Video: RRSP and TFSA: using financial education to improve decision-making


Pierre-Carl Michaud: While we load up the slides, how many of you have a registered savings plan, an RSP? How many have a TFSA, a tax free savings account? So a lot of you, hey? How many hold both? A lot of you, okay, so that's what we're going to be talking about today, that choice that we make when we're saving between these different types of plan.

Obviously in the general population, a lot of people also make contributions to these plans. It's important to know that 100 billion every year is contributed to RSPs and TFSAs. Nowadays, more in TFSAs than in RSPs despite the fact that TFSAs have been introduced more lately. So this is an important choice that we make as we save, in particular for retirement. That's what we'll be focused on today.

As you know, a basic understanding of marginal tax rate is key to understanding how you should place your money between RSPs and TFSAs. They are other considerations, but one of the key one you've learned, of course, is that if you're going to pay a lot of taxes today, you should probably use an RSP and save up on taxes once you take out the money in retirement and face lower marginal tax rates. If the opposite is true, of course, you probably would consider a TFSA from tax purposes.

In many circumstances, we think of marginal tax rates and retirement as being low relative to when we work. We'll show in a second that that's not true, and it's much more complicated than we think.

So here, a simple case, Quebec male working, age 40, and we use a calculator that we developed to compute the effective marginal rate, taking into account all sorts of programs that we have in Canada, tax credit and whatnot. And what you find basically is we try to line up the marginal tax rate (inaudible) effective comparison you would make by levels of earnings, and what you can see is that it's not always the case, that you face lower marginal tax rates in retirement compared to when you're working. So it's a complex choice, and a good understanding of those things are important.

Now mistakes can be very costly. We worked through a series of example in the paper, and one of them shows that for a thousand dollar contribution at age 30, if you're gonna be one of the guys that faces high effective marginal tax rates in retirement, lower end of the distribution, in terms of retirement income you're gonna get in retirement, you're gonna get about the third of what you would get with a TFSA if you choose the RSP wrongly. That's a lot of money. We think about giving incentives to save to people, but choosing how we save money is also a very important consideration, and tax consideration should be at the core of this.

So in this project, we're interested in understanding how Canadians make those choices, and so we designed an experiment to learn how individuals make choices between TFSAs and RSPs. So we parted with asking Canadians, Delvenia is a Toronto-based company here, and we filled it online with 3,000 Canadians in experiment where we're going to try to learn about this. We embed within the experiment, where people are going to be making choices, a randomized financial education intervention. So this is not a real intervention where we look at people's behaviour; we're going to look a little bit at how they make choices in an experiment, but we're going to do as if we're in the field, we're going to be randomizing an intervention, which I'll talk about in that respondents we'll see.

We'll try to teach the definition of RSPs and TFSAs. It's gonna be a very short video that we actually borrowed from Desjardins who gave us the permission to use the video, and teaches us just a little a bit about marginal tax rates. Simple, simple stuff. Then we're going to give them a second video. Some of these folks are going to get a second intervention. They're going to learn a little bit about clawback and some of the retirement income system provisions, and we're not going to lead them initially to think that these things, you know, would favour TFSAs relative to RSPs, but we'll just try to be neutral and say, guys, it might be that if you're planning on receiving some of these benefits, it would be important to pay attention to marginal tax rates because these could have a real consequence.

So we look at effects on the knowledge. So the first thing we measure after we do the intervention is measuring the knowledge that people have of RSPs and TFSAs. And then we look at the quality of choices, and that's an important point to bring home that very often, and we did allude to the point, saving more is not always a good idea. It's not always an optimal choice. So having a good benchmark where in the experiment we can say, well if we increase that outcome that necessarily means that we've increased well-being is very important or the quality of choices.

So what's the context? Individuals receive an unexpected refundable tax credit, they're forced to invest in RSP or TFSA and they have enough contribution room, they have to withdraw around the age of 70, and we specify all of the payoffs, so there's no uncertainty in terms of what they might get. And we give them visual aids, and here I did not reproduce the screenshot, but we give them visual aids for marginal tax rates on the side. We don’t tell them where they're marginal tax rates is, but we tell them a little bit what the schedule look like. And then we ask them which one do you choose.

We repeat six times by respondent and we randomize both the amount and the return so that we can look both at cross respondent but also within respondent, does the quality of choices change. So have a control with no intervention. We have a first treatment where we only show that Desjardins video which tells them basically what these products are, and then the second treatment which gives them more information on means testing.

Now, what's the effect of the intervention on simply knowledge: the first thing I want to point to is that basically knowledge of whether contributions are deductible in an RSP is relatively low in the population; 65% -- 64% of respondents actually understand that. Our withdrawal tax when you use an RSP; 67% of respondents get that correctly. Our returns tax; 45% of respondents understand that. Now luckily for us once we look at other things there, individuals don’t have the right answer, but we're not targeting these outcomes in terms of the intervention, so that's a nice placebo test that actually our interventions are not gonna have any effect on those behaviours. But if you look at the effect of these two interventions, there very short ones, one minute each, you increase by 14 percentage point the number of correct answers to whether contributions are deductible; 14% withdrawals tax and overall on the score, over 100 on these questions by seven percentage points. So knowledge does increase in the treatments relative to placebo, they're relative to control.

Now second choices. Without intervention, you can see this on the third line of the first panel of results, I was told this table was going to be a mess when I presented it, I didn’t have time to make nice crafts, so you will have to bear with me. But basically, traces are a toss-up. If you look in no intervention, in 50% of cases, people have the correct answer. So basically, if they had, you know, tossed up a quarter, they would have gotten about the same result.

Now when we do the video only intervention, the number of correct answers increases, but not so much. When we do both the video and the slide, it increases by a lot, and the slide was about the means testing. And so we see a six percentage point increase, that big or small, well given the intervention didn’t cost much, I would say that's a pretty nice result in terms of increasing the quality of choices.

Now as economists, of course, we can do much more. We measure preferences prior to the intervention, using an incentivized experiment to measure both intertemporal preferences and risk aversion, and then we can look at optimal choices, also given people's preferences and we get about the same result. So whether we use just marginal tax rates or more complex decision rule we get improved choices.

So what are the takeaways so far from this experiment, because we want to do much more with the data? Well, the first thing and we probably knew this but it's good to know that knowledge of TFSAs and RSPs is limited, and that's despite the fact that we contribute about 100 billion a year in these things.

Choices are close to random without an intervention; that should be preoccupying because we saw that these mistakes can be large when you don’t choose the right products, so you could be saving a lot for retirement, making a huge effort, sacrificing consumption today, but in fact you're getting a very lousy return in terms of how much you're saving.

A brief educational intervention was able to increase the quality of choices by 10%. And here I want to emphasise the measure. We were able to construct a measure of quality by basically constructing the experiment in such a way that we have a good counterfactual for what should be the optimal choice, and that's very important.

Randomization was key to the evaluation. Of course, a lot of folks in academia understand that if you're going to publish that kind of work, you need to randomize. A lot of people in the field know that as well, but very often we see a lot of evaluations where these things are not done. And that's very important for the credibility of assessing whether an intervention has effects.

And that might play a role into the debate whether financial education works or not. In some instances, when evaluations are done using methods that do now allow to really estimate the effect of those interventions, where we don’t have a clear outcome to measure, that measures quality, it might be the case that we get nothing but that's not necessarily a conclusion that leaves us to think that education doesn’t work.

So I'll stop here. Just acknowledge a few collaborators at HEC who have worked on this project. Thank you.

Page details

Date modified: