>> $0.01-$0.02 per transaction * millions of transactions per day has got to make some accountants happy.
You'd think, but actually no. That's because it's per transaction, so the accumulated amount is well pennies compared to the turnover.
For example, say you always rounded in favor of the company. That means an average of 2c gain per transaction. If you do 2 million transactions a day that's... 40 grand! yay. Except that if the average transaction is say $50 (which seems kinda low) that means your turnover for the day was $100 million.
So, in short, if you are making a lot from the rounding error, well you're making a lot lot from the actual sales.
Single turnip is ridiculous sure but that is hardly the average sub 5 $ purchase. I have seen people check out with single pieces of fruit or like a gallon of milk regularly. Plus, in my state, you could easily get a single deli item for lunch or breakfast for ~5$ or under. Basket size (which is the term groceries use for the average purchase total) is about 50$ although it has gone up in covid19 times. That is the mean though- there is a ton of ~5$ and sub purchases
I work near a grocery store and quite a few people by a salad or a small snack for lunch. Some also quickly go and some small things that have run out, i.e. some milk, before heading home.
This might also be a currency difference, through; 5 Euro in Germany can get you quite far (at least outside of metropolitan areas).
I have seen people purchase a single Brussels sprout in order to use a coupon that requires 'any purchase' to get a free item. Getting two of the same item is more common, but having someone go back to find the smallest vegetable that exceeds the tare isn't unknown.
This is similar to how HFT firms make a profit. Except it's generally less than a penny but it's per share per transaction (trade). We can thank the Sub-Penny Pricing Rule of Regulation NMS for this feature. An order for 1000 shares of XYZ stock that is front-run for half a penny per share is $50 dollars profit. Rinse and repeat.
>We can thank the Sub-Penny Pricing Rule of Regulation NMS for this feature
Are you sure? I did a quick search and came up with this
>One of the rules in Regulation NMS is a new Sub-Penny Rule: “which establishes a uniform quoting increment of no less than one penny for quotations in NMS stocks equal to or greater than $1.00 per share to promote greater price transparency and consistency. . . . In particular, Rule 612 addresses the practice of “stepping ahead” of displayed limit orders by trivial amounts. It therefore should further encourage the display of limit orders and improve the depth and liquidity of trading in NMS stocks.”
which suggests the opposite of what you're claiming. Also, I'm not sure how you'd even make money this way, considering that NBBO requires brokers to execute their customer's trades at the best available price.
The NBBO ensures you get the best price available on public exchanges. Public exchanges are not the only places where stocks are traded.
If the best price of a stock on exchanges is $25.00, but an HFT can buy it at $24.99 on a dark pool, they could buy it cheaper and sell it to you, and pocket the difference. When they pay your broker for the privilege of doing so, it's called payment for order flow. As a retail trader you have no visibility or access to dark pools.
Aside from dark pools, NBBO updates have latency which may be exploited:
Not matter how they do it, the fact that they are willing to pay your broker to execute your order is proof that they have some way to make money from your orders.
Your criticism of my attributing the cause of the 'feature' to Reg NMS is fair. Although it is worth looking at the entirety of Rule 612, especially paragraph (c)[0].
You are correct that brokers are required to execute customer trades at the best available price aka the NBBO. The issue here is that the NBBO is relatively slow as compared to data feeds offered by various exchanges and many brokers use the NBBO simply to satisfy Reg NMS Rule 603(c)[1].
If price data was able to travel instantaneously then the NBBO might represent the true best price(s). But the laws of physics say it isn't so and orders can be executed at a less-than-best price yet still at or better than the NBBO.[2]
Fortunately, more are aware of this these days including the SEC which released a proposed order calling for exchanges to submit revised NMS Plans for consolidated data earlier this year.[3]
You'd think, but actually no. That's because it's per transaction, so the accumulated amount is well pennies compared to the turnover.
For example, say you always rounded in favor of the company. That means an average of 2c gain per transaction. If you do 2 million transactions a day that's... 40 grand! yay. Except that if the average transaction is say $50 (which seems kinda low) that means your turnover for the day was $100 million.
So, in short, if you are making a lot from the rounding error, well you're making a lot lot from the actual sales.