Results
Descriptive Results
The effects of the recession can be seen in the variation of mean levels of the three economic well-being variables across the three waves for all subsamples of GUI98 (see Table
1) and GUI08 (see Table
2). For GUI98, strain increases between Wave 1 and 2 before going down slightly between Wave 2 and 3, although it does not return to pre-recession levels. Similarly, there is a sharp increase in the proportion of households experiencing some type of material deprivation during the first years of the recession which only decreases moderately in the aftermath of the recession. The average household income drops substantially in the first years after the start of the recession and continues to decline in the later years albeit not as rapidly.
Table 1
GUI98: means and standard deviations for our key variables by sex
Strain | 2.93 | 1.09 | 3.70 | 1.17 | 3.49 | 1.24 |
Log income | 9.76 | 0.50 | 9.57 | 0.52 | 9.49 | 0.56 |
Material deprivation | 0.11 | 0.32 | 0.25 | 0.43 | 0.23 | 0.42 |
Internalized SDQ | 3.21 | 3.00 | 2.88 | 2.93 | 2.91 | 2.79 |
Externalized SDQ | 4.89 | 3.44 | 4.31 | 3.49 | 3.71 | 3.22 |
Strain | 2.99 | 1.12 | 3.80 | 1.17 | 3.57 | 1.21 |
Log income | 9.68 | 0.58 | 9.52 | 0.55 | 9.42 | 0.56 |
Material deprivation | 0.15 | 0.36 | 0.30 | 0.46 | 0.27 | 0.44 |
Internalized SDQ | 3.50 | 2.99 | 3.20 | 2.95 | 3.94 | 3.16 |
Externalized SDQ | 4.11 | 3.24 | 3.53 | 3.23 | 3.13 | 2.96 |
Table 2
GUI08: means and standard deviations for our key variables by sex
Strain | 3.75 | 1.14 | 3.91 | 1.13 | 3.38 | 1.11 |
Log income | 9.66 | 0.53 | 9.62 | 0.53 | 9.72 | 0.54 |
Material deprivation | 0.31 | 0.46 | 0.35 | 0.48 | 0.23 | 0.42 |
Internalized SDQ | 2.69 | 2.25 | 2.72 | 2.62 | 3.28 | 3.18 |
Externalized SDQ | 5.72 | 3.41 | 5.38 | 3.62 | 4.99 | 3.73 |
Strain | 3.81 | 1.12 | 3.95 | 1.11 | 3.40 | 1.08 |
Log income | 9.62 | 0.53 | 9.59 | 0.52 | 9.71 | 0.52 |
Material deprivation | 0.31 | 0.46 | 0.35 | 0.48 | 0.23 | 0.42 |
Internalized SDQ | 2.52 | 2.18 | 2.56 | 2.44 | 3.20 | 2.96 |
Externalized SDQ | 5.20 | 3.34 | 4.50 | 3.22 | 3.77 | 3.14 |
In the GUI08 cohort, strain increases from Wave 2 to 3, coinciding with the height of the economic recession, before declining to its lowest level in Wave 5 in the aftermath of the recession. This pattern is mirrored in changes in income and material deprivation, with income declining from Wave 2 to 3, before increasing at Wave 5 to its highest level, while material deprivation increases from Wave 2 to 3, before declining at Wave 5. Differences in internalized behavioral difficulties between boys and girls are small, while boys hold higher levels of externalized behavioral difficulties than girls at each time point.
In GUI98, differences in the average levels of internalized and externalized behavioral difficulties can be observed for boys and girls. On average, girls have higher levels of internalized behavioral difficulties than boys. Conversely, boys exhibit greater externalized behavioral difficulties than girls. Despite variability in individual trajectories, behavior seems to improve on average as the study children grow older with the means of externalized and internalized difficulties decreasing over time.
In GUI08, boys exhibit higher levels of both internalized and externalized behavioral difficulties than girls at all time points. However, the difference in internalized behavioral difficulties between boys and girls narrows over time, with the difference at its smallest at Wave 5.
Within-wave correlations are all in the expected directions in both GUI98 (Table
3) and GUI08 (Table
4). In both cohorts, higher household income is associated with fewer behavioral difficulties at each wave. Additionally, material deprivation and greater financial strain are related to increased internalized and externalized behavioral difficulties at each time point. Furthermore, the three economic well-being measures show statistically significant, albeit only moderately strong correlations. This suggests that these dimensions of economic well-being are indeed related but not necessarily the same.
Table 3
GUI98 within-wave correlations between the key variables by sex at each wave (male below diagonal; female above the diagonal)
Strain W1 | 1 | −0.43*** | 0.33*** | 0.21*** | 0.19*** |
Log income W1 | −0.45*** | 1 | −0.24*** | −0.15*** | −0.12*** |
Material deprivation W1 | 0.29*** | −0.15*** | 1 | 0.15*** | 0.11*** |
Internalized SDQ W1 | 0.20*** | −0.11*** | 0.15*** | 1 | 0.36*** |
Externalized SDQ W1 | 0.18*** | −0.11*** | 0.10*** | 0.34*** | 1 |
Strain W2 | 1 | −0.47*** | 0.48*** | 0.19*** | 0.17*** |
Log income W2 | −0.46*** | 1 | −0.32*** | −0.14*** | −0.16*** |
Material deprivation W2 | 0.47*** | −0.30*** | 1 | 0.14*** | 0.12*** |
Internalized SDQ W2 | 0.17*** | −0.10*** | 0.15*** | 1 | 0.41*** |
Externalized SDQ W2 | 0.14*** | −0.12*** | 0.10*** | 0.35*** | 1 |
Strain W3 | 1 | −0.46*** | 0.49*** | 0.14*** | 0.13*** |
Log income W3 | −0.47*** | 1 | −0.34*** | −0.11*** | −0.12*** |
Material deprivation W3 | 0.48*** | −0.32*** | 1 | 0.16*** | 0.12*** |
Internalized SDQ W3 | 0.14*** | −0.11*** | 0.13*** | 1 | 0.43*** |
Externalized SDQ W3 | 0.11*** | −0.09*** | 0.10*** | 0.35*** | 1 |
Table 4
GUI08 within-wave correlations between the key variables by sex at each wave (male below diagonal; female above the diagonal)
Strain W1 | 1 | −0.46*** | 0.46*** | 0.09*** | 0.13*** |
Log income W1 | −0.48*** | 1 | −0.35*** | −0.12*** | −0.16*** |
Material deprivation W1 | 0.50*** | −0.37*** | 1 | 0.08*** | 0.09*** |
Internalized SDQ W1 | 0.13*** | −0.14*** | 0.13*** | 1 | 0.31*** |
Externalized SDQ W1 | 0.13*** | −0.14*** | 0.11*** | 0.32*** | 1 |
Strain W2 | 1 | −0.42*** | 0.49*** | 0.09*** | 0.11*** |
Log income W2 | −0.44*** | 1 | −0.34*** | −0.09*** | −0.14*** |
Material deprivation W2 | 0.50*** | −0.34*** | 1 | 0.12*** | 0.12*** |
Internalized SDQ W2 | 0.13*** | −0.11*** | 0.14*** | 1 | 0.35*** |
Externalized SDQ W2 | 0.14*** | −0.12*** | 0.14*** | 0.34*** | 1 |
Strain W3 | 1 | −0.45*** | 0.45*** | 0.16*** | 0.12*** |
Log income W3 | −0.47*** | 1 | −0.33*** | −0.09*** | −0.11*** |
Material deprivation W3 | 0.45*** | −0.33*** | 1 | 0.16*** | 0.12*** |
Internalized SDQ W3 | 0.17*** | −0.12*** | 0.16*** | 1 | 0.42*** |
Externalized SDQ W3 | 0.13*** | −0.12*** | 0.12*** | 0.45*** | 1 |
Multivariate Results
Fixed effects models are run separately using the GUI98 and GUI08 samples stratified by sex. For the GUI98 cohort, results are largely similar for boys and girls but only partly confirm our first hypothesis that economic well-being is related to increased behavioral difficulties (see Table
5). Against expectations, household income does not seem to be associated with the behavior of the young person. This is the case for both boys and girls, and even holds in the income-only model. Conversely, children who grow up in households with greater financial strain appear to have more externalized behavioral difficulties. Increased financial strain is predictive of increased externalized behavioral difficulties for boys and girls, even after adjusting for material deprivation and income. For material deprivation, distinct effects can be seen by sex. Material deprivation is not linked to any internalized behavioral difficulties for either boys or girls. However, it is associated with greater externalized behavioral difficulties for boys, although this is not the case for girls.
Table 5
Results from the fixed effects regression predicting internalized and externalized behavioral difficulties for boys and girls, GUI98
Wave (baseline, Wave 1) | | | | | | | | |
Wave 2 | −0.433*** | −0.375*** | −0.366*** | −0.424*** | −0.711*** | −0.639*** | −0.602*** | −0.721*** |
| (0.087) | (0.079) | (0.079) | (0.088) | (0.086) | (0.076) | (0.080) | (0.087) |
Wave 3 | −0.417*** | −0.377*** | −0.368*** | −0.402*** | −1.277*** | −1.232*** | −1.206*** | −1.281*** |
| (0.097) | (0.091) | (0.092) | (0.097) | (0.096) | (0.091) | (0.097) | (0.100) |
Strain | 0.083 | | | 0.09 | 0.154** | | | 0.134* |
| (0.052) | | | (0.052) | (0.055) | | | (0.057) |
Material deprivation | | 0.045 | | −0.003 | | 0.343** | | 0.267* |
| | (0.126) | | (0.123) | | (0.125) | | (0.126) |
Log income | | | 0.017 | 0.069 | | | −0.044 | 0.056 |
| | | (0.133) | (0.136) | | | (0.129) | (0.130) |
Constant | 2.485*** | 2.752*** | 2.595 | 1.796 | 4.382*** | 4.812*** | 5.322*** | 3.845** |
| (0.332) | (0.266) | (1.360) | (1.456) | (0.310) | (0.262) | (1.287) | (1.348) |
Observations | 2792 | 2792 | 2792 | 2792 | 2792 | 2792 | 2792 | 2792 |
Wave (baseline, Wave 1) | | | | | | | | |
Wave 2 | −0.398*** | −0.345*** | −0.316*** | −0.395*** | −0.659*** | −0.576*** | −0.574*** | −650*** |
| (0.086) | (0.077) | (0.078) | (0.088) | (0.084) | (0.079) | (0.083) | (0.088) |
Wave 3 | 0.353*** | 0.389*** | 0.420*** | 0.370*** | −1.035*** | −0.976*** | −0.974*** | −1.021*** |
| (0.097) | (0.093) | (0.102) | (0.104) | (0.094) | (0.090) | (0.106) | (0.108) |
Strain | 0.092 | | | 0.089 | 0.103* | | | 0.112* |
| (0.052) | | | (0.053) | (0.050) | | | (0.054) |
Material deprivation | | 0.14 | | 0.102 | | 0.004 | | −0.052 |
| | (0.127) | | (0.128) | | (0.129) | | (0.130) |
Log income | | | 0.05 | 0.101 | | | 0.007 | 0.048 |
| | | (0.142) | (0.142) | | | (0.177) | (0.177) |
Constant | 3.755*** | 4.017*** | 3.566** | 2.771* | 3.763*** | 4.093*** | 4.031* | 3.284 |
| (0.280) | (0.239) | (1.355) | (1.374) | (0.263) | (0.198) | (1.704) | (1.752) |
Observations | 2956 | 2956 | 2956 | 2956 | 2956 | 2956 | 2956 | 2956 |
Contrary to the first hypothesis and in line with model results from the GUI98 cohort, fixed effects models run using the younger GUI08 cohort indicate that household income is not associated with behavioral difficulties (Table
6). However, in contrast to the GUI98 models where increased financial strain is found to be associated with increased externalized behavioral difficulties in both boys and girls, financial strain is not associated with either internalized or externalized behavioral difficulties among the younger age sample of GUI08. An increase in material deprivation is found to be associated with an increase in boys’ internalized behavioral difficulties, even after adjusting for income and financial strain. However, this association is not present for girls, nor for externalized behavioral difficulties.
Table 6
Results from the fixed effects regression predicting internalized and externalized behavioral difficulties for boys and girls, GUI08
Wave (baseline, Wave 1) | | | | | | | | |
Wave 2 | 0.03 | 0.029 | 0.041 | 0.025 | −0.361*** | −0.370*** | −0.372*** | −0.370*** |
| (0.055) | (0.054) | (0.054) | (0.055) | (0.074) | (0.074) | (0.074) | (0.074) |
Wave 3 | 0.620*** | 0.613*** | 0.594*** | 0.626*** | −0.775*** | −0.758*** | −0.752*** | −0.763*** |
| (0.076) | (0.073) | (0.074) | (0.076) | (0.080) | (0.080) | (0.080) | (0.080) |
Strain | 0.064 | | | 0.046 | −0.022 | | | −0.057 |
| (0.046) | | | (0.049) | (0.053) | | | (0.056) |
Material deprivation | | 0.245* | | 0.226* | | 0.121 | | 0.133 |
| | (0.100) | | (0.105) | | (0.114) | | (0.115) |
Log income | | | 0.026 | 0.077 | | | −0.221 | −0.238 |
| | | (0.113) | (0.113) | | | (0.127) | (0.135) |
Constant | 2.225*** | 2.386*** | 2.235* | 1.462 | 5.373*** | 5.230*** | 7.411*** | 7.746*** |
| (0.392) | (0.316) | (1.083) | (1.145) | (0.408) | (0.351) | (1.240) | (1.407) |
Observations | 3620 | 3620 | 3620 | 3620 | 3620 | 3620 | 3620 | 3620 |
Wave (baseline, Wave 1) | | | | | | | | |
Wave 2 | 0.042 | 0.043 | 0.046 | 0.042 | −0.707*** | −0.696*** | −0.701*** | −0.705*** |
| (0.056) | (0.056) | (0.055) | (0.056) | (0.072) | (0.072) | (0.072) | (0.072) |
Wave 3 | 0.680*** | 0.677*** | 0.661*** | 0.672*** | −1.408*** | −1.439*** | −1.424*** | −1.410*** |
| (0.070) | (0.068) | (0.069) | (0.071) | (0.081) | (0.079) | (0.081) | (0.081) |
Strain | 0.013 | | | 0.023 | 0.055 | | | 0.065 |
| (0.041) | | | (0.042) | (0.049) | | | (0.050) |
Material deprivation | | 0.026 | | 0.025 | | −0.095 | | −0.133 |
| | (0.103) | | (0.104) | | (0.115) | | (0.116) |
Log income | | | 0.164 | 0.178 | | | −0.085 | −0.065 |
| | | (0.112) | (0.115) | | | (0.133) | (0.138) |
Constant | 2.206*** | 2.249*** | 0.687 | 0.462 | 4.680*** | 4.912*** | 5.701*** | 5.289*** |
| (0.355) | (0.324) | (1.156) | (1.226) | (0.388) | (0.362) | (1.406) | (1.494) |
Observations | 3588 | 3588 | 3588 | 3588 | 3588 | 3588 | 3588 | 3588 |
In line with the second hypothesis, results varied substantially by age cohort. However, while the anticipated association between financial strain and child behavioral difficulties among older children holds, the hypothesized relationship of resources with the development of behavioral difficulties in younger children only partly holds, with material deprivation associated with externalized behavioral difficulties for boys, and no observed effect for income on either internalized or externalized behavioral difficulties for either boys or girls.
Sensitivity Analyses
The results from the main analysis provided only limited support for the expectation that the three economic well-being variables would be related to greater behavioral difficulties in early-to-middle childhood and adolescence. In order to confirm that the findings were robust to different operationalizations of the economic well-being variables and to check whether results remained stable when operationalizing variables in a similar manner to Schenck-Fontaine and Panico (
2019), three additional models were run as robustness checks. In the first robustness check, material deprivation was measured as a continuous inversely weighted variable, so that infrequent experiences of material deprivation were given a greater weight than more common experiences of material deprivation. The second robustness check operationalized all economic well-being variables as dummies. The low income dummy distinguished between those who are 60% below the sample median and those who are above it, material deprivation was operationalized as in the main analyses and the financial strain dummy combined those who said they had difficulty or great difficulty making ends meet as opposed to all other categories. In the third robustness check, all economic well-being variables were again operationalized as dummies but based on different dichotomizations. This time, the financial strain dummy variable was operationalized such that any experience of strain was classified as experiencing financial strain, the alternative deprivation dummy was coded as in the main analyses, and the income dummy was based on the lowest income quintile.
Results remained largely stable to the different operationalizations of income, material deprivation and financial strain. Further details on the robustness checks employed and the results are available on request from the corresponding author.
Discussion
A large body of literature has investigated the relationship between child behavior and economic well-being. However, the predominance of cross-sectional approaches and a focus on financial indicators has made it difficult to identify association over time and overlooked the potential heterogeneous effects of different aspects of economic well-being. This study sought to address some of these shortcomings by investigating the links between different experiences of family economic well-being and the development of internalized and externalized behavioral difficulties in two cohorts of young people between the ages of nine and 17 (GUI98) and three and nine years old (GUI08). The analysis covered a period of major economic fluctuation in Ireland: from just before the outbreak of the global financial and economic crisis (i.e. 2007/08) to the depths of the Irish recession (i.e. 2011-13) and, finally, economic recovery (i.e. year 2015 and after). This allowed us to disentangle the effect of economic well-being from time invariant confounders by measuring how changes in economic well-being within families (as opposed to between families), were associated with changes in children’s internalized and externalized behavioral difficulties. Furthermore, to develop a more comprehensive understanding of the heterogeneous effects of economic well-being, household economic well-being is conceptualized as multidimensional. This was done by using the available measures of disposable household income, material deprivation and financial strain to tap into the financial resources, living conditions and psychological financial stress aspects of economic well-being.
The effects of the recession were clearly visible in descriptive statistics of the key variables, with significant deterioration in economic well-being at the height of the recession. Moreover, in all waves, lower economic well-being in the three measures was associated with greater behavioral difficulties. However, fixed effects estimation provided only limited evidence for a relationship between income and child behavior in either of the cohorts. Meanwhile, an increase in household financial strain was related to greater externalized behavioral difficulties in boys and girls at later ages (GUI98) but had no association with either internalized or externalized behavioral difficulties in boys and girls at younger ages (GUI08), or internalized behavior at older ages (GUI98). Additionally, household material deprivation was linked to greater externalized behavioral difficulties in boys at later ages (GUI98) and internalized behavioral difficulties in boys at earlier ages (GUI08), but not found to be associated with internalized or externalized behavioral difficulties in girls in either age cohort. Interestingly, the results for income, material deprivation and financial strain remained even when adjusting for the other two economic well-being variables. Moreover, these findings were largely robust to different operationalizations of the income, material deprivation and strain variables. This suggests that each measure of economic well-being holds an independent effect that is not captured by either of the other two measures. This finding adds to the body of evidence on the importance of conceptualizing and measuring household economic well-being as multidimensional (Guio,
2018; Whelan et al.,
2001), and is of policy interest, given it indicates that different experiences of economic well-being determine distinct child behavioral outcomes. While household income is a key measure of financial resources, it does not necessarily capture the material living conditions of families or the psychological aspects of coping with a given set of resources and needs. Hence, while household income remains an important and relevant indicator, it is important to consider it in combination with other aspects of household economic well-being when looking at child development. This has implications for theoretical models that incorporate household income, which may need to be adjusted to account for more multidimensional measures of economic well-being.
Placing the findings in the wider framework of economic well-being literature, the effects of material deprivation and financial strain corroborate the findings of existing studies. In line with FSM studies (Chzhen et al.,
2021; Ponnet,
2014), a significant association was found between financial strain and externalized behavior outcomes among adolescent boys, net of other household characteristics. Besides, the finding that adolescent boys seem to be more likely to externalize difficulties is of particular relevance since a recent study among a large and representative sample of Swedish youth (Plenty et al.,
2021) indicates that externalized behavioral difficulties for boys are associated with later-life likelihood of being Not in Employment, Education or Training (NEET). Similarly, the finding that material deprivation had an effect on younger boys’ internalized behavioral difficulties is of policy interest, given findings from previous research highlights the detrimental effect of economic hardship experienced in early ages on both short and long-term behavioral, cognitive and attainment outcomes (Duncan et al.,
2010; Kiernan & Mensah,
2009).
However, the null finding regarding the effect of disposable household income on child behavior in either age cohort is somewhat surprising. While it is consistent with the finding of no significant income poverty association in fixed effects regression estimates with child behavior problems observed in recent research (Schenck-Fontaine & Panico,
2019), it contrasts with the body of observational evidence on the association between household income and child outcomes from the FSM literature. A potential explanation could be that the dependent variables do not capture the effect of changes in income on child outcomes, with prior research indicating that income poverty has a low association with emotional and behavioral development, but a stronger association with other child outcomes including cognition, educational attainment and self-rated health (Plenty & Mood,
2016). Furthermore, it could be that other factors, such as household debt, moderate the effect of financial resources, with income alone not able to accurately capture this effect (Conger et al.,
1994). An alternative explanation is that income is associated with behavioral difficulties but is not the cause of them. This would explain why some studies (e.g. Schenck-Fontaine & Panico,
2019) observed income effects in models that rely on variation between households, but failed to observe the same effects when drawing on variation within households. Findings are also in line with the results of a recent systematic review of causal evidence of income effects on child outcomes (Cooper & Stewart,
2021), which documented limited evidence of direct household income effects on children’s behavior as opposed to via mediators such as parental mental health. This would suggest that there are other unobserved confounding factors that are captured by the income variable in between-household analyses, which also has implications for policy. Indeed, tax-benefit mechanisms can have a more direct effect on household incomes than on the actual living conditions or subjective evaluations of making ends meet.
Finally, this study holds several limitations, pointing at promising and relevant avenues for further research. First, the findings presented in this article might be specific to the Irish context and the Great Recession since Ireland experienced a unique and extreme change in economic climate. The possibility that these findings do not translate to different time periods, cohorts or countries cannot be dismissed, nor can it be categorically claimed that they hold in periods of milder economic decline. However, in the light of COVID-19 and the current cost-of-living crisis, these findings bear relevance for future studies. Second, by adopting fixed effects estimators, the models only look at change within families, and do not provide insight into whether the effect of the measures of economic well-being on child behavior differ by socio-economic status. It could be, for instance, that higher socio-economic households that experience income declines and increased financial strain are able to shield their children from the effects of such economic hardship owing to greater resources available to them prior to the onset of economic hardship, while change in income and financial strain among lower socio-economic households will more directly translate into effects on child behavior. The interaction between socio-economic background and economic well-being on child behavioral outcomes could be an interesting avenue for future research. However, given the primary interest of this article was to identify whether different experiences of economic well-being had independent effects on child behavioral outcomes, it was decided that it was not within the remit of this article to explore. Third, while fixed effects estimators, by virtue of the use of the individual as his/her own control, do not require a baseline value, the reliance on only within-individual variation over time discards all between-individual variation that could be used to estimate effects, leading to larger confidence intervals and less precise estimation when within-variance over time is small. It is therefore possible, particularly in the instance of Cohort08 where the sample timeline started later in the economic crisis, that effects which were present were not identified in the fixed effects models. Lastly, this study, while identifying different associations between measures of economic well-being and child behavioral outcomes, does not delve into the mechanisms through which these associations manifest themselves. Such an exploration is beyond the scope of this analysis – however, understanding how these different experiences of economic hardship lead to the development of varying child behavioral outcomes could prove an exciting avenue for future research.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.