Teaching Personal Finance DOES Work

It was a devastating day when I learned that teaching personal finance, or personal finance education if you will, does not work. As in, you can tell people about compound interest, how to budget, how to balance the books, and bad forms of debt, but this doesn’t actually do anything towards having them make better financial decisions. A sad day indeed.


In behavioural science it’s known that endlessly throwing information at people doesn’t work. People either get overwhelmed if they do engage with it, or just don’t bother with it at all as it’s too much to begin with. So in that light, if personal finance education simply “throws” financial information at people, it’s not super surprising it doesn’t work. However, it has always seemed rather counterintuitive to me that if you were taught about compound interest, you wouldn’t be able to see the benefits of saving and investing, and see the detriments of high interest loans, such as paydays and credit cards (or the modern version: buy now pay later apps). I should maybe backtrack a bit. The finding that teaching personal finance doesn’t work came from a meta-analysis conducted by Fernandes et al. (2014). They analysed the relationship of financial literacy and of financial education to financial behaviours in 168 papers covering 201 prior studies. So you could argue that that is quite a lot. Issue is, they find that interventions to improve financial literacy explain only 0.1% of the variance in financial behaviours studied, with weaker effects in low-income samples (you can likely blame structural factors for this). A specific point made by Fernandes et al. (2014) is that financial education, like other education, decays over time. This means that education that was given when someone was 10 years old won’t have much of an effect when they’re 20. If you want to change impact at the age of 20, you’re better off teaching someone at 19.75 years of age. If you can. They find that even large interventions with many hours of instruction have negligible effects on behaviour 20 months or more from the time of intervention. All results so far are a kick in the teeth for personal finance education and the behavioural sciences that have integrated this approach into their policy making. But wait, there is more! Although the initial sample of 168 papers and 201 studies seems very impressive, the majority of these studies was in fact correlational. This means that it is impossible to claim a causal relationship for financial education on financial decision-making and that other things may be going on. The authors, however, did dive deeper into this. The correlational studies that measured financial literacy find stronger associations with financial behaviours. After having conducted three empirical studies and they find that the partial effects of financial literacy diminish dramatically when also controlling for psychological traits that have been omitted in prior research or when one uses an instrument for financial literacy to control for omitted variables. The authors use this finding to argue that financial education (as studied to date) has serious limitations that have been masked by the apparently larger effects in correlational studies. As a result they do not support financial education as a generality, but make the argument for “just in time” types of financial education tied to specific behaviours it intends to help. To continue with an example I’ve already given before: you’d be taught about compound interest when applying for a credit card or payday loan. Not entirely sure when someone would teach you about the benefits of compound interest for saving or investing as you’d have to be interested in these types of financial services before receiving the education, but I understand the point Fernandes et al. (2014) are trying to make. I just didn’t want to believe it. Because I am a massive proponent of personal finance education (I suffer confirmation bias, it is what it is).


Now since this devastating paper there have been many more experiments, randomized control trials (RCTs), surveys and other approaches to the study of the (possible) effect of personal finance education on financial decision-making. And I wouldn’t be surprised if this was partially due to the fact that no one was keen on accepting the initial outcome. I won’t go into the results of the many individual studies that have shown there to be an effect of personal finance education on financial decision-making, because Kaiser et al (2020) have done this for me. They do find an effect, quite a large one. Kaiser et al analyse the results of 76 RCTs with a total sample size of more than 160,000 individuals and find that personal finance education has positive causal treatment effects on financial knowledge and financial behaviours. Looking at knowledge and behaviour separately, the authors find that these treatment effects on financial knowledge are similar in magnitude to the average effect sizes realized by other educational interventions (math, reading). The effect sizes of financial education on financial behaviours are comparable to those realized in behavior-change interventions in the health domain or those aimed at fostering energy conserving behaviour. I have linked both the papers in this article through hyperlinks, so you can read them for yourself if you’re not convinced or want more details. Whether you decide to do this is obviously up to you. But even after reading both papers’ result sections you might still wonder: why are these outcomes so drastically different?


Everything in the social sciences depends on context. There’s rarely an exception and the analysis of the effectiveness of personal finance education is not one of them. The samples in both papers are drastically different. As I mentioned before, a lot of the studies analysed by Fernandes et al. (2014) were correlational. They only looked at 13 RCTs. And there’s good reason for this: there probably weren’t any more, or at least not any more of reputable origin. Six years later, the field of personal finance (education) has exploded: its growth has been exponential. Kaiser et al (2020) had many more studies to choose from and limited their sample to “only those RCTs that have been published in top economics journals, when restricting the sample to only those studies with adequate power to identify small treatment effects, and when employing an econometric method to account for the possibility of publication selection bias favoring the publication of statistically significant results.” It’s not even comparing apples to oranges, it’s comparing Blockbuster to Netflix. Additionally, Kaiser et al (2020) do not find differences in treatment effects for low-income individuals and the general population. They also do not find strong evidence to support a rapid decay in the realized treatment effects, although they also do not find support for the sustainability of long-run effects either. The interpretation for this can go a couple of ways, but the way I see it: the decay isn’t rapid, but stuff you learned at age 10 for one class might not exactly stick around until 10 years later. Like most other topics of education. So there is most definitely an argument to be made for repetition, longer periods of (financial) education, as well as timely education.


So what are you supposed to do with these wildly different results? I do believe the main driver is the difference in sample. Comparing correlational studies to RCTs was unlikely to yield the same result (hindsight bias much?). Had it yielded the same result (there being no effect of personal finance education on financial knowledge and behaviour) that would’ve been a proper final nail in the coffin of personal finance education. However, that wasn’t found. So there’s something to see for the method in which we research effectiveness of interventions, or in this case, education. I also believe that the personal finance education given 10 years ago (the ones analysed in Fernandes et al. (2014)) is not the same as the financial education given now, which makes active use of behavioural insights and novel scientific findings, in addition to having to address many more novel and complex financial products and services in ways which are approachable to the average consumer. There has been an increase in initiatives providing this education and training because the financial landscape has become so complex. The one thing that we do have to keep in mind is that finding an effect of personal finance education in an RCT, does not guarantee this effect size when applying the intervention (education) to the “real world.” This is what Fernandes et al. (2014) meant by finding reduced effects when controlling for specific variables such as income (low) or personality traits (conscientiousness). Some RCTs control for this, and others do not. But in the real world, when being faced with people who need (or want, or are simply receiving) financial education, you don’t get to pick what personality trait and income group you work with. And you better hope that what you’re teaching helps each and every kind of individual, and doesn’t become detrimental to some of them… Another concern raised by the authors of both meta-analyses is that even to this day, there’s very few studies analyzing the long-term effects (and I mean LONG term effects) of these types of interventions. How far can the effects of one-off or continuous personal finance education be pushed? Which form is most effective? What duration is most effective? Are there ways to extend its shelf-life? There are no conclusive answers to this yet. This is not a limitation of the meta-analyses nor really of the field of study. Conducting these types of longitudinal studies comes with a high level of complexity and general struggle, relating to needing large amounts of funding and massive sample sizes (to deal with dropout), as well as much higher than average levels of patience and determination to continuously keep track of your participants over the year(s). Moreover, all this high level effort is not at all a guarantee for finding anything of note (easy to publish). Can you imagine doing years of work and not being able to publish properly? It’s not for the faint-hearted. Anyway, we’re getting off track here.


Although the points raised by Kaiser et al (2020) on why the analysis by Fernandes et al. (2014) was lacking are valid, we might want to, once more, be cautious in how far we want to extend this finding. However, finding a large effect size of financial education on financial decision-making to the extent that it is “of economic significance” through the meta-analysis of 76 RCTs with over 160,000 participants is great. Personal finance education: 1 – Haters: 0

Behavioural Science