How do mental health issues affect the likelihood of food security? This question is difficult to answer empirically, for (at least) two main reasons:
- Endogeneity/unobserved factors. For example, individual, family, and neighborhood characteristics (e.g., family stability, access to health care, exposure to violence) may influence mental health and the likelihood of food insecurity. Furthermore, the direction of causation is unclear, as mental health problems may lead to reduced employment likelihood, which in turn leads to food insecurity; conversely, food insecurity can increase stress and increase the likelihood of mental illness.
- Measurement error. Many studies of mental illness rely on surveys and self-report measures of mental illness. This can lead to significant measurement error, particularly because stigma can lead to misreporting of mental health conditions.
How do we solve these twin problems simultaneously?This is what a paper is about Jensen et al. (2023) Try to solve (see also demo here). A clear solution is to use instrumental variables, but finding valid instruments is problematic because most factors associated with mental illness are also directly related to food insecurity. Furthermore, measurement error is more problematic when the key exposure variable (in this case, the presence of a mental illness) is binary.
The solution used by the author is adopted in Creed and Hill (2009) and Kreider et al. (2012). They applied these methods to data from the National Health Interview Survey (NHIS). They focused on patients who self-reported “nonspecific psychological distress (NPD),” according to the Centers for Disease Control and Prevention. Kessler (K-6) Scale.
In a standard OLS regression model (see below), endogeneity may exist because the “treatment” (mental illness) may be correlated with the error term. Furthermore, the measurement of mental health status (D) is subject to uncertainty.For example, let D* It equals 1 if the individual is indeed suffering from mental distress and 0 otherwise.However, the researchers only observed Dwhich is self-reported distress.
A key way researchers have solved this problem is by using part identification methods. The goal is to estimate the following average treatment effect (ATE):
In this equation, Y(D* = 1) Represents potential food security outcomes for adults in distress; Y(D* = 0) Represents food security outcomes when adults are not in distress.
I guess there is something wrong with this equation. To understand why, let's break down these values.let us assume P(Y=1|D*=1)=P(Y(1)=1|D*=1)*P(D*=1). If the true odds of mental distress—P(D*=1)——If known, this quantity can be estimated.However, the term phosphorus[Y(1)=1|D*=0]because it estimates counterfactuals that are unobserved in the data (i.e., the food security levels of individuals without mental illness if they did have mental illness).
The second problem is that we don't really observe D*, so the first term is inestimable.The authors break the term down into evaluable components [i.e., P(Y=1,D=1)] and measurement error term. Because sigma is often present in mental illness, it is likely that mental illness is underdiagnosed rather than overdiagnosed.The authors claim that once it can be assumed that there are no false positives, then θ1+=θ0+=0. The authors also assumed that the ratio of true to observed nonspecific psychiatric disorders was the same as that reported for all types of psychiatric disorders. To do this, they use data reported by SAMHSA.
It also imposes 3 different types of assumptions:
- Monotonic Processing Selection (MTS). This means that people who actually suffer from mental illness are (marginally) less likely to be food insecure than people who don't actually have mental illness.
- Monotone Instrumental Variable (MIV). Here, they hypothesize that people living in areas with fewer food stores are (weakly) less likely to be food secure.
- Monotonic treatment response (MTR). On average, psychological distress does not improve food security.
Using these methods, the authors found:
Applying relatively weak monotonicity assumptions to the underlying food security outcomes, we find that mitigating SMI would increase food security rates by at least 9.5 percentage points, or 15%.
You can read the full article here There is also a useful summary slide (I borrowed it extensively) here.
