, all three levels of the Canadian courts have wrestled with extensive social scientific and legal evidence for and against disciplinary spanking.

In 2004, the Canadian Supreme Court agreed with two lower-court decisions to retain the right of parents to use reasonable force to correct their children (Section 43), while limiting legal forms of reasonable force (1). The court saw value in permitting parents to use open-handed swats to a child’s buttocks to correct child misbehavior, but only as developmentally appropriate (i.e., between the ages of 2 and 12).

This commentary provides a concise overview of physical discipline research since 2004. In Part I, we raise some general methodological considerations that consumers of this research should attend to. We do this to set up Part II, in which we review the specific studies most frequently cited to argue for and against spanking bans. In Part III, we raise additional considerations for parents and professionals concerned about the well-being of Canada’s children. We conclude by lauding the appropriateness of Canada’s current approach to the legislation of disciplinary spanking.

Part I: Methodological Considerations

Spanking is a corrective action, that is, an action taken to correct (or ameliorate) a problem. Examples of corrective actions include: chemotherapy (to correct cancer), methylphenidate (to correct ADHD symptomology), and spanking and other disciplinary responses (to correct child misbehavior). When evaluating studies of corrective actions, it is imperative to know: how the corrective action is being defined, the presenting problems in the population receiving the corrective action, the dosage regimen, and whether the corrective action works alone or in the context of a comprehensive treatment package. Also critical, is an understanding of whether the evidence obtained permits causal inference.

Definition, age, regimen/context

How researchers define spanking typically reflects their predisposing views on spanking. Researchers who consider any parental use of physical force to be abusive sometimes fail to distinguish spanking from actions that all discipline researchers view as abusive (e.g., kick, beat up). When defining “physical discipline” or “physical punishment” for review articles, these researchers often cast a wide inclusion net, lumping spanking together with many forms of severe and abusive punishment. Specific examples of these problematically-broad definitions have been identified by Baumrind et al. (2) and Larzelere et al. (3). Broad definitions can make non-harmful types of physical discipline appear harmful due to the conflation of spanking with severe and abusive physical punishment.

In contrast, researchers who believe that spanking can be administered in a non-abusive way distinguish “customary spanking” from overly harsh/abusive discipline, and sometimes make even finer discriminations. For example, some use the term “conditional spanking” (4) to refer to spanking used to back-up milder techniques (e.g., for refusal to cooperate with time-out).

Researchers opposed to spanking have also tended to amalgamate children aged 0–17 in their analyses (5,6). Lumping children with infants and teens (for whom physical discipline is not developmentally appropriate) elevates group means for undesirable outcomes and reduces the means for desirable ones. Physical discipline may appear harmful for the full group, even if it was neutral or beneficial for children in a specific age range.

In contrast, researchers who strive to distinguish spanking from abuse are more likely to limit their investigations to children aged 18 months through 11 years. As such, research conducted by this group is more directly applicable to the current legal debate in Canada over non-abusive spanking with children ages 2–12.

Four categories of evidence

Predisposing views also impact the types of evidence that researchers create, cite, and emphasize. Four broad categories of evidence are distinguished here (to set up our presentation of the specific studies most cited in the current spanking debate).

The first category is correlational evidence. Because corrective actions are typically applied only when there is a problem to correct, both cross-sectional and longitudinal studies will virtually always yield a positive correlation between the use of a corrective action and the severity of the problem it is intended to correct. (Methylphenidate use will always be positively correlated with ADHD symptomatology; discipline use will always be positively correlated with frequency of misbehavior.) As with corrective medical actions, no causal inferences can be drawn from correlational studies of spanking and child outcomes.

The second category is baseline-partialled longitudinal (BPL) evidence (aka controlled longitudinal (3)) evidence). When predicting the future severity of a problem, BPL studies statistically control for (i.e., partial out) the severity of the problem at baseline. BPL studies are often used to make causal inferences in the social sciences, even though most BPL analyses are known to reduce -- but not eliminate -- inherent statistical biases.

The most common type of BPL study used to assess spanking relies on ANCOVA-type statistics. ANCOVA-type longitudinal analyses have been shown to be biased against corrective actions (i.e., tend to make these actions appear harmful: 7–9). ANCOVA analyses of corrective actions yield a positive association between baseline correction and subsequent severity because people with more problems at baseline typically have more problems at follow-up, compared to those who had no need for the corrective action in the first place. ANCOVA only partially reduces the selection confound. This apparent harm (i.e., positive association) has been documented not just for spanking, but also for nonphysical disciplinary responses, psychotherapy, methylphenidate (9–11) and out-of-home placements (12).

Another type of BPL study relies on change-score (aka “slope”) analyses wherein the corrective action is used to predict change in the severity of problems from baseline to follow-up (i.e., the slope of the line from baseline to follow-up). Change score analyses are typically biased in favor of corrective actions. Because a corrective action is usually applied when a problem is more severe (e.g., a child is particularly defiant), there may be natural regression toward the mean prior to follow-up. Statistical methodologists have been exploring ways to overcome the biases particular to these two BPL methods (discussed later).

The third category of evidence is randomized experiments. Randomized experiments are the standard in medical research but occur rarely in child discipline research. We know of only four randomized experiments of spanking, conducted when clinical researchers were searching for alternatives to the traditional use of spanking to enforce compliance with a chair time-out. In these studies conducted by Roberts and his colleagues (13–16), clinically-referred 2- to 6-year-olds who left chair time-out prematurely were prescribed “two swats” versus an alternative “back-up” procedure, and then returned to the chair. Unfortunately, these studies are now quite dated (1981–1990) and an anti-punishment bias among many contemporary parenting researchers will likely preclude their replication.

The fourth category of evidence is a literature review. Some reviews are narrative (box-score). Authors of narrative reviews simply count the number of studies documenting harmful- or beneficial-looking results. Other reviews are meta-analytic. Authors of meta-analyses (MAs) calculate and provide average effect sizes. Regardless of the type, a review is only as good as the individual studies on which it is based. Causal inference cannot be drawn from reviews of bivariate correlations; reviews of ANCOVAs or change-scores are subject to the biases inherent in these two baseline-adjustment methods. Consequently, reviews of randomized trials are strongly preferred in medical research (17).

Part II: Studies Used to Argue For and Against Spanking Bans

Studies cited in Canada’s 2004 Supreme Court decision

The 2004 court decision emphasized two reviews and two BPL studies. The first review was a narrative review from a scientific consensus conference on corporal punishment (18). The second review was a preliminary draft of Gershoff’s (5) MA based exclusively on bivariate correlations obtained from studies of children 0–17 (mean ages 1–16). The individual studies were the first two BPL studies of physical discipline (19, 20). Like the authors of the two reviews, the authors of the two BPL studies came to some shared and some contradictory conclusions.

Studies emphasized to promote spanking bans

Since the 2004 Supreme Court ruling, researchers who are unconditionally-opposed to spanking have tended to emphasize three types of studies.

1. Meta-analyses of bivariate correlation coefficients

Some of the most frequently cited studies of physical discipline are Gershoff’s initial MA (5) and an updated version co-authored with Grogan-Kaylor (21). The update is an improvement over the initial version, in that 24% of the studies in the update were longitudinal, and many studies lumping spanking with severe corporal punishment were omitted. But correlations are still correlations, which the researchers acknowledge “do not rule out the potential for a child elicitation effect” (p. 455). Thus, “causal links between spanking and child outcomes cannot be established by these meta-analyses” (21, p. 464).

Despite these acknowledgements, the American Academy of Pediatrics (AAP) and the American Psychological Association (APA) repeatedly cited these unadjusted correlations in their empirical rationale for officially opposing all spanking in 2018 (22) and 2019 (23, 24). In response to various critiques of this evidence (25, 26), the APA Task Force cited a lack of evidence of any beneficial outcome of spanking (27).

2. Baseline-partialled longitudinal studies (mostly ANCOVA-type)

A more recent narrative review (6) went beyond bivariate correlations by tallying the number of BPL studies that reported harmful- and beneficial-looking associations of baseline physical punishment with subsequent child outcome. Children had mean ages of 6 months to 14 years when punished. Because the vast majority of these studies employed ANCOVA-type statistics (known to be biased against corrective actions), it is not surprising that 69% of the studies reported harmful-looking effects. Only 1% found more beneficial- than harmful-looking outcomes; 30% had mixed significant and non-significant outcomes. The researchers presented this tally as “compelling evidence that physical punishment is harmful to children’s development and wellbeing,” asserting that all countries should therefore join UN efforts “by prohibiting physical punishment in all forms and all settings” (6, p. 361).

3. Attempts to correct known bias

Attempts to overcome the selection bias inherent in ANCOVA-type analyses of spanking have included the use of: propensity score methods (28), entropy balancing (29), and fixed-effects regression (30). Unfortunately, systematic comparisons have revealed that propensity score methods do not necessarily overcome the selection bias (31, 32). For example, Lin (33) found that both prescription medications for mothers and therapy for mothers still looked harmful (i.e., were positively associated with their own subsequent depression severity) according to propensity-score matching. Entropy balancing can be understood as a flexible propensity score method (32). Fixed-effects regression can be less biased than ANCOVA-type analyses when researchers attend carefully to temporal sequence, but many datasets do not lend themselves to the hypotheses researchers want to test. In one application of fixed-effects regression to physical discipline (30), a spanking-ban advocate demonstrated that above average spanking for a specific child in the past week was associated with above-average antisocial behavior during the past three months (i.e., temporal direction was reversed, providing evidence for child elicitation of spanking rather than a spanking effect).

Despite recent advances in approximating unbiased causal effects in longitudinal studies, eliminating selection bias remains a difficult challenge (32). Studies have demonstrated the importance of partialling out the most relevant covariates (31), identifying treatment assignment mechanisms, and designing studies to focus on similar comparison groups (34) (e.g., equivalent presenting problems, such as oppositional defiance). Meta-analyses of BPL studies suggest that the most important covariates to partial when conducting spanking research are baseline child adjustment and child age (3, 35, 36).

Summary of the evidence used to support spanking bans

The most cited empirical evidence used to promote spanking bans are reviews of children aged 0 to 17 years that are either correlational or based on statistical methods known to be biased against corrective actions. Efforts to eliminate these biases have not been convincing.

Studies emphasized to support Canada’s current age-specific approach

Researchers who support Canada’s existing age-delimited approach to spanking (i.e., only between the ages of 2 and 12) tend to focus on four lines of evidence.

1. Meta-analyses of baseline-partialled longitudinal studies

Three MAs of physical discipline have gone beyond correlations by adjusting for baseline differences on child outcomes (3, 35, 36). These MAs also distinguished customary spanking from more severe physical discipline. Two were limited to children aged 18 months through 11 years (3, 36).

In all three of these MAs, the researchers emphasized the trivial effect sizes obtained from BPL studies of customary spanking. Although the majority of qualifying studies in each MA (45–79%) did report a harmful-looking effect, spanking explained less than 1% of the variance in child outcome (0.64% for externalizing problems; 0.16% for internalizing, cognitive achievement, and social competence in the most recent MA (3)). Effect sizes this small can easily be explained as residual confounding (37, p. 99) due to researchers’ failure to include all relevant covariates.

The researchers also addressed the biases inherent in the usual ANCOVA analyses and in change-score analyses. The two most recent MAs computed effect sizes for the two (oppositely-biased) methods. As expected, average ANCOVA results suggested harmful-looking effects of spanking, and average change-score results suggested beneficial-looking effects of spanking. In other words, all “effects” were consistent with each method’s known bias and too small to overcome the opposite bias in the other method.

Per Angrist and Pischke’s (38) assertion that these two methods likely bracket the true effect, the researchers concluded that the average effects of customary spanking on child outcomes are very near zero. If these results occurred for a medical intervention, researchers and physicians would next attempt to evaluate the effectiveness of various implementations of the corrective action (as a function of presenting problems, dosage regimen, other patient characteristics, etc.). This would be consistent with the likelihood of genuinely heterogeneous effects demonstrated by Imbens & Xu (32). Unfortunately, such nuances are largely ignored in the research on parental discipline.

2. Comparisons of spanking with alternative discipline tactics

Another way to evaluate spanking is to compare it with its alternatives -- investigated with the same methods in the same families. The only MA comparing effect sizes for various types of physical discipline with alternative disciplinary tactics (4) indicated that children benefitted the most from conditional spanking (i.e., spanking used only to back up milder techniques). For increasing compliance and reducing antisocial behavior, conditional spanking outperformed 10 of 13 alternatives. Not surprisingly, children were harmed the most when physical discipline was overly severe or was parents’ predominant response to children’s misbehavior. Most studies in this MA had investigated customary spanking (i.e., “spanking” as defined by the parent), which performed in the middle. Child outcomes associated with customary spanking did not differ from those associated with other disciplinary responses (e.g., privilege removal, reasoning), except in one retrospective study favoring customary spanking for drug abuse eight or more years later (39).

Subsequent to this MA, there have been three more BPL studies that compared spanking with alternative disciplinary techniques. Two utilized longitudinal data from the US (10, 40). The third (11) is the only known BPL study of physical punishment in Canada. Table 1 is an updated synthesis of the results of these three studies with other published mean effect sizes, demonstrating the pattern of biases typical of all corrective actions, c.f. (41).

Table 1.

Standardized Associations of Three Corrective Actions with Externalizing-Type Problems by Type of Evidence^a

Type of Evidence	Spanking	Nonphysical Punishments	Psychotherapy
Non-randomized studies	.20***b	.17***d	.10**d
Cross-sectional correlations
Longitudinal correlations	.16***b	.19***e	.16***e
ANCOVA/beta-type controls for initial differences	.07***b	.05*e	.12*e
Within-person changes (slope)	−.04*b	−.05*e	.00e
Randomized trials	−.35*c	−.63*f	−.24***g
Corrective action vs. control group

^aAll coefficients are standardized. For all cell elements, positive coefficients indicate harmful-looking associations (e.g., spanking is correlated with more externalizing-type problems); negative coefficients indicate beneficial-looking associations.

^bMean effect sizes (r, b) from Larzelere, Gunnoe, & Ferguson (36, p. 2044, Table 1) for externalizing. These values are similar to correlations reported by Gershoff & Grogan-Kaylor (21) and to the ANCOVA-type coefficients reported by Ferguson (35).

^cDifference between the effect size d for conditional physical punishment vs. the control condition (child release from timeout) for compliance reported by Larzelere & Kuhn (4, p. 20, top half of Table IV; d = −.74 is equivalent to r = −.35).

^dUnweighted average of effect sizes for three nonphysical tactics and effect size for therapy with the antisocial outcome in Larzelere, Cox, & Smith (10, p. 6, Table 2).

^eNonphysical punishment is the unweighted average of three effects from three studies by Larzelere and colleagues (each study contributing one effect). The first effect is for one item assessing two tactics (11, p. 185, Table 1, antisocial). The second is the average of three tactics (10, p. 8, Table 5, continuous antisocial). The third is the average of three tactics (40, p. 530 text & 531, Table 1, externalizing). Therapy is the unweighted average of three effect sizes from the same three publications/same DVs.

^fMean effect size for time-out only for conduct problems reported by Larzelere et al. (55, p. 300; d = −1.62, equivalent to r = −.63).

^gMean effect size reported by Weisz et al. for conduct problems (56, p. 227; g = −.50, equivalent to r = −.24).

^*p < .05.

^**p < .01.

^***p < .001.

To create Table 1, we calculated five mean effect sizes (cross-sectional correlations, longitudinal correlations, ANCOVA betas, change-score slope coefficients, and treatment effects from randomized trials) for three types of corrective actions (spanking, nonphysical discipline techniques, and psychotherapy) for externalizing problems (the most frequent outcome in studies of parental discipline). As seen in the first three rows of the tabled data, correlations and ANCOVA betas make all three corrective actions appear harmful, although ANCOVA reduces the bias somewhat, relative to correlations. In contrast, the fourth and fifth rows show that slope coefficients from change-score analyses and randomized trials make all three corrective actions appear beneficial.

3. Attempts to correct known bias

In addition to measurement issues, ANCOVA-type analyses are biased by between-child differences that do not change during the study period (7,8). Recommendations for overcoming this bias include slope-type change-score innovations (e.g., random-intercepts cross-lagged panel model). To date, these recommended methods have produced either non-significant associations between spanking and child outcomes or have reversed the apparent effects of spanking on externalizing problems from harmful- to beneficial-looking (7, pp. 4–6 of online Supplemental Material; 42).

4. Randomized trials

The only method permitting researchers to confidently infer causation is a randomized experiment. This makes the four trials by Roberts and his colleagues (13–16) very important – and a focus of debate in the spanking literature. Promoters of spanking bans have emphasized the small sample sizes and attempted to explain away the benefits of spanking documented in this series of studies (21, 24). Defenders of Canada’s current approach acknowledge the small sample sizes but emphasize the large beneficial effect of this spanking on children’s compliance to both parental commands and time-out (3).

Compliance to time-out is critical because the effectiveness of behavioral parent training (BPT; the most empirically-supported non-medication treatment for oppositional defiant disorder and ADHD) depends on time-out cooperation. Cooperation with time-out permits parents to quickly phase-out or reduce the need for disciplinary spanking over time (a goal shared by all spanking researchers).

Summary of the evidence used to support Canada’s current approach

Researchers who support Canada’s existing age-specific approach to spanking emphasize randomized trials and effect sizes obtained in MAs that go beyond unadjusted correlations. They tend to limit their studies to children (excluding infants and teens) and strive for transparency concerning biases inherent in various statistical approaches. Moreover, they have investigated nonphysical discipline techniques to make recommendations for replacing spanking but have not found any that is more effective than customary spanking.

Part III: Additional Considerations

In this section, we address two concerns likely raised with our review of the most emphasized studies. We also speculate on some possible reasons for differences in child abuse and delinquency statistics across countries with and without spanking bans.

1. If customary spanking and nonphysical punishments “work the same way,” can’t we just ban spanking?

Canadian parents currently have the option to use both physical and nonphysical punishment with children ages 2–12. We believe they should retain both options for two reasons.

First, equivalent mean effects do not mean that different types of discipline work equally well for all children (just as different ADHD medications with equivalent mean effects do not work equally well for all children with ADHD). The fact that both spanking and a brief barrier-enforced room isolation increased compliance with chair time-out in Roberts’ randomized studies (13–16) has often been cited as evidence that parents do not “need” spanking. In the medical field, when two corrective actions are both demonstrated to be effective, both are kept available as treatment options. Having options in medicine is critical because individual clients often respond better to one type of treatment than another. Similarly, in Roberts’ lab, some children responded better to one back-up condition than the other (16).

Second, corrective actions sometimes work better in combination. Most children who intially refused to cooperate in Roberts’ studies developed the self-control to cooperate when the parental directive to remain in chair time-out was consistently backed-up with a more forceful consequence (3, 43). The combination of a less forceful corrective action (time-out) with a more forceful one (spanking or brief room isolation) permitted the phasing out of the more forceful action (16). Consistent with these randomized results, two non-randomized studies found that never-spanked children in certain parenting contexts appeared less well-adjusted than those whose spanking had been phased out by ages 9 or 11 (44, 45).

2. Do spanking bans protect children?

Spanking bans are often promoted as a way to improve children’s lives by reducing their risk of assault. Whether they actually do this is unclear.

Spanking bans do seem to be related to parental reports of disciplinary tactics. In Bussmann et al.’s comparison of five European countries (three with spanking bans and two without), parents reported less use of customary spanking in countries where spanking was banned (46). This was particularly true for Sweden, the first country to ban spanking (in 1979) and the country with the most extensive publicity campaigns to promote awareness of the ban. Patterns for more severe physical discipline were somewhat less consistent. Seven years after their spanking ban, more German parents reported “beating with an object” (5.2%) than French parents (4.5%, no ban) and more “severe beating” (9.0%) than Spanish parents (4.4%, no ban). (Statistical significance was not reported in the original article).

While parental reports across countries suggest that Sweden’s ban may be protecting Swedish children, Swedish crime statistics suggest a more complicated story. Within-Sweden, across-time comparisons suggest that several types of assault increased markedly for the first cohort of children raised after the ban, and then again for the second cohort. Sweden not only posted increased allegations of child abuse (which may or may not (47) be attributable to changes in definition, social awareness, and reporting) but also increased allegations of violence perpetrated by minors. From 1981 to 2010, Sweden saw a 22-fold increase in alleged cases of child abuse and a 24-fold increase in alleged criminal assaults by minors against children ages 7–14 (47).

We do not presume that Sweden’s ban caused these apparent increases. Non-experimental data does not permit causal inference, and these increases are much too large to be reasonably attributed to any one cause. A thorough analysis would need to consider many factors including budgets for child protection and forensic services, the general economy, changes in family structure, media exposure, immigration patterns, secularization, etc. The ban may be a partial contributor to the dramatic increase in Swedish assaults on children. At minimum, the ban failed to take Sweden in the direction that Swedish legislators intended.

In contrast, Canada’s 2004 legislation was followed by a decrease in child abuse. Trocmé et al. (48) reported that substantiated physical child abuse throughout Canada (excluding Quebec) decreased from 5.31 cases per 1000 children in 2003 to 3.19 cases per 1000 in 2008. Within Quebec, Clement et al. (49) documented a significant decrease in “minor physical violence” and a nonsignificant decrease in “severe physical violence” from 2004 to 2012. Again, we do not presume that Canada’s 2004 legislation caused this decrease. Indeed, rates were already falling in the period from 1999 to 2004. Our point is simply that Canada’s age-specific ban on physical discipline was followed by a decrease in substantiated cases of child abuse, whereas Sweden’s total ban was followed by a large increase in alleged child abuse cases.

If these two bans did increase risk for Swedish children and decrease risk for Canadian children, how might we explain this difference? One possible explanation is that Canadian parents have retained the authority necessary to help young children grow into responsible adolescents and adults, whereas the authority of Swedish parents was undermined. Such undermining is a serious concern. In a review of studies from the US and Scandinavia, Patterson and Fisher (50) concluded that most children’s disposition to comply with adult directives increases steadily from ages 2–8, and parents’ failure to secure compliance about 70% of the time foreshadows serious dysfunction.

One manifestation of family dysfunction is child abuse, which may stem (in part) from parental anger over limited authority/control (51). Baumrind observed that permissive parents were more likely than authoritative parents to admit to “explosive attacks of rage in which they inflicted more pain or injury upon the child than they had intended” (51, p. 35). The possibility that customary spanking helps suppress children’s misbehavior before parents’ escalating frustration leads to undisputed abuse is supported by Bussmann et al.’s (46) analyses predicting “mild” and “severe” violence in three European countries that have banned spanking. As shown in Table 24.2 of the Bussman report, parents’ knowledge that mild corporal punishment was prohibited in Sweden, Austria and Germany was negatively associated with the use of mild corporal punishment (b = −.26, p < .001), but positively associated with severe physical punishment (b = .09, p < .001).

Perhaps some parents in the ban-countries followed the law until their children’s behavior became unbearable and then temporarily “snapped,” surprising themselves with an atypical burst of severe punishment. Atypical bursts may help explain the marked increased in child abuse allegations in the Swedish crime statistics following the 1979 ban. They may also help explain data from a cross-country comparison of hospital admission codes. Gilbert et al. (52) analyzed hospital trends from 1979 to 2005/6 in six industrialized countries/regions. Sweden had a higher rate of hospital admissions than Canada for maltreatment injuries in children ages 1–11, even as it had a lower rate of violent child deaths.

Low parental authority also increases the risk of serious antisocial behavior related to early oppositional defiance. Unquelled “authority conflict” in childhood has been identified as one of the earliest (53) and strongest (54) predictors of criminal delinquency. The possibility that nonphysical punishment alone was insufficient to deter some Swedish children’s early antisocial tendencies may help explain Sweden’s marked increases in alleged minor-on-minor assaults following the ban.

Support for Canada’s Existing Age-Specific Approach

Since Canada legislated its current approach to physical discipline in 2004, some groups have continued to lobby for a ban on all spanking. As demonstrated in this commentary, the evidence underlying this well-intentioned advocacy is weak.

Methodologically sophisticated MAs of BPL studies suggest that the mean effect of customary spanking on subsequent child adjustment is near zero. More importantly, the only randomized studies of spanking have shown that back-up spanking can be part of an effective treatment protocol for those children at the highest risk of chronic delinquency. Admittedly, this latter evidence is dated. Social scientists should now be attempting to replicate these randomized studies and to design new ones to develop more effective discipline programs. If exact replication is too controversial, there are methods used in medical research to approximate unbiased causal effects when ethical concerns prohibit full randomization (e.g., randomized consent, encouragement designs, regression discontinuity designs).

Finally, we know of no evidence to suggest that total bans protect children more than Canada’s current approach. In contrast to the increased allegations of assaults on Swedish children following Sweden’s total ban on physical discipline, Canada’s child abuse rates have continued falling since Canada’s age-specific ban. We believe this is because Canada currently has one of the best approaches in the world. We hope that Canada will continue to lead with empirically-informed, developmentally appropriate legislation that empowers parents to help children become well-adjusted adults.