We conducted secondary data analyses of survey data collected by YouGov (an international Internet-based market research and data analytics firm; https://today.yougov.com/about/about/) through their proprietary opt-in Internet panel. The YouGov opt-in survey panel is comprised of 1.2 million U.S. residents who have agreed to participate in YouGov’s web surveys. Panel members are recruited by a number of methods (e.g., web campaigns using internet banners, pop-ups, e-mails, permission-based email campaigns, partner-sponsored solicitations, telephone-to-web recruitment such as random-digit-dial-based sampling, and mail-to-web recruitment using voter registration based sampling) to help ensure diversity in the panel population and engage hard-to-reach populations. Participants are not paid to join the YouGov panel, but do receive incentives through a loyalty program to take individual surveys. The YouGov Panel Management team ensures the quality of the panel by removing panelists who speed through surveys or answer in ways that are determined to be obviously not genuine.

YouGov’s sampling frame is constructed using demographic information from the American Community Survey; voter registration data; and, computer internet, and smart-phone usage data from the Bureau of Labor Statistics’ Population Survey. YouGov recruits a random stratified sample from their sampling frame for specific surveys. Individuals responding to YouGov’s invitation are matched to the most similar member of the YouGov panel using propensity scores. YouGov has shown that their methods generate nationally representative samples and estimates that are closer to real percentages than other methods such as RDD surveys (Kellner, 2004; Rivers & Bailey, 2009). A more detailed description of YouGov’s U.S. panel and sampling strategy is available elsewhere (Rivers & Bailey, 2009).

For this survey, YouGov invited a random stratified sample (n = 5709) of their panel members and received 3566 responses (62.5%) of which 3043 (53.3% of those invited) completed the survey. Respondents completing the survey were then matched down to a sample of 2500 using YouGov’s sampling frame and matching methods as described above. There were some demographic differences between the YouGov sample and the U.S. population (Table 1); specifically, the YouGov sample has slightly more women, is older, less educated, and has somewhat lower incomes. There were also more women and participants with lower income among caregivers in the YouGov sample and caregivers were also younger compared to the U.S. population.

Overall, there were more similarities than differences between the three groups. The majority of parents (63%), regardless of race/ethnicity used CP with proportions lowest among Whites (59%) and highest among Latino (73%; Table 2). However, most used it seldom or sometimes, with variations in frequency of use by education levels statistically significant for Blacks, X² (8, N = 69) = 21.37, p = .006, and Whites, X² (12, N = 307) = 24.89, p = .015. Specifically, Blacks with high school degrees or less and Whites with graduate degrees or more were 1.3 and 1.4 times more likely to report never spanking.

Caregivers across race/ethnicity believe CP is used in higher proportions (89–91%) and with greater frequency by other parents in their community (Table 2). Variations in this perception by educational attainment was only statistically significant among White caregivers (X² (12, N = 314) = 24.34, p = .018), with the least educated and most educated believing more parents used CP.

The YouGov survey asked respondents to think of people whose opinion they respected and asked how strongly these people would agree or disagree (five-point Likert scale) with the statement “spanking your child on the bottom is a necessary part of parenting”. Sixty-nine percent of Blacks, 50% of Whites, and 47% of Latinos believed that people whose opinion they respected would agree or strongly agree with this statement (Table 2). Differences in agreement by education were statistically significant for Latinos, X² (12, N = 112) = 23.24, p p = .026, and Whites, X² (12, N = 307) = 25.77, p = .012. Specifically, Latinos with a graduate degree were much less likely to believe they ought to spank. Among Whites, those with some college but no degree and less education were more likely to believe they ought to spank.

Finally, Table 3 presents bivariate correlations between use of CP and perceived frequency of use and perceived opinion of people respondents respected (i.e., ought to spank). Overall, correlations between use of CP and perceived frequency of use (r = .35) and perceived opinion of CP by people respondents respected (r = .43) were statistically significant (p = .000). When examined by race/ethnicity, the strength of these correlations were higher for Blacks and Latinos and lower for Whites.

To better understand beliefs underlying the use of corporal punishment among those with lower education (i.e., the higher risk group), we conducted and analyzed data from 13 focus groups (n = 75; ranging from 2 to 9 participants) segmented by race/ethnicity (i.e., Black, Latino, or White), gender (i.e., mothers or fathers), and population density (i.e., rural or urban). Participants were recruited in five different states (i.e., Maine, New York, Connecticut, North Carolina, Florida) from a variety of community-based organizations and educational institutions that serve low-income families, including a Parent Resource Center, early childcare center, Head Start Center, elementary schools, and a YMCA. Parents were recruited via word of mouth from staff and through flyers that were distributed in printed versions to parents and posted on sites’ social media accounts. The recruitment flyer indicated: (a) that we were looking for parents of children from 0 to 5 years of age for a group discussion on different ways of managing children’s behaviors; (b) this discussion would last about an hour and a half; (c) onsite childcare would be available, if needed; (d) participants would be paid $50 for attending a group; and (e) the date, time, and place of this group discussion.

Interested parents were asked to respond to a screener which asked: (1) whether the participant was a mother or father; (2) race/ethnicity; (3) approximate household income; (4) level of educational attainment; (5) marriage status; (6) age of youngest child; and (7) contact information so that the research team could follow up to confirm their participation. To qualify for the study, parents had to have at least one child under six, a total household income below $60,000; and could not have obtained a degree beyond an Associate’s Degree (although two exceptions were made). Parents could provide their answers to this screener via phone, email, website, or as a paper version attached to the flyer. Eight to 10 eligible parents were invited to participate in each focus group with the goal of having eight parents per group.

Facilitators conducting focus group discussions had previous experience conducting research with low-income parents and were matched to the gender, language, and race/ethnicity of participating parents. Facilitators participated in a 3-h training session, conducted via video call with live mock participants, in which they practiced excerpts of the discussion guide.

Another trained researcher observed each discussion group and recorded relevant data such as nonverbal communication, body language, and participant discomfort that would not otherwise be captured by a recorder.

Before starting each focus group, participants read (or were read) the consent form explaining what they would be asked to do, that the discussion would be recorded, the risks and benefits of participating, how their privacy would be protected, and their right to choose or not to participate and asked to sign the form. During each session, parents were provided with notecards to record thoughts or opinions they were uncomfortable sharing aloud and a manila envelope with which to return their comments anonymously. The focus group research protocol was approved by CDC’s IRB.

We developed a focus group discussion guide to ensure that all groups covered the same topics. The discussion guide asked about methods parents used to discipline their children and probed the benefits and harms of different methods, especially hitting young children. It also asked how common hitting children was in their community, the situations in which it occurred, and the age when parents usually began hitting.

Focus group discussions were transcribed verbatim. Observers’ notes were added to the transcripts. Analysis was guided by principles of Grounded Theory (e.g. Pidgeon & Henwood, 2009). As such, concepts and categories emerged from the data (i.e., they were not preconceived or defined beforehand), and similarities and differences were identified by comparing groups. Four trained members of the research team identified the concepts and categories to code for following an iterative process with input from other members of the project team. First, the research team independently reviewed a sample of three transcripts, which included representation from all key audience variables, and generated a proposed list of codes. These code lists were compiled and compared at a qualitative debrief meeting, which resulted in an initial list of 51 unique codes. Further pruning based on a careful comparison of codes to the research objectives and feedback from the research team resulted in a list of 31 codes. Using this revised list, four of the researchers analyzed a sample transcript to pilot test the codebook. Inter-rater reliability was generally low across all research pairs at the sentence level (Cohen’s Kappa between .25 and .41). Further revisions were made to the codebook to: (a) eliminate redundant and unclear codes; (b) refine codes that were too broad or general; and (c) provide additional details and examples that would guide accurate coding. The result was a final codebook with 28 codes.

Two additional rounds of coder calibration were conducted using excerpts from different transcripts, with qualitative debrief meetings following each to discuss challenges interpreting specific codes, difficult passages, and inter-rater reliability. In each round, the research team made further clarifications to the codebook to better guide the interpretation and application of qualitative codes in order to increase coder reliability. Following the second round of calibration, an adequate Cohen’s Kappa was obtained for each pair of coders (over 0.55). After inter-rater reliability was reached the research team started coding full transcripts.

Three authors coded between 3 and 11 transcripts each, with two authors independently coding each transcript. Inter-rater reliability was high across the three pairs (weighted mean = .81 using the total number of sentences coded by each research pair) on the final coding, ranging from .78 to .82. Inter-rater reliability was also consistently high across all 27 codes in the full analysis sample, ranging from an average of .73–.91. Finally, three authors independently identified similarities and differences by race/ethnicity within each code. Cohen’s Kappa for each pair of coders ranged from .76–.88. Disagreements were discussed until a consensus was reached.

Focus group participants also expressed infrequent use of CP. Many parents in the focus group discussions reported only using CP after other strategies had failed. As this Black dad said, “That [CP] shouldn’t be your go-to.” However, for the most part, parents felt use of CP depended on the seriousness of the behavior. Examples of “serious” behavior included danger to the child (e.g., rushing out into a busy street, reaching out to touch a hot stove, hitting another child), tantrums, biting or pulling parent’s hair, being disrespectful with parent, persistent disobedience, and stealing. Parents also mentioned trying different strategies to find what works and recognizing that children are different. This White dad’s words exemplify this approach, “You have to find what works with each child. Not saying that you have to spank them, or that’s the other alternative, but not every punishment works the same with each kid all the way up through.”

Parents differed as to when it was appropriate to start using CP. The earliest age proposed was a White mom who mentioned slapping her 11-month-old on the hand. Another White mom said she had spanked her 1 year old “the first time she told me “no”. White fathers, on the other hand, thought three or four-year old children was when it was appropriate because “…once they get older and can understand why it’s bad”. Black parents agreed the time to start using CP was when children started walking because that was when “they start touching everything”. On the other hand, the ages mentioned by Latino parents for starting CP varied from one and half years old to 5 years old.

This mixed-methods study found CP to be used by the majority of parents albeit infrequently and mainly as a last resort with highest proportions among Latino caregivers. Across the three racial/ethnic groups, survey participants believed a greater proportion of parents used CP than what they themselves reported using and believed people whose opinion they respected would find it acceptable. Perceived frequency of use of CP in the community and acceptability were significantly associated with use of CP. We also heard that Blacks, Latinos and Whites believe parents use CP because that was how they had been parented and maybe had no knowledge of other disciplinary strategies. However, all groups mentioned CP’s negative effects and identified non-physical and more positive strategies.

Before discussing potential implications of these findings, several limitations with both the survey and focus groups must be acknowledged. The demographics of YouGov’s random sample from their opt-in internet panel was slightly different from the U.S. population, even after weighting the sample using demographic information from various sources and propensity scores and therefore, the findings may not be nationally representative. In addition, just over half of those invited to participate responded and completed the survey so self-selection bias may be an issue. Survey responses are also subject to social desirability biases, which were not measured in this survey. As for the focus groups, although these were conducted in five different states, these were limited to the eastern coast thus our findings may not be relevant in communities in other geographic regions with different cultural norms, values, and beliefs. In addition, our Latino focus group participants in the rural area were mostly of Mexican origin; in both urban and rural Latino focus groups there was some variability in the number of years in the U.S. Given great heterogeneity among Latinos based on country of origin and acculturation status, our findings may not be generalizable to some Latino populations. For example, Lee and Altschul (2015) found that immigrant Latino parents were less likely to spank their young children than Latinos born in the U.S. Future research might focus on similar questions with other Latino communities with different levels of acculturation.

Despite these limitations, these findings are supported by several strengths of the study such as its use of mixed methods; large sample size for the survey; homogeneity within focus groups in terms of race/ethnicity, gender, and level of education, diversity across focus groups in terms of race/ethnicity, urban/rural, and geographic region; matching of focus group facilitators by race/ethnicity and gender; and high rates of reliability in the coding of focus group transcripts.

The survey findings show higher proportions of caregivers using CP compared to other studies (Regalado, et al., 2004; Slade & Wissow, 2004; Wissow, 2001) but these other studies examined the issue in younger age groups. Contrary to most studies which find use of CP to be higher among Blacks (Berlin et al., 2009; Grogan-Kaylor, & Otis, 2007; Lorber et al., 2011; MacKenzie et al., 2011; Regalado et al., 2004; Wissow, 2001), we found Latinos reporting use of CP in higher proportions than Blacks or Whites. This may be an artifact of US-born Latinos having greater access to the internet (Brown, López, & López, 2016) and therefore being able to opt in to participate in online surveys more than foreign-born Latinos. As mentioned before, foreign-born Latino parents are less likely to spank their young children than Latinos born in the U.S. (Lee, & Altschul, 2015).

On the other hand, our qualitative findings are consistent with other research showing CP is widely accepted in certain situations (Lubell, Lofton, & Singer, 2008; Taylor, Hamvas, & Paris, 2011). Similarly, our focus group participants also mentioned situational factors in which CP might be used such as stress, seriousness of the transgression committed by the child, and repeated child misbehavior as well as perceived high frequency of use in their community (Taylor et al., 2011). However, our focus group participants also noted some negative effects of CP such as teaching children that hitting is acceptable and noted many positive, non-physical alternatives. Interestingly, focus group participants perceived use of CP declining and some referred to its use as “outdated”. This might suggest that parents are already learning about better alternatives for disciplining children. Changing social norms is a promising strategy for reducing CP and preventing child abuse (Fortson et al., 2016). Educational campaigns to change social norms might take advantage of this perception of CP being outdated and newer methods being more effective.

Randomized controlled trials (RCTs) have shown the effectiveness of delivering information on the risks of CP and alternative non-violent disciplinary strategies in home visits (Minkovitz et al., 2003) and primary care settings (Chavis et al., 2013; Holden, Brown, Baldwin, & Croft Caderao, 2014; Scholer, Hamilton, Johnson, & Scott, 2010), even more so when discussed with a clinician (Scholer, Hudnut-Beumler, & Dietrich, 2011). However, mass media educational campaigns might be more effective at changing social norms and reducing support for CP at the population level. We are not aware of any mass media campaign in the US focused on changing social norms around CP. However, there is increasing evidence for the effectiveness of mass media campaigns to change behaviors in other areas of public health such as tobacco, nutrition, physical activity, birthrate reduction, and road safety, particularly when campaigns are combined with complementary policies that support behavior change such as an excise tax or other disincentives to engaging in the behavior (Wakefield, Loken, & Hornik, 2010). Nevertheless, campaign messages need to be tested for their effectiveness on changing social norms and behavior change. For example, a pre-post evaluation showed a media campaign in Canada reduced approval of CP but may not have reduced the use of CP (McKeown, 2006). Perhaps testing messages in RCTs before engaging in a campaign can help identify more effective messages (e.g., Evans, Falconer, Khan, & Ferris, 2012).

In addition to testing messages, credible messengers for different racial and ethnic groups might also be tested. Focus group participants reported trusting other parents’ recommendation more than doctors’ or experts but Latinos and Whites also mentioned trusting academic institutions. A RCT could establish which messenger is most effective in changing social norms and reducing the use of CP. Such efforts could help reduce the frequency of CP and its negative consequences.

In sum, we found many more similarities than differences between Black, Latino, and White parents in regards to their use, perceived use, and necessity of use regarding CP (i.e., “ought to use”). The findings suggest social norms change efforts might focus on parents with less education and influencing perceptions around whether they ought to spank. Changing social norms around whether parents ought to use CP may encourage some parents to seek information on alternative strategies. Instead of CP, positive parenting strategies could be more widely disseminated (e.g., Centers for Disease Control and Prevention, 2014). However, while parenting practices are important for raising well-behaved children, it may also be important to recognize, understand, and improve the context in which some parents are raising their children. Structural interventions such as strengthening household economics and family-friendly work have the potential to improve the context and increase safe, stable, nurturing relationships and environments for all children (Fortson et al., 2016).