Interest Rates, Mortgage Refinancing and Consumption

This is a discussion in a comment thread which I think is worth pulling back. I claimed (as I often do) that interest rates do not have noticeable effects on consumption. Please notice the “noticeable” — it is an appeal to standard econometric analysis of aggregate time series.

OK the discussion

January 16, 2015 10:14 am


One other thought while readily admitting I am out of my element here.

A couple of weeks ago I did a private refi for a buddy of mine. He was at 7.5% on 300gs, paying PI of $2084 a month. Now he is at $1578 while keeping the term the same at 4%.

That is $500 a month of potential consumption caused by low interest rates. And knowing my buddy, he will spend it. 🙂

part of my reply

I will claim that consumption (as officially defined) is not noticeably correlated with interest rates. The key word is “noticeably”. I assume that EMichael’s friend’s consumption will increase. However, as far as I know, the effect of such events on aggregate consumption is so small that it doesn’t show up in the aggregate data.

The correlation between the achieved 3 month real interest rate and the consumption disposable income ratio is -0.13. OK I should look at the mortgage rate, but this is a blog comment.

Always I am discussing how to model the economy. I don’t claim that my story is a complete description of everything which happens in the economy. I am discussing which effects are small enough that they can be neglected when modelling.

January 16, 2015 6:07 pm

i would guess that while emichael’s friend’s consumption would increase
(unless paying interest is “consuming” a financial product), the guy collecting the interest would decrease his consumption…. unless he got paid off and so increases his consumption with the money he now has “not lent.” in that case maybe the new mortgage holder would have less consumption than the first…

update: the interesting comments continue. Marko knows a lot more on the topic than I do.

January 17, 2015 4:38 am

You might have better luck seeing an effect if you looked at durables – and , more specifically , auto sales. A common use of a sizable windfall via mortgage refi seems to be splurging on a new set of wheels. Another common use is paydown of high-interest debt , like credit cards , which doesn’t increase current consumption , but bodes well for consumption in the future ( assuming that lower debt load households will tend to consume more than higher debt load hhs ) .

Also , you have to consider the after-tax impact of lower mortgage payments – you probably only get about 70-75% bang for the buck after tax , which makes for a pretty good hit to “noticeability”.

A couple of recent papers :

– See more at:

End update

My long confused and confusing reply to Coberly after the jump (and down in the thread)

Robert Waldmann
January 16, 2015 10:48 pm
This comment thread is getting interesting enough to maybe pull back to the blog somehow.

@Coberly it is true that in standard mainstream macro new Keynesian DSGE models refinancing doesn’t affect consumption. Consumption is modelled as the choice of a large number of identical people (called the representative agent) who have average income, wealth and consumption. This means that the mortgage interest nets out.

However, such models are not realistic.

Sticking with the discussion of the meanings of words (not the interesting part of the thread) I note your odd use of the word “guy”. This is normally used to refer to a human being. But the mortgages in question belong first to a bank then are packaed and sold to … well in the end probably a pension fund or maybe a very rich guy.

In the standard models (at least those popular pre-2008) this doesn’t matter, but in the real world it matters a lot.

The old Keynesian way of discussing this issue is the concept of the marginal propensity to consume — the effect of of a change in someone’s income on their consumption. if debtors have a higher marginal propensity to consume than creditors, lower interest rates cause higher consumption (as the reduction in interest rates acts like a transfer from creditors to debtors).

Let’s pretend the bank didn’t package the mortgage into a mortgage bond and just kept ownership (so much for reality). Then the reduced flow of mortgage payments are going to a firm. Some will go to shareholders as dividends and some won’t. In some simple models this shouldn’t matter, but in reality it does. Undistributed profits aren’t personal income (they should cause personal capital gains eventually). The fact that consumption is more correlated with personal disposable income than with GDP is a hint that undistributed profits have less effect on consumption than personal income.

Also the dividends may go to a defined benefit pension fund and so not appear as anyone’s personal income. Finally, the dividends which go to individuals mostly go to wealthy individuals whose marginal propensity to consume is low.

Even old Keynesian models tended to ignore the distinction between stuff belonging to corporations and the personal belongings of the shareholders. Only when dealing with data was the distinction between GNP minus taxes and personal income minus taxes made (the difference is undistributed profits of corporations plus depreciation of capital which is ignored in GNP and GDP hence the G).

My this reply is getting long and confusing and probably confused. I’d say the effect of refinancing on consumption is ignored entirely in widely used models for exactly Coberly’s reasons. That more realistic models and evidence supporting them strongly suggests that there is an effect, and finally, the data on interest rates and aggregate consumption suggest that ignoring the effect is a reasonably approximation.

Finally a regression. Consinc is the ratio of consumption to disposable pesonal income. rmort is supposed to be the real interest rate charged on mortgages and is the standard 30 year conventional mortgage rate minus 4 times the past quarters consumer expenditure deflator inflation. Obviously the way I deal with inflation is absurd, but this is a blog post. Data on the mortgage interest rate is available only from 1971 on.

. reg consinc rmort

Number of obs = 173
R-squared = 0.0132
consinc | Coef. Std. Err. t
rmort | -.0009941 .0006577 -1.51
_cons | .8957366 .0037146 241.14

here is the scatter


In my heart, I am sure interest rates affect consumption through mortgage financiing. But the aggregate data (analysed in a maybe ok for a blog post level) suggest that this isn’t a big deal.