Nick Szabo's papers and concise tutorials

Accounting Controls

Copyright (c) 1997 by Nick Szabo
permission to redistribute without alteration hereby granted

Outside of the financial cryptography community, and long predating it, there is a deep tradition of protocols used in the course of performing contracts. These protocols consist of a flow of forms ("data flow", canonically displayed in data flow diagrams), along with checks and procedures called "controls". Controls serve many of the same functions as cryptographic protocols: integrity, authorization, and so on. This article uses "control protocols" or simply "controls" to refer to this combination of data flow and controls.

Control protocols, and the professions of auditing and accounting [ 2 ] based on them, play a critical but ill-analyzed role in our economy. Economists lump them, along with other costs of negotiating and ensuring the performance of contracts, under their catch-all rubric of "transaction costs". But without controls, large corporations and the economies of scale they create would not be possible. Controls allow a quarrelsome species ill-suited to organizations larger than small tribes to work together on vast projects like manufacturing jumbo jets and running hospitals. These control protocols are the result of many centuries of business experience and have a long future ahead of them, but the digital revolution will soon cause these paper-era techniques to be dramatically augmented by, and eventually integrated into, smart contracts.

Controls enable auditing of contract performances, allowing more precise inference of the behavior of an agent. Auditing is costly, so it is undertaken by random sampling. Economists study the substitutability between the probability of verifying a breach and the magnitude of legal fines, where physical enforcement is used. Conceivably, one could substitute increasingly high penalties for increasingly rarer and less expensive auditing. However, this is not robust to real-world conditions of imperfect information.

Since controls primarily address the implicit contracts between employees and employer, there is little mapping from contract to control. A secondary function of controls to to monitor contracts with other organizations. Here there is some mapping, but it is confounded by the integration of the two functions in most controls. Rather than based on contractual terms, controls are typically based on managerial authorization.

Controls are typically based around amounts of money and quantities of goods. A canonical control is double entry bookkeeping, where two books are kept, and there must be arithmetic reconciliation between the books. To conceal an irregularity, necessary to omit from both sides, or to record entries offsetting the irregularity.

Notice that there is a problem distinguishing error from fraud. This problem crops up in many areas in both auditing and smart contracts.

To illustrate, here are two common control techniques:

Imprest: this is a family of controls involving the receipt or disbursement of bearer certificates (usually notes and coins). One example is the protocol used at most movie theaters. Entry is segregated from payment by introducing tickets and establishing two employee roles, the ticket seller in a booth, and the ticket stub salesman at the entrance. Periodically, a bookkeeper reconciles the number of tickets with the total paid. Discrepancy again indicates fraud or error.

Customer audit: Techniques to get the customer to generate initial documentation of a transaction. For example, pricing goods at $.99 forces the employee to open the cash register to make change, generating a receipt.

A complete control protocol typically features the generation of initial documentation, segregation of duties, and arithmetic reconciliation of quantities of goods, standard service events, and money.

Of these, the segregation of duties deserves special comment.

It has long been recognized that an intermediary is more trustworthy when it is distributed. In a large business, transactions are divided up so that no single person can commit fraud. Segregation of duties is an instance of the principle of required conspiracy. For example, the functions of warehouse/delivery, sales, and receipt of payments are each performed by different parties, with a policy that each party reports every transaction to a fourth function, accounting. Any singular reported activity (e.g., delivery without receipt of payment) indicates potential fraud (e.g., a delivery was made to a customer and the payment pocketed instead of being put into the corporate treasury). Segregation of duties is the auditor's favorite tool. Where it is absent the auditor cries "foul", just as a good engineer would react to a single point of failure. Many cryptographic systems have rightfully gone down to commercial failure because they ground down to trust in a single entity rather than segregating functions so as to require conspiracy.

There are least three significant differences between the scope and emphasis of smart contracts and controls. Controls are paper-era protocols designed around static forms, place little emphasis on confidentiality, and are based on management authorizations rather than one-to-one relationships.

Smart contracts can be based on a wide variety of interactive protocols and user interfaces, and can be involved in a wide variety of kinds of contractual performance. Control protocols, developed in the era of paper, are based on static forms passed as messages and processed in tables and spreadsheets. Controls focus on money and counts of standardized goods and service events, easily recorded by numbers and manipulated by arithmetic, while mostly ignoring other kinds or aspects of contractual performance. Checksums on numbers, the basis of reconciliation, are crude and forgeable compared to cryptographic hashes. Electronic Data Interchange (EDI) keeps these static forms and maintains reliance on controls. It uses cryptographic hashes for nothing more sophisticated than integrity checks on individual messages.

Controls place little emphasis on confidentiality, at least in the modern accounting literature. The emphasis on confidentiality in paper-era protocols is lacking because violation of often implicit confidences, via replication of data, was much more difficult with paper. Furthermore, technologies for protecting confidentiality while auditing were not feasible. Businesses traditionally trusted accounting firms with confidences, a trust that has eroded over the last century, and will erode still further as accounting firms start taking advantage of the vast amounts of inside and marketing information they are collecting from their customers' databases during audits. Using paper-based protocols in a digital world, there are few effective controls against the auditors themselves. Post-unforgeable transaction logs and multiparty secure computation indicate the possibility of cryptographic protocols to implement less relavatory but more effective auditing trails and controls; their use may be able to ameliorate the growing problems with data mining and breach of confidentiality.

Auditors place quite a bit of trust in management to authorize transactions in a secure and productive manner. Objecting to this dual trust in management and distrust of employees inherent in the accounting tradition, there has been a trend in the last two decades towards a loosening of controls as a part of hierarchy flattening and empowerment of professional employees. Unfortunately, loose controls have led to several recent scandals in the banking and investment trade. The most recent view is that there must be a learned tradeoff between controls and empowerment.

These traditional protocols have a long future ahead of them, for the deceptively simple reason that they have a long past. They are highly evolved, hundreds of years old (double-entry bookkeeping, for example, predates the Renaissance). Smart contracts will incorporate many techniques and strategies from control protocols, such as generation of an initial record, segregation of duties, and reconciliation. It will not be long, however, before smart contracts start augmenting and transforming traditional business procedures, making a wide variety of new business structures possible and in the long run replacing traditional controls.

2. George H. Bodnar and William S. Hopwood, 1987.Accounting Information Systems. 3rd ed. Boston: Allyn and Bacon.

(Derived from "Formalizing and Securing Relationships on Public Networks" , by Nick Szabo)

There is also a formal language useful in accounting for derivatives and other sophsticated contracts.

Please send your comments to

Nick Szabo's papers and concise tutorials