Wednesday, January 25, 2023

Targeting on Household Income Doesn't Make Sense and Has to Go

The Malaysian government's policies on social welfare has been for the past decade been anchored on household income categories. We've become used to talking about the B40, M40 and T20 of the household income distribution, and in popular discourse these categorisations have become synonymous with the poor, the middle class, and the rich.

But the more I've thought about it, the less satisfied I've become. The fundamental problem is household heterogeneity - households are not the same in terms of their size, their composition, or the number of income earners relative to dependents. Two other problems are that a small minority of households have non-family members while some dependents that don't live within the household (primarily elderly parents living separately). These differences cause significant variation in household expenditure, that are simply not captured by anchoring policy on income.

I can't publish any of the data that I have since I don't have any permission to do so, so the following will have to be in general terms.

Just to give an idea of the level of differences I'm talking about here, the prototypical nuclear family of two working adults and one or two children is less than a quarter of Malaysian households. If we include those households with more than two children, its only a little over half. Nearly a third of households have three or more income earners. Taking averages hides all these significant differences.

The key variables for the impact on household expenditure are the number of income earners, and the size of the household. The former is self-explanatory - the more income earners there are, generally the higher household income will be, even if there is also an increase in expenditure (working adults spend more than children or retired parents). The size of the household on the other hand has a direct impact on household expenditure - based on EPF's Belanjawanku, an additional child adds nearly RM1k to minimum household expenditure.

All this matters. Let's take a concrete example to show why (these are not hypotheticals, and exist in the data):

  1. A single person with no dependents earning RM4k would be in the B40 category (income deciles based on the 2019 household income survey);
  2. A two income earner household bringing home RM8k with one child would be in the M40 category;
  3. Three persons earning a combined RM12k with 4 children would be in the T20 category.
In terms of living standards, each of the above is roughly equivalent after taking into account minimum household expenditure. And yet our policies assume the household in the B40 group "deserves" more assistance than the household in the T20 group. If anything, the T20 household in this example is probably the more vulnerable, since the loss of one income earner would have a substantially bigger impact on household finances. Just to underscore this point, our RM4k B40 household above would be considered borderline T20 if we compare based on individual rather than household income.

All things equal, we would expect a T20 household to be in a better position financially than one in the M40 or B40 groups. But things are NOT equal. While a lot of households do conform to the stereotypical view of household income categories, nevertheless a T20 household is not necessarily rich and a B40 household is not necessarily poor. There's enough heterogeneity between households that anchoring policy on income alone results in significant inclusion and exclusion errors.

So what should be done? There are I believe two ways to adjust our approach:
  1. Use household income per capita, instead of household income. This has the virtue of taking directly into account the size of the household when defining the target group. For our examples above, this would be RM4k, RM2.7k, and RM1.7k respectively, which makes more sense from a policy priority perspective.
  2. Use a categorical approach instead of income ie target children and the elderly directly. This avoids the threshold problem in both household or per capita income targeting, but does require a substantial change in the philosophical approach to social assistance.
There are flaws and additional data and administrative overhead in both of these alternatives, but either would be a better alternative to household income alone.


  1. I agree with you. This should be amplified

  2. you really need to write more often! missed your articles

  3. Here is complex model we use to calculate decile per household, these considering nutrition, CPI and many more factor.

    But too complex to implement the bantuan to rakyat.