Predictors of Spanking across 60 Low- and Middle-Income Countries: A Multilevel Analysis

Background and Purpose: Ending violence against children is a major public health concern worldwide. Of particular importance to research and policy in recent decades is the reduction of spanking. Domestic and international research suggests that spanking is harmful for children, and 59 countries to date have banned spanking in all settings. While the effects of spanking are well established, the sociodemographic predictors of spanking have been less explored. Understanding factors that predict spanking will help inform prevention and intervention efforts. This study explored micro-, meso-, and macro-level predictors of spanking across 60 low- and middle-income countries (LMICs).

Methods: Data came from the 4th and 5th rounds of UNICEF Multiple Indicator cluster Surveys (N=337,849). Respondents were the head of the household (HH); if HH was unavailable, a spouse or caregiver was interviewed. Spanking was measured with a modified version of the Parent-Child Conflict Tactics Scale, which asked respondents whether they or anyone in the household had “spanked, hit, or slapped [child] on the bottom with a bare hand” in the past month (0=no, 1=yes). Child variables included child age in months and child sex (0=female, 1=male). Household variables included sex of the HH (0=female, 1=male); HH education (1=none [comparison category], 2=primary, 3=secondary-plus); number of household members; whether respondent believes children need physical punishment to be raised properly (0=no, 1=yes); household wealth score (standardized within each country); and whether the household was in a rural or urban community (0=urban, 1=rural). Country-level variables included child homicide rate (per 100,000 population over 12 months); human development (measured with the Human Development Index); unemployment rate; and gender inequality (measured with the Gender Parity Index). A multilevel logistic regression analysis was conducted in Stata, which provided odds ratio (OR) coefficients.

Results: A one-month increase in child age was associated with a decreased odds of spanking (OR = 0.96, p < .001), and the child being male was associated with an increased odds of spanking (OR = 1.06, p < .001). A one-person increase in household members, believing that children need physical punishment, and living in a rural area were associated with an increased odds of spanking (OR = 1.01, p < .001; OR = 2.55, p < .001; OR = 0.91, p < .001, respectively). A one-unit increase in household wealth score and the head of household having a secondary education or more were associated with a decreased odds of spanking (OR = 0.97, p < .001; OR = 0.93, p < .001, respectively). After accounting for micro- and meso-level factors, no country-level variables were associated with spanking.

Conclusions and Implications: Findings suggest that attitudes toward physical punishment were the strongest predictor of spanking, after accounting for many micro-, meso-, and macro-level variables. Intervention and prevention efforts may benefit from attempting to change individual level, or family level attitudes toward physical punishment in order to reduce spanking in LMICs. Families with a large number of household members, that have male children, and live in rural areas may also need attention in order to reduce spanking worldwide.