BMO released a report on Friday about choosing between fixed and variable-rate mortgages.

BMO says, “Over the past 30 years it has been more cost-effective for borrowers to have a variable-rate mortgage 82% of the time.”

That appears true according to BMO’s assumptions. We did a slightly different test though, and will talk about that in a moment.

There is a problem with these types of studies, however, and that is sample size. There have been very few cases where history resembles today. In fact, we’ve never been witness to a monetary policy rate near 0%.

Nonetheless, BMO economist, Doug Porter, had this to say in support of variable rates:

- “The current outlook for inflation remains benign, which will likely keep price pressures at bay well into 2011.”
- The difference between 5-year fixed mortgage rates and variable rates “is now close to an all-time high.”
- “There is also some risk to locking in as fixed rates could fall if the economy performs worse than anticipated.”

On the other hand, BMO’s report made several points upholding fixed rates. It said:

- “Short-term rates are at extreme lows and pressure is likely to build for higher rates in the year ahead.”
- “The Bank of Canada’s overnight rate is now as low as it can go, so there is no further downside for variable rates.”
- “Although inflation hasn’t been a problem since 1991, there is a risk of an inflation flare-up as global central banks keep the pedal to the policy metal, and amid record government deficits. The Bank of Canada could be forced to raise interest rates aggressively, driving variable mortgage rates higher, but leaving Canadians with fixed rates unscathed.”
- Fixed rates were beneficial twice in recent history… “through the late 1970s and briefly in the late 1980s.” Both cases were “ahead of a period of rising interest rates, as is the case now.”

That last point is where you can look at things in two ways. BMO arrived at its conclusions by comparing posted rates to prime rate (their chosen proxies for fixed and variable mortgage rates). Prime is a good approximation for variable rates but *no one* pays posted fixed rates anymore.

So we did the same study using discounted fixed rates (i.e., 1.5% off of posted, which seems reasonable). The results were very different…click on the chart below.

With these assumptions, there were at least six periods in recent history when fixed rates beat variable rates at prime. Put another way, discounted fixed rates would have outperformed prime rate almost half the time. (Mind you, if big discounts to prime were available once again, variable rates would fare better.)

BMO declares that the optimal choice “depends on the individual.” That, of course, is true as always.

Interestingly, if we forget about chart data for a moment, it appears BMO presents more arguments in support of fixed rates than it does for variable rates. In addition to its comments above, BMO says:

- “The moderate extra cost of peace of mind you can get from a fixed rate may be a price worth paying.”
- “There is also a reasonable scenario where fixed rates may actually prove to be a cheaper alternative at this point.”

Despite all of this, BMO says its “core view” is that:

“The most likely economic and interest rate outlook will ultimately again slightly favour the variable rate option.”

But what does “slightly favour” mean? If we assume a 55% probability, then it’s little better than a coin flip.

For most people, if they faced a 45% probability of losing money, they’d insure against it. Fixed rates provide such insurance, and if rates rise 2.5% or more over the next few years, that insurance will start looking pretty good.

Our perspective isn’t meant to be a blanket endorsement of fixed rates, but the above is definitely something to think about if you’re on the fence.

I’m not sure how you get to your conclusion that:

“there were at least six periods in recent history when fixed rates beat variable rates at prime”Or that“discounted fixed rates would have outperformed prime rate almost half the time.”It looks to me like you are ignoring

how longthe FRM was lower than the VRM. If you mean there were 6 instances, however brief, where fixed rates were lower than variable, then, yes, you are right. But that’s not the correct way to compare VRM and FRM, because people don’t get mortgages for brief periods. See that 1995 blip on your graph. It is meaningless when comparing VRM and FRM.The fact is, for

at least2 of those 6 instances, the person still would have been better off with a VRM because the VRM would have been lower for the bulk of the mortgage.Second, you don’t seem to consider

how muchlower or higher the VRM and FRM are relative to each other. If the FRM is lower than the VRM by 0.1 points for half of the term, but the VRM is lower than the FRM by 3 points for the other half, the winner is VRM.Finally, as you well know, Canadian monetary policy has change enormously since the 1970’s (inflation targeting, etc.). It is apples and oranges to put them all together.

In essence, all you have done is produce Moshe Milevsky’s graph, but without his statistical analysis.

Here is his original paper.

http://www.ifid.ca/pdf_workingpapers/WP2001A.pdf

Here is his latest research.

http://canadahometrust.com/Fixed%20or%20Float.pdf

Which is not to say that the BMO study doesn’t contain some iffy assumptions & conclusions.

You are absolutely right — it does.

Hi Bob,

Thanks for the questions. Per your points:

BMO stated there were two times when fixed rates were more beneficial. BMO did not qualify those instances by the amount of time that fixed rates were better. We’re merely saying that, based on a simple comparison using discounted fixed rates, there were six periods.

The length of the window varies in each instance. You chose a period for your example (1993-94), which just happens to be the shortest (i.e. there was only a five month window where you could have got a fixed rate that would be better than the average variable rate over the next five years). As you can see on the graph, the other five “windows of opportunity” lasted significantly longer.

Regarding your point about time, the time that fixed rates were lower is built into the graph to a large extent. As you can see, the proxy for variable rates is a five-year moving average of prime rate in the future. Going back to your example, that means that in February 1994 you would have theoretically got a 5-year fixed at 5.7%. The average variable rate would have been 6.65% at that point, for the next five years. Obviously, that doesn’t take compounding into account (which is important when fixed and variable rates are close). However, the main point was simply that there are more instances where discounted fixed rates performed better than was mentioned in BMO’s report.

Monetary policy was not a key consideration in BMO’s study so we haven’t got into that either. If you want to look at monetary policy today, then one might ask his/herself how rising interest rates would affect a variable-rate mortgage versus a fixed rate. In our environment, with a 0.25% floor on the overnight rate, one could argue that fixed rates have a higher probability of outperforming (not our prediction, just something to consider).

Regarding Moshe Milevsky, we always give deference to him and in no way should our chart be on put on a same analytical plane as his research. That said, like BMO, Moshe’s charts in the link you provided was based on posted versus prime. Again, I’d submit that a discount off of posted rates makes for a fairer comparison because it represents what people would actually pay today.

In sum, this story is meant to provide both a high-level overview of rates over 40 years, and another look at the BMO research that’s been making the papers lately. The future is random, compounding matters, and the sample sizes are insufficient (in BMO’s study and ours), but it’s an interesting overview nonetheless.

Cheers,

Rob

Ah yes, my apologies.

I looked at that chart before my morning coffee and didn’t notice that the VR line was a future moving average rather than simply Prime at the given time.

A discounted 5-year rate is definitely a better comparison than posted.

Rob

I have thought the same thing many times when reading Prof. Milevsky’s papers. I know discounts were not always available but can’t see the logic of using posted rates to represent fixed rates today. This is the first thing that struck me about the BMO report too.

Take care

Marc

Does the graph also show when VRM’s were prime minus a discount as well?

Hi TSP,

Not this graph, because most VRMs are still around prime. As noted, however, big discounts to prime would increase the odds for variables.

Cheers,

Rob

“Prime is a good approximation for variable rates but no one pays posted fixed rates anymore.”It is that last “anymore” wherein lies the problem. Prime is a decent proxy for

currentvariable rates, but it hasn’t always been the case. Prior to Summer 2008, discounts of 0.6-0.9 were standard for VRM. Post-2008 it has fluctuated up to Prime +0.8.Likewise, while posted rates are currently a poor approximation of fixed rates, they haven’t always been. In the 1970’s and much of the 1980’s, for example, there were far fewer discounts offered from posted rates. So is Prime vs Posted -1.5% a reasonable average to use across the last 20 years? I’m not entirely convinced.

It is pretty hard to model historical trends without access to accurate historical data.

Aren’t there historical data somewhere that could tell us what was the

actualtypical VRM or FRM offered per year?Hi Ted,

Thanks for the note. We had a similar thought process initially. In the end, it seemed more extrapolative to applying a constant discount from posted rates and attempt to normalize all the data.

You’re absolutely right that people never used to get discounts on fixed mortgages. However, with a few notable exceptions, it’s not because lenders couldn’t afford to give them.

It’s also worthwhile to consider some of the factors that drove rates in the 70’s and 80’s, which include, but are not limited to:

* Less lender competition

* Fewer securitization options

* Inflation that was astronomical by historical standards

If these impediments didn’t exist, people would have received much better discounts from posted rates. None of these things should be factors going forward. Therefore, it’s important to somehow reflect that in the past data if we’re going to rely on past data to foretell the future.

As for variable rates, prime is a pretty good proxy. People paid prime rate on variables for many years. Discounts and (more recently) premiums eventually became common, but prime rate is a good approximation that takes into account these peaks and valleys.

The lending environment has changed significantly over the course of 40 years, so it’s difficult to find one perfect approach to back-testing. What we can say for sure is that people don’t pay prime – 0.60, prime +.80%, or posted rates nowadays. Therefore, consistently applying today’s fixed-rate discounts to the past doesn’t strike us as being out of the ballpark.

Unfortunately I’m not aware of any database that has historical discounted rates for fixed and variable-rate mortgages. I wish one existed!

Cheers,

Rob

Hi Rob, you said:

Unfortunately I’m not aware of any database that has historical discounted rates for fixed and variable-rate mortgages. I wish one existed!

Not that it is public data, but wouldn’t banks have that specific historical data?

Hi O.H., yes, it’s possible.

Its funny how in every report its stated that rates cannot go any lower than they are currently, that we’re at historic lows, yada yada. Well, guess what. They can. And they have been for almost 30 years. Sure, rates may rise a little in the next 12 to 24 months but theres nothing to say they wont then turn and come right back down again. Economies cannot support high rates anymore or even average rates. Its a sctructural problem in the financial industry but one thats not likely to suddenly get fixed in the next few years. Variable rates have been the better choice on ‘most’ occasions for almost 30 years now. I dont see that changing all of a sudden just because some mortgage brokers are screaming to get people into fixed rates now.

Salut,

Al you say, “Sure, rates may rise a little in the next 12 to 24 months but theres nothing to say they wont then turn and come right back down again.”

There is nothing to say they won’t keep going up!

You must understand how low 0.25% really is. These are emergency measures. Structural changes will not keep the Bank of Canada at this level for long.

Hi Al,

Thanks for the note. It was going so well until the last sentence. :)

There has undoubtedly been a structural shift, but I think you’d agree that specific predictions are futile. Better to focus on risk management based on what reasonably

couldhappen.In other words, assuming rates went up ___%, and assuming we don’t know how long they will stay there, what effect could this have on the client’s finances?

For the record, this isn’t really a broker driven issue because most brokers generate their opinions of the market from economist and analyst research. Every economist in Canada (that we’ve heard of anyway) has publicly stated that rates will rise. It’s a matter of when, how long, and how high.

In the end, it’s important to manage risk on an individual basis and not apply blanket recommendations (or rate forecasts) to all clients. That said, different opinions “make the market” so we appreciate your perspective.

Cheers,

Rob

Call me a conspiracist but don’t banks have more to gain by putting people in variable mortgages? If rates rise won’t they make more money than if people today get fixed rates?

I just wonder if BMO is pushing variable rates for this reason. It seems so counterintuitive to get a variable rate right now.

J

Jon, I believe the way banks operate, they win either way (fixed or variable). Correct me if I’m wrong, but if the cost to borrow increases (variable), it does for both the bank and the consumer (latter based on prime). For those who know the specifics, is it a kind of floating spread?

@ Jon – That would only be true if banks could somehow accurately predict future interest rates.

In which case they could do away with lending altogether and make a killing on the bond market.

Generally, when you shift risk to a lender by fixing your mortgage, you pay for it. So, as O.H. noted, it’s not like banks are upset when people take out fixed rate mortgages.

Al R

Rob,

This is a very interesting article that I only ready today (late by 9 days).

Regarding the graph, is it possible to extrapolate (dotted line instead) the Prime for the final 5 years moving averages using prime rates for the available reamining period instead of 5 year.

Also, if there are any more views, since the last comment on 28Oct, it would be appreciated.

Thanks.

Binu

Forgot to add…

Discounts on prime is happening again. I got my mortgage application ammended P-0.10% instead of =P that I had 2 weeks or so back.

When will my mortgage holder send out a renewal notice. I had a 5 yr fixed at 5.5 that is up for renewal Feb 1 2010. I see rates advertised as low as 3.98 for the same terms, if I go through a Mortgage broker, how likely am I to get the best advertised rates. Do I start shopping now 3 months early & can I lock in these rates, or wait to see what happens in the new year? I believe my credit rating to be near excellent.

It depends on the lender. A lot of lenders send out renewal letters 60 days before maturity.

You are always best to shop early (120 days before renewal) in case rates go up.

Call a broker and bank and go with the best overall deal.

Hi Binu:

Per your question:

“Regarding the graph, is it possible to extrapolate (dotted line instead) the Prime for the final 5 years moving averages…”

Maybe I misunderstood what you’re asking. The moving average indicates the average rate 5 years hence. Since we don’t know what future rates will be, we can’t extend the line beyond where it is now.

Does that make sense?

Cheers,

Rob

Rob,

The Prime graph (yellow line) stops at 2005. My suggestion was to extrapolate it based on rates that are available to date (and not the 5-year moving average for just this period)

Thanks

Binu