So I looked at the serial-numbers lists that are available online (tw-db, typewiterdatabase and an article in an older issue of ETCetera). The numbers suggest sequential numbering, starting their numbering in 1906 and numbering sequentially right up to 1941. In that case, the serial-numbers should give the annual production volumes. (From these numbers, I've pegged the start of the '3' at 20,000 to get a ~6000 run in 1912. On the other hand, earliest reported serial number for a '3' is 12,172. There's more confusion on the numbers and volumes, e.g. from the serials the 1919 I get 81k and not 153k. The serials up to 1921 seem a bit suspect, ending with 999.)
Ignoring all the uncertainties and confusion and just taking the volumes from the serial numbers, these can then be plotted over time:
Putting this in a graph shows the meteoric rise of the little portable. It also shows that it's success comes to a rather abrupt stop, with some minor 'revivals'. Looking at this graph with some related events in mind, I've speculated a bit about some of the how and why.
I can just imagine that the introduction of the Remington Portable with a full keyboard was a nasty shock for Corona. Coming on top of Underwood entering the fray in 1919 with a 3-bank and Corona still busy seeing off the Baby Fox. Their numbers were heading for 100k/year and suddenly the year 1921 ends with 20% down. (There would be some lag, with production reflecting sell-in to the trade feeling a drop in sell-out (from trade to consumer) a bit later, so the decline in sales may have started in e.g. late 1919.)
Plotting the volumes of these three early compact portables shows that the decline in sales of the '3' was not so much a decline in the market of compact portables overall, but much more likely due to the new competitors in the market.
With competition in the compact portable market-segment now from both Underwood and Remington, Corona need to respond. These competitors are too large to drive off with advertising or price promotions. Trying to sue them into oblivion (Fox) would have been out of the question for the same reason (even if there had been a possible case). Even worse; their products compare very well against the little '3'.
If work had not already started on making their own 4-bank offering, it was doubly important to do so now.
In the mean time, a bunch of improvements is made to the '3' over 1921-1923. The machine is widened, with shift keys at both left and right and the carriage is also widened. Automatic ribbon reversing in 1922 (thanks to Mr Kernighan for pointing out that datapoint!) from serial number 500,000 (?) up, would have allowed for some 'news to the trade' to drum up sales. With these improvements the '3' compares a bit better to the competition.
Maybe this contributed to a slight rise again in '22, maybe the 10k less in '21 reflected the trade still having too much stock from '20. Who can tell... Around that time (1923?) amongst other improvements an extra two keys are added.
They needed to live off their old product until they could introduce their own full keyboard portable with the 'Four' in 1924. Then in 1924 immediately the volume of folding '3' machines plummets. Advertising from then on is showing the new flagship product 'Four' model only.
On an aside; with their market (and likely their margins) under pressure, working on their own response, they would be investing in the new product and tooling. This probably means that the financial position of Corona would be at a relative low in 1924, with the launch-costs of the 'Four'.
With the introduction of the 'Special' in (I think?) 1929, the level is again at a respectable 20k, showing a drop off over the following couple of years as would be expected after a new product introduction. Adding a splash of color and fashionable paneling makes also the '3' product sell again (a little). Some examples from a quick image search:
What is quite remarkable is the variation; a choice of both base and panels in colors, but also plain lacquered in one color with no crinkle panels. Again odd is that these are all marked 'Special' on the front panel but no longer have the automatic reverse. Seems Corona was having a ball with product-diversity. This is not good for cost control today, and my guess is this was not good back then either.
The year 1934 again shows a small spike in volume. This may be just the random fluctuations of a low-volume end-of-life product, or it may have been a new advertising push to counter the very low-cost market-segment. The price is by then also quite low, maybe this contributes to a bit of volume again.
However with the price now down to USD 24,50, even without any tooling cost amortization the margins would not have been great anymore. Material and labor still needed to go into the machine at about the same level.
Keeping a type in production over such a long period after 1924 with such small volumes (and margins) raises the question if that wasn't at least partially a sentimental, emotional decision.
Or rather; I suspect that nobody in the Corona company would ever dare to suggest to axe the folding Corona 3 - their core, founding product.
It happens :)