At least two out of three children younger than five living in low- and middle-income countries (LMICs) are exposed to physical punishment (also known as corporal punishment) in the home, early childhood care and education centers, or schools [1,2,3]. Extensive evidence from meta-analyses and reviews of cross-sectional and longitudinal studies—conducted mostly with samples from the USA and other high-income countries (HICs)—suggests that physical punishment relates to an array of detrimental child, adolescent, and adult outcomes [4,5,6,7]. In addition, the United Nations (UN) has firmly stated that physical punishment is a form of violence against children and a violation to children’s rights .
Despite evidence that physical punishment might be harmful, there is still some academic and extensive societal debate on the specific consequences of physical punishment. Most of the ongoing controversies are fueled by concerns of internal and external validity. First, there is disagreement on whether physical punishment causes worse child, adolescent, and adult outcomes versus whether confounding factors that plausibly influence both the likelihood of punishment and later individual outcomes (e.g., children’s initial levels of behavior) can fully explain observed associations between physical punishment and such outcomes [9,10,11]. Second, there has been a vast underrepresentation of samples from LMICs in prior meta-analyses and systematic reviews about physical punishment [5, 6, 9, 12]. This lack of studies in LMICs calls into question the generalizability of past findings given that 90% of children in the world live in LMICs and that physical punishment is prevalent in these countries . Finally, most meta-analyses to date have focused on physical punishment at home and less is known about the potential links between physical punishment in schools and child, adolescent, and adult outcomes, with some recent exceptions [13, 14].
Defining physical punishment considering a global perspective
Physical punishment is the use of physical force intended to cause some degree of pain or discomfort to correct or punish a child’s behavior [4, 15,16,17]. As such, physical punishment could vary in its frequency and severity, and could include actions like spanking, hitting a child with objects, or forcing a child to stay in an uncomfortable position, among others. A key feature of physical punishment is that adults have the intention to punish, correct, or control the child’s behavior [4, 16]. Some scholars have tried to distinguish between physical punishment (i.e., spanking) and abuse (e.g., hitting with hard objects or hitting the child frequently) based on how socially normative are different behaviors . Yet, these US-centered distinctions, which have been common in the literature [9, 19], are far from universal and are difficult to operationalize (e.g., there is not a clear threshold that divides physical punishment from abuse and the same behaviors are not normative across countries/cultures). Despite this, to date little is known about the specific forms of physical punishment that have been studied in research conducted with samples from LMICs, where, in contrast with US-centered perspectives, behaviors like hitting with objects are as socially normative as spanking [3, 20].
Theoretical and empirical links between physical punishment and child, adolescent, and adult outcomes
Developmental and educational theories and empirical evidence indicate that physical punishment can compromise children’s, adolescents’, and adults’ development, learning, and well-being through several biological and social mechanisms. According to traditional developmental and educational theories like social learning theory , social information processing theory , and attachment theory , by using physical punishment, caregivers and educators (intentionally or unintentionally) model aggression as a means to solve problems, inculcate in the child expectations of aggression, and erode the attachment bond, with downstream negative consequences on the relationship between children and their caregivers and children’s social-emotional development, behaviors, and mental health.
Contemporary neurodevelopmental models like the dimensional model of adversity  indicate that physical punishment can also affect children’s social-emotional and cognitive skills through neural mechanisms. The dimensional model of adversity predicts that exposure to threatening experiences, such as physical punishment, influence neural networks that facilitate the rapid identification of and response to environmental threats, including heightened response to negative emotional cues in brain regions that tend to underlie social and emotional processing and some cognitive functions [25, 26] Furthermore, the model predicts that the neural consequences of exposure to threatening experiences scale in relation to the severity of the threat involved. Finally, neurodevelopmental perspectives also indicates that threatening experiences might be more consequential if they occurred early in life, when the brain is more malleable and sensitive to experiences and contexts . Consistent with these models, nascent evidence from neuroimaging studies shows associations between experiencing physical punishment early in life and atypical brain structure and function [28, 29], in ways that may lead to downstream behavioral, emotional, and cognitive consequences.
These theoretical perspectives also align with a growing number of studies from LMICs that have shown consistent associations between physical punishment and individual outcomes, which do not seem to vary across settings. A rapid review of 42 studies using samples from LMICs concluded that there is robust evidence on the associations between physical punishment and individual social-emotional and mental health outcomes, but the evidence for cognitive outcomes is scarcer and mixed . In addition, studies using nationally representative samples for more than 49 countries across LMICs indicated that social normativeness does not modify the associations between physical punishment and individual outcomes [30, 31].
With these theoretical perspectives and empirical findings, we can hypothesize that: (1) physical punishment will likely be more strongly associated with social-emotional and mental health outcomes than with cognitive outcomes, (2) physical punishment could lead to stronger consequences if it occurs in early childhood relative to later in life (3) different forms of physical punishment could lead to different consequences that will likely scale in relation to the severity of threat involved, and (4) the same mechanisms linking physical punishment and individual outcomes might be relevant in different settings (e.g., in LMICs and HICs, between regions and countries, or if physical punishment takes place at home vs. schools).
Issues of internal validity and effect sizes
Despite consistent evidence on the associations between physical punishment and negative individual outcomes, establishing credible causal links between physical punishment and child, adolescent, and adult outcomes is not straightforward. It would be unethical to randomly assign children to physical punishment vs. non-physical punishment conditions and to date it has proven impossible to identify arguable exogenous sources of variation for physical punishment to conduct instrumental variables or regression discontinuity designs. For example, programs aimed at preventing physical punishment tend to include other components that might promote positive parenting (for example, content on the importance of play and/or emotional communication), therefore making them endogenous (see as examples the ACT Raising Safe Kids program , the Irie Toolbox [33, 34], and Parenting for Lifelong Health [35, 36]). Therefore, researchers have mostly relied on observational designs, with some exceptions using fixed effects models [11, 37] and matching methods, in an attempt to improve the internal validity of estimates [38, 39]. Yet, none of these approaches rules out all potential confounders (i.e., characteristics that might simultaneously affect physical punishment and outcomes), even if researchers have longitudinal data [38, 40]. Failing to control for most confounders will likely lead to overestimating the association between physical punishment and different outcomes. For example, maternal depression will likely have a positive correlation with physical punishment and negative association with individual outcomes. As such, failing to account for it may lead to an artificially inflated estimate of the relation between physical punishment and individual outcomes. A similar situation will arise with variables related to socioeconomic status (SES), self-efficacy, and even genetics.
Given these challenges to establishing causal links between physical punishment (and other developmental characteristics/exposures) and child, adolescent, and adult outcomes, researchers have increasingly recognized the importance of strong theory and assessing the sensitivity or robustness of estimates to the inclusion of covariates, multiple methodological approaches, and different identifying assumptions [11, 38, 41]. While even meta-analyses of physical punishment have recognized the importance of sensitivity/robustness checks in research on the consequences of physical punishment, to date all meta-analyses have included only one effect size per outcome per study, therefore making it impossible to test how robust are effect sizes within studies.
Besides allowing to assess the robustness of estimates, the inclusion of more than one effect size per outcome per study is useful to exploit all available data, increasing statistical power and leveraging informative within-study variability. Including multiple effect sizes in a meta-analysis is not entirely straightforward, as conventional meta-analytic methods assume independence of effect sizes. However, multiple effect sizes from the same study (e.g., different outcomes) are likely to be non-independent for different reasons, including correlations between sampling errors (due to the use of same sample) or nesting within the primary study . Under such circumstances, the results from conventional meta-analytic methods are inappropriate and could even be misleading [42, 43] and researchers have recommended the use of multilevel random-effects models to analyze datasets that include more than one effect size per study .
The proposed systematic review and meta-analysis
The proposed study has two main objectives. The first aim is to conduct a systematic review of the literature examining the associations between physical punishment in childhood or adolescence and child, adolescent, and adult outcomes in LMICs to describe the quantity and characteristics of studies, including geographic distribution, definitions of physical punishment used, methodological approaches, and main findings, among other basic characteristics. The second objective is to conduct a series of meta-analyses of the associations between physical punishment and child, adolescent, and adult outcomes in LMICs. We will conduct searches in multiple languages and databases to find more studies from LMICs in addition to those considered in prior meta-analyses. Furthermore, we will include all relevant effect sizes and use state-of-the-art multilevel random effects models to analyze the data. In addition, these meta-analyses will, for the first time in the literature, include physical punishment both at home and in schools. Finally, we will conduct moderation analyses to assess variability in the links between physical punishment and child, adolescent, and adult outcomes in LMICs.
Research questions and hypotheses
RQ1: What are the main characteristics of the published research on the associations between physical punishment and child, adolescent, and adult outcomes in LMICs regarding (a) geographic distribution, (b) the different forms of physical punishment that have been studied, and (c) methodological approaches?
We hypothesize that prior research (a) has not been widespread in different LMICs, but is scarce and has concentrated in specific countries, (b) has likely examined multiple forms of physical punishment, including spanking, hitting children with objects, and pinching the child, among others, and (c) has mostly relied on linear regression models with conventional covariate adjustment.
RQ2: What are the average associations between physical punishment and a range of child, adolescent, and adult outcomes in LMICs?
We hypothesize that all forms of physical punishment will associate with detrimental individual outcomes.
RQ3. Do the associations between physical punishment and child, adolescent, and adult outcomes vary by (a) different forms of physical punishment (e.g., spanking, hitting a child with objects), (b) developmental period at time of physical punishment [0–2 years; 3–5 years; 6–10 years; +10 years], (c) whether punishment occurred in homes or in schools, (d) region (i.e., East Asia and Pacific, Europe and Central Asia, Latin America & the Caribbean, Middle East and North Africa, North America, South Asia, Sub-Saharan Africa) or country income group (i.e., low-income, lower-middle income, upper-middle income), and (e) methodological approach (e.g., data structure and analytic strategy)?
We hypothesize that the association between physical punishment and individual outcomes will (a) vary according to different forms of physical punishment, (b) be stronger if physical punishment took place early in life relative to later developmental periods, (c) be similar regardless of whether punishment takes place in homes or in schools, (d) not vary across countries or regions, and (e) will be stronger in studies with weaker internal validity (e.g., cross-sectional, observational with poor covariates) relative to more internally valid studies (e.g., longitudinal, rich set of covariates, experimental, quasi-experimental) due to issues of selection bias.