With the way that people currently think about things, the greatest threat of inflation comes from the supply side. That is, a recurrence of the 1970s. To justify (and more thoroughly define) this, though, I think a more in-depth discussion of what inflation is might help.

There's a somewhat famous accounting identity (which is to say, an equation that is true by definition) that mv = pq. Here, m is the amount of money, v is the frequency with which money is spent, q is the total quantity of goods produced and sold, and p is the average price of goods. The equation mv = pq is true because both sides are equal to nominal GDP: mv is the total amount of money spent, which is to say, the total price of good purchased, while pq is the total price of goods sold. Of course, more goods can't be bought than sold or vice versa.

This equation is also useful for thinking about inflation, because in this context, inflation is defined as changes in p.

Now, the meme that an increase in the money supply causes inflation comes from the assumptions that v is relatively fixed, and that q cannot be changed by changes in m.

I'll first briefly talk about v, as it's the simplest, and in a deceptive way, among the most important. First, there are institutional constraints on the velocity of money, for example, people are only paid so often, dividends only pay out so often, and anyway, people only want things at certain rates. Taking that into account, v primarily represents whether people are net saving or net borrowing. Money in savings doesn't get spent, hence saving reduces velocity. On the flip side, leveraging money allows it to be spent more often. The relevance of this will come up later though.

So, what happens if the money supply changes? In principle, any of the three other variables can change to accommodate it. In the real world, though, it isn't true that changes in the money supply drive changes in the price level, or at least, it's not true that the other two variables aren't affected by change in money.

Money velocity can be accounted for with two observations. First, in general, people prefer to save, and the amount of savings scales with income. Second, people want to maintain or modestly improve their lifestyle, which means, people want to spend money with the same or modestly increasing frequency. If income growth doesn't keep up, people will take on credit to compensate, which is to say, if money supply goes down, money velocity will tend to go up to compensate, and if money supply goes up, people will save more (driving money velocity down).

Now then, let's assume that people are spending more money, meaning, mv is going up. What happens? Either p must go up, or q must go up, or some combination. That is, there's either output growth or inflation, or both.

But which happens?

Assume there's two companies A and B that both make widgets, and assume that there's unmet demand out there for widgets. Which is to say, there was one level of spending (i.e., one value of mv), and at that level people were deferring their desires for widgets in favor of eating, and now there's another, higher level of spending, either because there's more money, or because it's being spent faster. At this new level of spending, people are now able to satisfy some of their desire for widgets, so they direct their newfound spending into the widget sector.

Of course, our companies aren't mind-readers; they're only capable of noticing changes in their widget sales figures, or in the final case, they're capable of seeing changes in their leftover inventory at the end of an accounting period. (Note that they can't know changes in aggregates before those aggregates are available; the first indication they'll have is changes in inventory.) So A and B see that they're selling out of widgets; what should they do?

In principle, there are two possibilities: they can charge more for widgets, or they can make more widgets. Let's say company A assumes that this is just inflation at work, and raises prices, while company B believes there's a change in demand, and they make more widgets. In this case, company A gets screwed and company B gets all the business. Hence, there's certainly no good reason to think that changes in m can't drive changes in q. This is, in any case, the principle behind stimulus spending.

What would cause both companies to raise the widget price? One answer is, if neither company is capable of making more widgets. This could be the case if there are no unemployed workers who want to work the widget assembly line, or it could be the case if there's an energy or land shortage, or if their existing factories are operating at capacity and they can't build more.

Another answer is if the cost of business is going up. The most obvious reason this could happen is if wages are going up (cost of inputs going up is also possible, and we'd have to ask the same question about the producers of their inputs: is it wages, or is it productive capacity?), but what would drive wages up? The answer is, if employers have to compete for employees, which is to say, if unemployment is very low. This is the origin of the idea of the "natural rate of unemployment," which is, the unemployment rate at which employees are capable of demanding higher wages because of the relative scarcity of potential employees.

The problem with the natural rate of unemployment is that there's no particularly good reason to say it should lie at any particular rate. Is it 4%? 6%? 8%? (There are also very sound proposals to effectively eliminate unemployment through a job guarantee that turns employment into a buffer stock, removing wage competition at the low end, but that's a whole other tangent.) Certainly, at least, we're nowhere near the natural rate, so this source of unemployment is not a relevant consideration for the foreseeable future.

So let's get back to the case where the widget manufacturers want to make more widgets but can't, so they raise prices. There are two reasons this can happen. On the one hand, demand for widgets could grow so high that the widget business can't expand any more (for example, maybe widgets require steel, and steel is already being used at the same rate it's produced, and in turn steel production is limited by the rate at which iron can be prospected and mined), which is demand-pull inflation. On the other hand, demand could stay the same, but the ability to make widgets could collapse, for example, due to resource supply shocks, which is supply-pull inflation.

This latter case, supply-pull inflation, was what happened to Germany in the 1920s, it's what happened to Zimbabwe recently, and it's what happened to the US during the 1970s. And that's always the more serious inflation risk, since contraction is the only solution. I'm not aware of any modern incidents of demand-pull inflation, and the only truly demand-pull inflation episode I'm aware of concerned overmining of silver in the New World. Hence I tend to operate under the belief that, unless the government is seriously devoted to creating demand-pull inflation, it's a total nonentity.

So sorry that was so long, but I think it's important to be thorough when it comes to the subject of inflation, because there's so much intuition out there, and length is the only way I know to compensate for that.