What does it mean? -- "If you lower interest rates, people will choose to hold more cash."

"If you lower interest rates," Scott Sumner says, "people will choose to hold more cash." He provides this graph showing "the demand for currency (as a share of GDP) and T-bill yields":

"Notice that the two variables tend to move inversely," he says. Sumner uses the graph as evidence of his claim that "If you lower interest rates, people will choose to hold more cash."

I have to ask: More cash, compared to what? More cash, compared to the money people have in total, I think. If you lower interest rates, Sumner is saying, people will choose to hold more of their money as cash.

The argument makes sense: People holding cash don't earn interest on it. If interest rates go down there is less incentive to hold money in an interest-bearing form, and more incentive to hold money as cash. If interest rates go up, there's more incentive to hold money in an interest-bearing form and less incentive to hold it as cash. This is his argument.

But Sumner is careful not to say "hold more of our money as cash."


If you wanted to show the relation on a graph, you'd want to show an interest rate. This, Sumner does. And you'd want to show the amount of cash people are holding, compared to money held in total. The ratio would show people's preference for holding money as cash. This, however, Sumner does not do.

Sumner's graph shows cash relative to GDP. The context variable is wrong. The graph does not show cash compared to money in total. The graph does not show what it must show to be relevant evidence.

The lines on Sumner's graph seem to support his statement about holding more cash, because the lines move in the directions they should move if his statement is true. But the blue line does not show the ratio that his statement describes.

Sumner's graph is not relevant evidence.


I pointed out the discrepancy on Sumner's blog:

Scott, why does your graph not show currency relative to *money* ?

You say that when interest rates fall, currency holdings rise. To my mind “currency holdings rise” is the same as “we hold more of our money as cash”. To look at “holding more of our money as cash” the graph would have to show currency relative to M1 or some broader money.

Sumner replied:

I use currency to NGDP, because I’m interested in explaining changes in NGDP, not the broader aggregates.

Of course he's interested in NGDP. The whole focus of his blog is NGDP targeting. That's fine. But it doesn't address the issue.

The ratio of currency to NGDP says nothing about the preference for holding money as currency. The ratio of currency to NGDP is not evidence of the desire to hold money as cash. Sumner's graph is not evidence of his claim. And his response to my comment does not resolve the matter.

Am I looking at this wrong? Or did Sumner try to bullshit me?


// Edit, 11 Jan 2019, changed
"More cash, compared to the money people hold in total, I think" to
"More cash, compared to the money people have in total, I think".
Now that I know what "holding" money means.