The title references a famous rap song: "Money, power, and respect / First you get the money / Then you get the #$(#%in', power / After you get the #$#(%in' power / Mutha*#(%#%s will respect you."
We recently covered the long shadow that checks cast on modern financial practice. The most important genre of check historically was the paycheck, settling an employer’s liability to their employee for services rendered. Most paychecks (and direct deposits, and earned wage access vehicles, and similar) are not sent directly by employers, but instead go through a payroll provider. Payroll providers are the original financial technology company, and were some of the first scaled adopters of computers generally.
You have almost certainly depended on payroll at some point in your life and probably never thought much of it. And it is fascinating! Let’s discuss a bit of history, a bit of politics, a bit of financial technology, and a heavy, heavy dash of extremely boring schlep in service of larger societal goals.
A bit of a disclaimer off the top: I’m indebted to a discussion with Ian Zapolsky at Check about some of the history and current practice of payroll processing companies. Check received an investment from Stripe, where I am an advisor (and past employee). Neither Check nor Stripe necessarily endorses what I write in my own spaces. Any errors in fact are, as always, my own. Credit for taste in financially-oriented rap music goes to ChatGPT.
Why does payroll exist, anyway?
Payroll, as an institution, exists to support the deputization of (mostly private) employers as tax collectors by the state.
In the United States, prior to the Civil War, almost all federal spending was funded by tariffs on alcohol or tobacco. The war was extremely intense and expensive, and the winners mostly chose to fund it by instituting a temporary income tax, at a flat rate of 3% of income above a relatively generous threshold. This might or might not have been legal, strictly speaking, much like the contemporaneous suspension of habeas corpus, but needs must when the future of the nation is on the line. (A note to non-American readers: you can Google this if you want to get into fascinating minutiae of U.S. history, but take note that a disturbing percentage of authors who are extremely invested in Abraham Lincoln’s supposedly tyrannical behavior on taxation are using that to distract the reader from the institution of slavery.)
The actual collection of this tax was from taxpayers directly. The combination of the tax being new, tax incidence being a relatively small portion of the population, and taxpayers being mostly on their own recognizance lead to a fairly large “tax gap”, which is the wedge between taxes theoretically owed to the government (per the law) and those actually reported by taxpayers.
America re-instituted the income tax in the early 1900s (via a constitutional amendment, this time) as a result of that most cherished of American traditions: widespread populist revolt over tax rates. Specifically, tariffs on imported goods hit the pocketbook of the emerging middle class the worst. The middle class exercised (and exercises) substantial political heft in America, and directed its representatives to move more of the tax burden to the wealthy. The term of art for this in tax policy is “progressive” taxation (where one’s effective tax rate goes up with income and/or wealth); any tax on consumption (including tariffs) will almost definitionally be regressive (the opposite), because the wealthy consume less of their income and/or wealth than other social classes.
The first implementation of it attempted to do lump sum collection, but taxpayers found this surprising and unpleasant. In particular, a large portion of the economy was still agrarian, and the farmer’s harvesting calendar is impossible to reconcile with a single fixed national tax deadline. (As both farmers and people who play Stardew Valley know in their bones, not all crops arrive on the same schedule.)
Enter tax withholding
Actors in government had a bright idea: instead of requiring a single lump sum payment of income tax annually, they could spread the payment over the course of the year. It would have been too operationally intensive for government to arrange this with every wage owner directly. Taxpayers would have revolted about the marginal paperwork burden, to say nothing of the constant reminder that they were paying taxes. So the state deputized their employer as a tax collector to accomplish it.
The mechanism would be withholding: the employer would, using a formula, estimate the taxes you would owe for your next payment, divide them ratably over the year, and withhold an estimated tax payment from each payment to you. They would “relatively swiftly” convey this payment, on your behalf, to the government over the course of the year. You would file a return once a tax year, reconciling your reported income and withholdings made by your employer, resulting in either you paying the difference to the government or, more commonly, receiving a refund from the government for tax overpayment.
This solves many operational issues for the government. It makes its own income much smoother and more predictable than under an annual payment regime. (This returned government to income schedules which more closely matched those of excise taxes. Although business owners might understandably feel otherwise, it is not in fact the case that all alcohol is consumed on tax day.)
Withholding also reduces fiscally consequential tax non-compliance. Most individuals who forget to file a return have already paid all taxes that are due, and may be forgoing a refund against their own interest.
It also has one side effect in the political economy, whose consequences we deal with a century later: withholding is quiet and sending in tax payments is loud. Most taxpayers, including relatively financially sophisticated taxpayers, do not carefully scrutinize their payroll withholding statements. Most do not understand how their payroll tax is calculated; it is simply a law of nature, handled by dutiful clerks somewhere. The only time they really think about tax withholding is once a year when the government pays them. In that blissful moment, their thought is not “Wow, I am paying so much in taxes!”
Tax withholding is one of a very few ways in which you can licitly bribe someone with their own money.
An aside about tax preparation software
It is widely believed in the tech industry that the reason the United States requires taxpayers to calculate their own tax returns, which is not required in many peer nations, is because Intuit (who make Turbotax, the most popular software for doing one’s taxes) spends money lobbying policymakers to oppose the IRS creating a competing product. People who believe this have a poorly calibrated understanding about the political economy of taxation in the American context.
I will have to take notice about uncontroversial but politically inflected facts about the world we live in to describe why you must use the software you use. If you’d prefer to not get politics mixed in with your finances and software, mea maxima culpa. That said, a democratically accountable government which deputizes the private sector to achieve state aims is invariably subject to the political process, and this is on net a good thing. To the extent one has a complaint about the outcome, one's complaint is not with some unaccountable or corrupt actor in a smoky backroom somewhere. It is with one's countrymen.
In particular, the tech industry zeitgeist that blames Intuit for us needing tax preparation software fails to understand the preferences of Congressional representatives of the Republican Party. Any fairminded observer of U.S. politics understands the Republicans to be institutionally extremely interested in tax policy (and tax rates in particular), in the sense that doctors are interested in heart attacks. Their most recent platform includes the quote “Republicans consider the establishment of a pro-growth tax code a moral imperative. More than any other public policy, the way the government raises revenue—how much, at what rates, under what circumstances, from whom, and for whom—has the greatest impact on our economy’s performance.” This is far from the only flag proudly planted by the elected representatives who enjoy the enthusiastic support of about half of Americans for their views on tax administration.
The specific policy implications of those shared values are frequently outsourced, in a fashion extremely common in Washington and critical to your understanding of U.S. politics. Washington has an unofficial ecosystem of organizations and public intellectuals who, by longstanding practice, have substantial influence on policy. When a Republican candidate promises to voters that they are anti-tax, as their voters (particularly in primaries) demand they must be, the thing they will offer in support of that is “Grover Norquist gave me a passing grade.”
Norquist runs Americans for Tax Reform, a non-profit political advocacy group which opposes all tax increases. ATR is institutionally skeptical of withholding, because they believe that withholding allows one to increase taxes by stealth. I don’t think it is excessively partisan to say that, if one phrases that claim a bit more neutrally as “withholding increases tax compliance by decoupling public sentiment and policy changes,” the people who designed the withholding system would say “I’m glad the National Archives makes our design documents so accessible. We wrote them to be read!”
And, relevant to the question of whether Intuit controls U.S. tax policy: it can’t, because that would imply they have wrested control from Norquist. Norquist considers a public filing option a tax increase by stealth and opposes it automatically. (I offer in substantiation ATR’s take on a specific policy, which was bolded for emphasis in the original: “Americans for Tax Reform rejects the use of unauthorized taxpayer dollars being used to expand the IRS into the tax preparation business and urges states to reject participation in the program.” You can find much more in the same vein.)
Anyhow, back on the topic of withholding taxes: they still have political consequences, including consequences directly relevant to how individual Americans interface with their government, more than a hundred years later.
Withholding taxes were an operational disaster in early implementations
Calculating, collecting, and remitting withholding taxes was extremely labor intensive in the early years. Businesses used to have entire departments of payroll clerks, who needed to do per-employee calculations manually or on primitive mechanical calculators.
Fun fact: this was one of private industry’s first scaled uses of women in the workforce, and the job title was “calculator.” AI took our jobs and thank goodness it did. The typical workload of a (human) calculator persists in the present day only in Japanese megacorps which, and this is in no way an exaggeration, use it as a form of a psychological torture to coerce employees to quit “voluntarily.” OK, I guess that is less of a fun fact.
Withholding was designed without substantial input from the business community. It was so hated that it almost triggered a "tax revolt"; refusal to pay taxes as an act of protest. After World War One was over, the War Revenue Act (which had traded phasing out withholding in return for phasing in more onerous income reporting requirements) lapsed, and planners got a few years to debrief what had gone wrong. They had a better game plan ready for 1935’s rollout of the Social Security Act, which assigned Americans unique(-ish) identification numbers for the first time. This time, user research included America’s relatively nascent large national employers like e.g. Ford and GE, who exactly dictated what they’d need from the government to make withholding happen on a scale of hundreds of thousands of employees. (Think how many rooms of clerks that implied! The imagination boggles!)
In addition to the clerks, Big Business demanded the government accommodate modern practices like using specialized hardware from the Computing-Tabulating-Recording Company. It had recently completed a corporate rebrand to reflect their cutting-edge technological supremacy: International Business Machines. To a very real degree, the spec for U.S. withholding taxes became “What can be quickly calculated on a machine that IBM can deliver production units of this year? Write that law and we will send a check to Washington, every month, like clockwork. Or have it your way. Write another law. Then, you can come to Detroit and explain to the boys they need to give you what's coming to you.”
So what happens in payroll, anyway?
In broad strokes payroll processing hasn’t changed in a hundred years:
A business owner, or their payroll department, first calculate the amount of money they owe to employees in a pay period. The length of that pay period is established partly by custom (two weeks is most common in the United States, monthly is more common in Japan) and partly by operational complexity. Weekly was considered too exhausting for payroll clerks to contemplate decades ago and most businesses haven’t moved to it due to the persistence of tradition. (They have made some changes to get workers their pay faster, which is a topic we’ll return to.)
During the process of running a payroll, the payroll department or payroll provider will apply current laws to calculate deductions from payroll. Almost all of these will be taxes or tax-adjacent, like contributions to tax-advantaged employee retirement plans. There are some important exceptions which you absolutely must support, like e.g. garnishing the wages of parents who are delinquent on child support obligations (or, for that matter, certain delinquent debts owed to the nation).
In parallel with this calculation, money has to fan out. Is the service provided by payroll providers the calculation, the recordkeeping, or the money movement? Yes.
In U.S. practice, payday is almost inevitably on a Friday. The payroll provider will make an ACH debit against the employer’s bank account within a few hours of them running payroll, ideally (from its perspective) on Monday. This moves the total cash outlay for payroll in a single transaction to a settlement account owned by the payroll provider. Then, on Thursday night or early Friday, it will schedule payment via ACH push or other mechanisms to each employee. Typically, that will have that money be spendable in their accounts at Friday. (ACH direct deposit has almost entirely replaced physical paychecks.)
These are not the only payments made by payroll providers, and one might argue they are not the payments made which actually justify the existence of payroll providers, but they are the ones most people think of when they think of payroll. We will return to the other payments in a moment.
Why did businesses largely choose to outsource payroll processing? Because jurisdictions with taxing authority saw the federal government succeed with withholding taxes and copied the effort. There are many thousands of jurisdictions in the U.S. with this authority. Even at very small scales of business, decisions which seem very minor about who you employ suddenly pull in huge amounts of the collected tax code of jurisdictions which might feel Very Far Away to you. Keeping up with what laws and rulings you're subject to and the changes to them very quickly becomes a full-time job. In addition to being basically the first fintech, payroll processing is the first regulatory compliance layer as a service. (You'd predict from this, mostly accurately, that payroll processors are far less economically dominant in e.g. Japan, where taxing authority and collection is far more centralized than in the U.S.)
“Where is the risk transfer?”
That was my first question for Ian, and we both laughed ruefully at it. Almost every movement of data and money in finance also includes an (often underanalyzed!) risk transfer as part of the price of the services.
The biggest risk in payroll is that the business’ bank account doesn’t have money to cover the payroll debit in full, in which case the debit will be “returned” by the bank, resulting in the payroll provider not getting the money in their settlement account. Now, counting days on your fingers like all good financial technologists, you might think this is unlikely to be a problem. Monday plus two business days for an ACH return window is Wednesday. The money can’t leave before Thursday. No overlap in timeline means no risk.
So here’s the rub: that Friday deadline is, by law and custom, very hard. If you, as a business owner, fail to have money delivered to employees by Friday, you have “missed payroll.” This is one of the most legally consequential screwups a business owner can do. How developed is the law in this area? Missing payroll is listed in the Bible (James 5:4, etc) as a “sin crying out to heaven for vengeance.”
The Monday deadline, on the other hand? Eh, business owners sometimes get busy. I probably averaged Wednesday for payroll submission at the last company I ran. “Thou shalt not submit thy CSV files on Wednesday” is not written in the Bible.
And so here is the product decision confronting every payroll provider: do we strictly impose the Monday deadline, or do we let favored customers slip it until, say, Wednesday? OK, maybe gulp Thursday afternoon? Or do we cede the custom of many, many employers to a provider capable of playing ball here?
The payroll industry overwhelmingly allows businesses to transform “late” payrolls into “on-time from the perspective of the employee” payrolls by taking balance sheet risk. They pay the employees before they know they will collect from the employer in a durable fashion. This is a considerable risk.
In the event a business misses payroll, the likelihood that it is not simply an operational stumble but actually insolvency is, over the universe of all firms, very, very high. Individual runs of payroll are very large relative to the fees earned by payroll providers; at my last company, two weeks of salaries/etc for engineers/etc was ~$200,000 against a payroll service charge of a few hundred dollars a month. Clearly, taking credit losses even at a very low incidence rate does not pencil out well. (A payroll provider will typically reserve the right to recover against employees if there is a return on a payroll run. Nobody likes that resolution. One reason: employees of a defaulting employer will frequently not have the money a day later, either, not least because their employer might have told them “Uh you probably don’t want to keep your last paycheck in the bank over the weekend, call it a hunch.”)
And so payroll providers, unlike most software companies in the we-run-a-complicated-spreadsheet business, will have to underwrite their customers as they onboard them. Most businesses in the economy are extremely thinly capitalized. Underwriting will often focus on reviewing books, demonstrable (past) cash flows versus anticipated payrolls, etc. Doing this well is make-or-break for payroll providers.
(Other risks payroll providers have to be cognizant of include account takeovers and fraud. “Payroll is a money hose”, as Ian mentioned, and by design will cause the payroll provider to spray money into accounts it has no history with and has minimal documentation on until the first transfer of four figures per account. A legitimate business which gets their payroll account taken over is in for a very bad day. A payroll provider which is convinced to run payroll on behalf of a business which either doesn’t exist or only exists to enable a later crime via a “long fraud” will also have a very bad day. There exist many people in finance and tech companies who think of little other than “How can we identify an account takeover before we or our client lose money in an irreversible fashion?”)
What about those other payments?
A lot of people have read about Warren Buffett’s investing genius over the years. One key to it was using the “float” earned by his insurance companies. Float is, in the insurance context, money which customers have paid you (premiums) which covers risks which you will (as an actuarial matter) owe them back someday. In the present day, you can invest it (subject to some constraints) and keep the returns for yourself.
Many people see float absolutely everywhere. “I bet a payment processor just rakes it in from float!” Well, not so much, because float being interesting requires a large amount of money multiplied by an extended timeframe multiplied by relatively high interest rates. Non-specialists tend to forget that “timeframe” bit and overestimate float by a factor of ten or more. (Payment processors are keenly aware that their customers want their money faster at virtually every margin and are keen on improvements to make this happen, even if they notionally cost a minor revenue opportunity.)
Which I mention to excuse the following: payroll firms actually do earn material amounts of float, in a fashion which is rare for financial technology businesses. Why? Precisely because payroll runs are large relative to fees for payroll services and holding times can be quite considerable.
We are not talking about the float between Wednesday and Friday primarily here. We’re talking about the float earned on “funds held for clients”, which mostly constitute money that they’ve withheld on behalf of particular employees but not yet remitted to the tax agency (or similar).
You can read the disclosures for your payroll firm of choice for the boring mechanics here. I recommend ADP’s, since it is a giant in the industry (~17% share nationwide) and publicly traded. ADP keeps client funds at a trust account at a trust bank that they established. (Was that decision voluntary? One might speculate that if you happen to have plural percentage points of all wages earned in the nation on your books on any particular Tuesday, almost all of it owed directly to the government, some people in Washington would really prefer you had it in a maximally regulator-legible wrapper versus another structure. Anyhow, OCC charter secured, congrats all around.)
So in economic substance, when your Chicago pizzeria runs payroll on Wednesday for payday on Friday, ADP constructively receives the money Thursday-ish and pays out the majority of it Friday-ish.
But the withholding taxes that the pizzeria owes to the IRS and to the Illinois Department of Revenue? Those ADP custodies until the next day they are required to pay it out, which will (in expectation) be about two weeks later, not about one day later. (Various payout schedules prevail for various taxes and various taxpayers, etc, but this suffices to show the general pattern.)
The money is custodied on behalf of the pizzeria—it is still their money—but the interest earned on it goes to ADP. It is a lot of dough: ADP earned more than $800 million on client funds in fiscal year 2023, representing about 18% of EBIDA. Payroll providers aren’t as dependent on interest income as discount brokerages but it is a very important part of the model. (The main revenue line, for ADP and other payroll providers, is the fee they charge businesses for services.)
An interesting detail here: payroll companies are a conveyor belt for money. That conveyor belt looks pretty short if you just count from the payroll being run to the final transfer of withholding tax. It starts to look much, much longer if you consider “Well, if a software company hires an engineer in 2023, and has attrition of 10% annually, in expectation haven’t they basically committed to paying something like 85-90% of current salary in 2025 once you count the impact of annual raises/etc?”
When you multiply that intuition over a large and diversified book of business, and subtract out the negligible churn rates of payroll providers, you (as a payroll provider) could get comfortable conceiving of your conveyor belt as being much, much longer than it looks at first glance. So should money on your conveyor belt be earning demand deposit interest rates? Oh clearly not, short-term Treasury rates will of course be higher and have even less risk. How about mid-term Treasury rates?
And if mid-term Treasury or agency MBS rates are higher than your corporate short-term borrowing rates, you could borrow short and lend long. How do you know you’ll have the cash flow to support the borrowing? Dude. Look at the conveyor belt of money which charges you nothing for borrowing from it.
Welcome to the wonderful world that is being a corporate treasurer at one of the fairly few companies where one’s fiscal asset allocation strategy is both really interesting and also really matters to results.
A detail of a detail here: the flipside of this being an excellent reason to be in the payroll business is, indirectly, one reason why payroll providers are so terrible. To do this licitly, you need to get a money transmitter license everywhere your customers want to pay as many as one employee, which (very quickly) means all 50 states and then some.
This is not hard relative to creating a fusion reactor, but is hard relative to creating a functioning Rails app. This means that the minimum cycle time to create a new functioning payroll provider is “a few years” unless you’re using an underlying provider via some sort of white label arrangement. Most ways to do that just end up replicating their infelicities, so really innovating requires all the licenses first, which implies years of boring scutwork prior to doing the valuable, fun, and lucrative bits of innovating on payroll software.
Where is the frontier in payroll?
It was a revelation a few years ago when one startup payroll provider (Gusto, about which I will say “I have used them as an employer and yet I have employer-side payroll stories that take longer to tell than ‘I ran payroll successfully’”) debuted a web application for employees which wasn’t terrible. Pedestrian applications of technology create enormous value in financial services, film at 11.
In terms of actually new capabilities, the hot new thing in payroll recently is called “earned wage access.” This is mostly for employees at the lower end of the socioeconomic ladder. We covered that the payroll cycle in the U.S. is generally two weeks, for legacy reasons owing to the workload capacity of payroll clerks. In principle, computers could trivially do it daily, but business owners largely have not decided to do that.
Employees low on the socioeconomic ladder are therefore, economically, extending their employers a loan of (on average) one week’s worth of wages at zero percent interest. They are themselves are often in debt, frequently paying high or unconscionable rates on that debt.
EWA attempts to make it incentive compatible to fix this. In it, the payroll provider convinces the employer to allow employees to opt-in to receiving their wages not as a paycheck or an ACH deposit into their bank account but as a deposit onto a prepaid card provisioned by the payroll provider. That deposit is made daily not biweekly. Critically, these wages are being advanced by the payroll provider and not actually coming from the employer, which is probably surprising. (This is free to the employer; no cost aside from the wages/etc you’ve agreed to pay.)
Why is this incentive compatible for the payroll provider, when it could be advancing wages to e.g. a pizzeria chef who might depart employment two days into that 14 day pay period? Because people who are relatively low on the socioeconomic ladder tend to spend most of their paycheck relatively soon after they receive it, and much of the spend will be by presenting that prepaid card to pay for e.g. groceries. Every business accepting the card will pay interchange to do so, in the neighborhood of 2-3%.
And so you can napkin math this out something like this: an employee earns 250 days of wages in the year. If they spend half of their wages via using the prepaid card (with the other half being tax withholding, rent, and similar things which will not result in the provider using interchange), the provider will earn something like 2.5 days of wages for doing financial engineering. They will also absorb, on average, 7 days on average of risk of separation from employment. And therefore, factoring in estimated annual risk of separation and anticipated recovery in event of separation, both of which can be measured empirically as you roll out the offering, there is some solution to the equation where the payroll provider earns a day’s wages without working a day.
Then, it does that for tens of thousands of employees in parallel. This is lucrative relative to earning the interest alone on a small percentage of a few weeks' wages, or on the rates businesses will tolerate to run payroll for one pizzeria employee at the margin.
Other fun innovations include giving people money in ways they find actually valuable. A major one for Check has been enabling instant EWA-style to employees’ Cash App accounts. (Cryptocurrency advocates will, inevitably, mention that this would all be so much simpler as less expensive if everyone just used stablecoins. Convincing U.S. workers to actually desire stablecoins as opposed to e.g. Bank of America dollars or Cash App dollars has been rather more evitable.)
Coming up next in Bits about Money
Merry Christmas, somewhat in advance! I’m planning on a recap issue for the year, likely on the Friday before Christmas.
If you’re looking for late Christmas gifts, wondering how to spend your continuing education budget while nervously glancing at the calendar, or perhaps want to book an expense in 2023 and save on taxes, Bits about Money offers paid memberships.
After the year in review, we’ll return to the usual potpourri of topics. One relatively close to the top of the list is the check-adjacent alternative finance products that are reinventing the paycheck cycle for people lower on the socioeconomic spectrum. (In contrast to EWA, above, some of them don’t require integration into the payroll provider and/or facilitation of the arrangement by one's employer.)
Speaking of payroll software, a thing I’d love to write at some point is why government payroll system modernizations are the most doomed and cursed of any software projects. Perhaps next year!
Want more essays in your inbox?
I write about the intersection of tech and finance, approximately biweekly. It's free.