Are working moms causing child obesity?

Childhood obesity is a growing problem, both physically and conceptually. But what are its causes? Specifically, people have argued that an increasing participation of women in the job market made mothers more absent from their children’s lives. This in turn allowed children to basically do much more what they wanted to do, e.g. watching TV, playing video games and eating fast food, sweets and snacks.

A plausible-sounding hypothesis but is it true? If a relationship exists, through what channel does maternal employment affect childhood obesity? And do the potential effects of maternal employment on childhood obesity depend on the mother’s education?

I ran into a paper by Fertig, Glomm and Tchernis (2006) that investigates these questions. The theoretical background is quite intuitive and is well summed up in the first paragraph of this post, so I will get straight to the methodology.

Using data on about 2,500 children the authors estimate the effect of certain activities on BMI,

$\ln{BMI_i} = \alpha_{0,A} + \alpha_{1,A} A_i + \alpha_{2,A} X_i + \epsilon_{A,i},$

where A is the activity in question and X denotes a bunch of control variables. Activities included for instance the number of meals per day or amount of TV watched. Then the following equations were estimated,

$A_i = \gamma_{0,A} + \gamma_{1,A} \ln{MWH}_i + \gamma_{2,A} X_i + u_{A,i},$

where MWH is maternal working hours.

So with the first equation, the authors basically estimate whether activity A has an effect on BMI (i.e. whether it contributes to obesity), then with the second equation they check whether activity A is influenced by maternal working hours. This way, one can decompose the channels through which maternal employment could influence obesity. The advantage of using a seperate equation for each activity is to avoid creating a multicollinearity problem. And if BMI was just regressed on MWH, one could not identify the channels.

Using the above specification, the effect of maternal employment on BMI is given by

$\frac{\partial{\ln{BMI}}}{\partial{\ln{MWH}}} = \frac{\partial{\ln{BMI}}}{\partial{A}} \frac{\partial{A}}{\partial{\ln{MWH}}} = \sum\limits_{A=1}^n \alpha_{1,A} \gamma_{1,A}.$

Each term in the resulting sum is the effect of one channel. This way, one can determine whether maternal employment causes higher childhood obesity through a certain channel (= activity) or not. The authors use OLS to estimate the equations where appropriate, for equations with count data they use a Poisson maximum likelihood estimation and for dichotomous variables they use probit. To estimate the standard errors of the coefficient products (i.e. alpha times gamma above), they use SURE.

So which activities have a significant effect on BMI? For the full sample only watching TV has a positive effect on BMI, while number of meals, being breastfed, miscelleanaeous passive activities and time spent on eating have negative effects at 1% or 5%. At 10% doing chores interestingly has a positive effect and socalizing has a negative one.

There is some difference between the children of mothers with low (< 12 years) and high (> 12 years) education. For children with less educated mothers, attending school and playing sports have a significant negative effect on BMI; watching TV is not significant, and shopping has a significant positive effect on the 10% level. For children with more educated mothers only time spent on eating, number of meals and TV are significant, the directions agree with the full sample results.

Why do we see this different pattern between these two groups? Well, to answer this we should see the effects of maternal employment on each of the activities. It seems that children of mothers with less education tend to be at school more, travel more and be in child care more, when the mother works. For children of more educated mothers, it is rather watching TV and time spent in child care that increase (and the latter more so than for the other group), when the mother works.

So let us see whether maternal employment significantly influences BMI and if yes, through which channels. For the full sample there are two channels which are significant, both indicating that maternal employment increases childhood obesity. Apparently, if a mother is employed, the child will eat fewer meals and watch more TV thus significantly increasing the chance of obesity.

Interestingly, the effect for children of less educated mothers is negative. I.e. the more a (less educated) mother works, the lower her child’s BMI will be. And this is through the channel of attending school. As established above these children increase their time spent in school as their mother works more and this helps push BMI down. For children with well-educated mothers it is only through the number of meals that maternal employment increases BMI.

Nice, so the authors established a significant relationship between maternal employment and childhood obesity, is our question answered? Not quite. Because the magnitude of the effect has to be studied as well. The authors note “the total elasticity of child’s BMI with respect to mother’s working hours is around one percent for all groups“. This means that if a mother works 20 hours more than previously, then the child’s BMI is expected to increase by 0.2 points, which is basically negligible (probably BMI naturally changes throughout the day/week that much).

So yes, maternal employment does have an effect on childhood obesity, but the effect is marginal at best. The authors say this agrees with past research and more recent papers I found also seem to confirm this. Some authors find no relationship, while others do. A review of the literature suggests only marginal effects as well (with the most extreme result saying that working 10 hours more per week will increase the probability of a child being obese by 11%; other research, however, suggests that this probability is somewhere between 0.5% and 4%). This review also cites a paper stating that changes in labor force participation contributed 10.4% to the rise in childhood obesity. There is also evidence that the timing of maternal employment may be critical.

The overall conclusion seems to be that maternal employment does increase childhood obesity, mainly via the channels of having fewer meals (which are either bigger or complemented by snacking in-between meals) and watching more TV. But even this weak relationship does not hold for children of less educated mothers. Clearly though, these channels are such that sufficient planning and an appropriate child care center or relative can easily eliminate them (and then the children can even take advantage of allo-parenting).

My verdict is that sure maternal employment made mothers spend less time on their children on average, but the size of the effect was not catastrophic on an aggregate scale. If we want to combat childhood obesity we must look at other factors.

Advertisements