5.3 Sample, variables and methods
5.3.2 Health measures
Child health. Finding appropriate measures of a child’s health status is a challenge. Health has many dimensions, such as mental, physical, chronic conditions, environmental conditions, nutrition and injuries. In developing countries, infant mortality rates, anthro- pometric measures and vaccination data are often used as benchmarks to estimate child health because they are easy to measure and highly correlated with the full health status. Studies on Western industrialized countries use LBW as an indicator of poor health at birth (see the studies cited in sections 5.1 and 5.2). Alternative measures of children’s health are bed days and hospitalization episodes. As there is still no operational, global definition of child health, it might be useful to use various measures once they become available.
This is a crucial advantage of the data we used here. The SOEP allows us to observe different types of health measures for the children (for other SOEP-based studies using
6
The sample size within one cross-section varies because of missing observations with regard to some health conditions.
similar child health measures, see for instance, Dunkelberg and Spiess (2009) and Cawley and Spiess (2008)). For all age cohorts, we observe anthropometric (health) measures such as weight and height of the child. Anthropometric health measures have the advan- tages that they are easy to administer and that potential measurement errors are more likely to be random (Kebede, 2005). Weight and height of the children are reported by the mother and not measured by an expert. Thus there might be reporting errors (see for instance, Strauss and Thomas (1996)), but we argue that the reporting error is low and random given the specific features of the German health care system. In Germany, preventive medical check-ups are offered to children on a regular basis form birth up to the age of five. They are free of charge. The weight and the height of the child are taken by experts at each check-up and documented in a medical record booklet that is kept by the family. 98% of SOEP children have had such regular check-ups. The average weight (height) at birth is 3338.39 gram (51.13 cm). Both measures increase with age. A 5-6 year old child has an average weight and height of about 21.46 kg and 116.57 cm (see Table 5.2).
Table 5.2: Descriptive statistics of children’s and parental health measures
0-19 months 26-47 months 62-81 months
mean std.dev. mean std.dev. mean std.dev.
child’s health
weight (in kg) 3.34 (0.60) 14.02 (2.53) 21.46 (4.07)
height (in cm) 51.13 (3.17) 93.25 (11.44) 116.57 (6.80)
disorder (1=yes) 0.05 (0.22) 0.46 (0.50) 0.42 (0.50)
worried about my child’s health1 3.43 (0.77) 3.46 (0.76) mother’s health
height (in cm) 167.26 (6.68) 167.36 (6.62) 167.95 (6.57)
weight (in kg) 68.36 (12.81) 68.16 (12.65) 69.82 (13.37)
self-rated health status2 (t − 1) 3.79 (0.75) 3.70 (0.80) 3.60 (0.74)
smoking 0.18 (0.38) 0.21 (0.41) 0.19 (0.39)
father’s health
height (in cm) 180.48 (7.11) 180.36 (6.95) 180.74 (7.54)
weight (in kg) 84.06 (14.35) 85.16 (14.65) 86.35 (15.88)
self-rated health status2 (t − 1) 3.85 (0.75) 3.75 (0.80) 3.68 (0.82)
smoking 0.36 (0.48) 0.34 (0.48) 0.30 (0.46)
Note: Standard deviations are in parentheses. 1worried about my child’s health varies from 1=(applies fully), 2=(applies more), 3=(applies less) to 4=(does not apply); 2Self-rated health status varies from 1=(bad), 2=(poor), 3=(satisfactory), 4=(good) to 5=(very good). For the 0-19 months old we observe the weight and height at birth.
Source: SOEP 2003-2008. Own calculations.
Moreover, in all cross-sectional samples, mothers were asked about any disorders their child had, for example, motor impairments or asthma (see Appendix Table D.1 for a detailed
description). We compute a dummy variable instead of using all dimensions separately, because the share of children with any kind of a disorder at birth is quite low. It takes the value of one, if the child has at least one disorder and zero, if the child has no disor- der. Again, we would expect that given the regular medical check-ups in Germany during the first five years of a child’s life, measurement errors are low and random. Descriptive statistics in Table 5.2 show that among newborns, the share of children with any kind of a disorder is only 5%, at the age of 2-3 years nearly every second child has at least one disorder, while three years later, almost 60% of the children suffer from at least one disorder. The main disorders are asthma, bronchitis and middle-ear inflammations. Some aspects of this health measure also reflects common illnesses during early childhood and explain why the percentage of children with any kind of a disorder at the age of two and older is eight times higher than for newborns.
Apart from these ‘self-reported health measures’, we also use ‘self-rated’ health measures (in the strict sense these are mother-rated measures of the child health): The mothers in the SOEP were asked in the first and second Mother and Child Questionnaire whether they were worried about their child’s health.7 They could specify to what extent the state- ments apply, ranging from ‘applies fully’ (1) to ‘does not apply’ (4). For the newborns as well as for the 2-3 year olds, the share of mothers who assessed the health of their children as ‘good’ or ‘very good’ is about 90 percent.
The SOEP also contains data on the utilization of health care services. The mothers were asked for the number of doctor visits and hospital stays. However, it is difficult to disentangle whether both variables capture the health of the child or whether they simple reflect the health behavior or risk aversion of the mother. Thus, we do not apply models based on these variables.
Parental health. As far as parental health is concerned we also use different health mea- sures (compare for such an approach based on the SOEP, Schwarze, Andersen, and Anger (2000)). First, the SOEP covers anthropometric health measures for all adults living in the household every two years. We consider the height and the weight of the parents: the former for the cross-sectional models, the latter for the panel models (Poskitt, 1995). Table 5.2 indicates that the height of the mother as well as the height of the father are sta- ble over time, which is one indication that the reporting error of this self-reported health
measure is quite low. The average height of the mothers is about 167.26 cm, while fathers are 13 cm taller on average. Regarding the weight of the mother, we do not observe a so- called pregnancy effect, as the weight only differs slightly between the three cross-sectional samples. In contrast, the weight of the father increases (slightly) over time.
The data also contain a self-rated health measure of the parents:8 Parents were asked how they assess their own current health status, using a scale from 1=(bad), 2=(poor), 3=(satisfactory), 4=(good) to 5=(very good). One might argue that self-rated health is not a valid health measures. However, the empirical health literature has shown that self- rated health measures are highly correlated with ‘self-reported health measures’ (see for instance Singh-Manoux et al. (2007) or Schwarze et al. (2000)).9 We use this indicator, because we would expect that is associated, at least to some extent, with the true health of the child. On average, the parents’ self-rated health lies on the upper bound of the scale, but the average value slightly decreases between the three cross-sectional samples for mothers as well as for fathers. For all three periods, it seems as if fathers assessed their own health better than the mothers.
Finally, we observe aspects of the health behavior of the parents: every two years, they were asked whether they smoked or not. In general, mothers and fathers differ with re- spect to their smoking behavior. For the newborns we observe that the percentage of fathers who smoke is twice as high as the percentage of mothers who smoke. But the percentage of nonsmoking mothers with very young children is three percent lower than that of nonsmoking mothers with 2-3 year olds. It may be that mothers stop smoking and then start again when their children are slightly older.10 Given this broad set of health measures, we do several estimations, using anthropometric and self-rated health measures as well as the information on any health disorders of the child.
Maternal reports of child health might depend on the mothers’ own health. Healthier mothers, for example, might tend to overestimate their children’s health. Given the prob- lem of reverse causality between parental ratings of children’s health and the self-rated
8In the main SOEP, self-rated health is measured by an internationally widely accepted scale. 9
Our descriptive evidence indeed supports this hypothesis: self-rated health is significantly correlated with both the anthropometric health and the self-reported health.
health of the parents, we use the self-rated health information of the parents one year before the health of the child was assessed.