Note: The author of the paper is Hans Byström, professor of economics at Lund University in Sweden, originally published in the Ledger 2019 Vol 4 journal.
The following is the translation:
Abstract: In this paper, I will discuss how blockchain can potentially affect the way credit risk is modeled and how real-time accounting using blockchain techniques can improve default predictions.
To demonstrate that this change has a (substantial) impact on well-known credit risk metrics, we use a simple case to compare the studies.
Most financial people have heard of bitcoin, a virtual currency. However, few people have heard of the technology behind the bitcoin – blockchain. Although blockchain technology has so far been used primarily as a conduit for bitcoin, blockchains can also be used for infrastructure for traditional financial products such as debt contracts and financial derivatives. (1)
In accounting, the blockchain can potentially improve the quality of information obtained by investors in two ways: one is to make accounting information more credible, and the other is to make information more timely. For trust, if a company keeps its financial records on the blockchain, opportunities such as accounting-related manipulation and fraud may be significantly reduced, and intercompany transactions will become more transparent. (2) For time, real-time updates of accounting information will be possible because the blockchain-based books will make every transaction in the company's books available immediately. (3) In addition, this information is not only available immediately to the company's internal staff, but also to external (selected) outsiders such as regulators. (4)(5)(6)
In this paper, my focus is on credit risk modeling and how the widespread use of future blockchains may affect how we model credit risk. As we all know, accounting information, such as balance sheets, income statements, etc. are not perfect. (7) There are some problems in accounting data, such as unclear and non-uniform accounting practices, managers engaged in creative accounting or reporting lags behind real events. Since most credit risk models rely on accounting data, companies can keep their books on the blockchain (whether public or private) with transparency, accuracy and timeliness. Significantly improve the credit risk model. (8)
Of course, no one knows if the above scenario will be achieved or when it will be achieved. In the analysis of this paper, I usually assume that (I) listed companies upload their financial data to a public blockchain, (ii) data uploads will be very frequent, possibly daily, and (iii) required for credit risk modeling Any accounting data is indeed uploaded to the blockchain. In other words, the focus of this article is not on whether the company will participate in future blockchain initiatives.
Second, the blockchain
In 2008, the author of the pen name Satoshi Nakamoto (Zhong Bencong) first published a white paper on Bitcoin, a cash-like but lacking central bank-backed digital currency that provides a peer-to-peer exchange of ownership, important The bitcoin does not depend on a central clearing house like a bank. Instead, each historical bitcoin transaction is stored in a globally distributed electronic ledger, which we call the blockchain, which records all transactions in the history of Bitcoin. (9)(10)(11)
Bitcoin books are called blockchains because new bitcoin transactions (or blocks) are added to the historical trading chain, which are added by special Bitcoin users (called miners). . The miners verify that every bitcoin transaction in the block is legal by solving a difficult cryptographic problem. (12) This innovative technology adds new transactions to the (bitcoin) blockchain and is verified by the entire decentralized network, which greatly reduces transaction costs. In addition, the Bitcoin blockchain is completely transparent and protected by sophisticated encryption techniques (using mathematical algorithms called hash functions) and the work of miners.
Third, blockchain and real-time accounting
A blockchain is basically a ledger that cannot change and destroy records. Therefore, it can be used as a trustworthy, constantly updated company accounting record book. (13) This is because blockchain technology can be used not only to transfer digital currency between buyers and sellers, but also to transfer ownership of any other asset between the two companies cheaply, efficiently and reliably. (14)
The financial statements are prepared on a regular basis and summarize what happened to a company's books over a certain period of time. The auditor then comments on the accuracy of the financial statements.
Outsiders such as investors and credit risk managers must believe that the audit work is thorough and fair, and that the company does not provide false information to the auditor. In other words, the concept of trust is crucial in the preparation of financial statements and audits. This is where the blockchain technology behind Bitcoin comes into play. (13)
If a company voluntarily (possibly due to market pressure) publishes all of its business transactions on the blockchain, each transaction has a permanent timestamp, then the entire book of the company will be immediately visible and anyone can own the company The transactions are aggregated in real time to the income statement and balance sheet. (2) That is to say, the auditor has done a lot of things in the accounting profession today, and the blockchain may be more efficient and timely tomorrow. By construction, if a company keeps all its transactions and balances in the blockchain, the blockchain itself can largely replace the auditor to confirm the accuracy of the company's accounting (avoiding potential moral and agency risks) ). Since past transactions in the blockchain cannot be tampered with, the issue of distrust is naturally removed from the company's financial statements.
In addition to the trust issue, the account book is automatically updated in real time, and each transaction (more or less) is immediately included in the company's blockchain, which may make the company's accounting information as timely and dynamic as the stock price. That is to say, due to the natural parallelism between blockchain and accounting, blockchain technology can improve the quality of investors' access to accounting information in two ways: one is to make information more credible, and the other is to make information more timely.
Blockchain and credit risk model
Currently, the two most famous credit risk models are the Atman Z-score model and the Morton (1974) model. (15) (16) The Z-score formula for predicting bankruptcy was proposed by Edward Altman in the late 1960s, using various corporate income and balance sheet variables (ie accounting information). Plus stock prices to predict whether the company will go bankrupt.
Z-score is a linear combination of five financial ratios, calculated as:
X1 = working capital / total assets; X2 = retained earnings / total assets; X3 = earnings before interest and taxes / total assets; X4 = equity market value / total book value of liabilities; X5 = sales / total assets;
The larger the Z-score value, the less likely the company is to default.
The Merton model, which also relies on accounting information and stock prices as input, but treats the company's equity and debt as contingent claims issued against the company's underlying assets. (17) In the Merton model
N（ ） is a cumulative normal distribution, and
VE is the market value of the company's stock; VA is the market value of the company's assets; D is the total amount of the company's liabilities; Tt is the maturity of the company's liabilities; rf is the risk-free rate;
In addition, stock volatility And asset volatility Associated by the following equation:
We can solve VA and Nonlinear equations (1) and (2). The default distance is defined as:
The larger the DD value, the less likely the company is to default.
4.1 Case Study
To demonstrate the effect of quarterly updates to near-instant updates from accounting information, I studied the two credit risk models described above and applied them to two well-known US companies: Apple and Groupon. Because the accounting information for these companies is sampled quarterly (that is, risk metrics cannot be updated multiple times every three months), I must simulate the daily movement of Z-score and DD metrics. (18) These daily changes are generated by sampling normal distribution random numbers.
In this way, I got a reasonable implementation of the possible blockchain real-time Z-score and DD in the future.
Figure 1 shows the Z-score and Merton Default Distance (DD) metrics for the two companies, with daily and quarterly accounting data. (19) (20) The volatility of the assumed daily risk change is selected based on the company's actual quarterly risk and the volatility of the debt change, so the (substantial) fluctuations in the two figures provide a reasonable The reality proves that the introduction of blockchains in corporate books, how the estimated risk measures will change and how much.
The volatility during the quarter was not significant, indicating that the credit risk model was improved when accounting information was changed from quarterly updates to daily updates. (twenty one)
Figure 1 Apple and Groupon's Z-score and Merton DD (default distance), including daily and quarterly accounting data from October 2014 to October 2015.
As mentioned earlier, the dynamics of Z-score and default distance (DD) in Figure 1 indicate that credit risk metrics can be significantly improved if real-time accounting based on blockchain is available. For example, the average change in Z-score and DD (for both companies during this time period) from one quarter to the next is 13% and 36%, respectively. In other words, even if we assume a little simpler, the actual Z-score and DD change linearly from quarter to quarter, and on any day between quarterly updates, the credit risk modeling error is on average compared to the actual credit risk level. 6.5% and 18%. Furthermore, if the stochastic process in this paper is used to model the intra-quarter changes in risk, the modeling error may be much larger than these numbers, as shown in Figure 1. An extreme example is the third quarter distance between Groupon and the default value, where the modeling error is much larger. Furthermore, in the case of the Merton model or Z-score, the error is even larger when considering the probability of default (PD) rather than the default distance. (22) As for the timing of risk assessment, regardless of the credit risk modeling method, the blockchain-based risk measurement will reach the credit risk level of the next quarter earlier, with Apple 75% and Groupon 67%.
On average, through real-time accounting, the risk level for the next quarter can be reached in about two months (that is, one month in advance), and in some cases, in a few weeks.
Figure 1 shows the risk dynamics of the Z-score and Merton DD metrics, but considering the more dynamic (daily) implementation of the Merton model, Z-score may be the most measurable measure of real-time accounting in practice. . It is even possible (to be quite outdated) that the Z-score method will undergo a renaissance due to the introduction of blockchains.
In fact, with the significant changes in accounting and auditing practices described above, current Z-scores may be replaced by new score methods that include other financial ratios or coefficients.
The entire area of bankruptcy forecasting is also likely to change, with a focus on new tools (smart contracts in blockchain terminology) or financial ratios that are directly adjusted for the likelihood of default. As outsiders have access to all of the company's transactions, the bankruptcy process can also undergo fundamental changes, and managers, creditors, investors, and regulators are all following new rules. Problems such as reflexivity may be greater than today's impact.
Finally, it should be emphasized that even if the company's business transactions published on the blockchain are limited, the credit risk model may still be affected in some way.
Above, I have discussed how the blockchain technology behind Bitcoin can improve credit risk modeling by improving trust and better timing of accounting data release.
If my proposal can be realized in the next few years, its impact on our approach to credit risk modeling can be enormous. Through a simple case study, I found that blockchain can also have a major impact on today's widely used credit risk metrics.
Acknowledgements: Thanks to the Marianne and Marcus Wallenberg Foundation and the financial support provided by Handelsbankens Forskningsstiftelser. Part of the paper was written when the author visited ESADE in Barcelona. The author would like to thank the five anonymous presenters.
Notes and references
1. A recent example was the decision by the Australian Stock Exchange to become the world's first market to use blockchain-settled stock trading in February 2016, see Financial Times (January 25, 2016) https:// Www.ft.com/content/fba2346-c370-11e5-b3b1-7b2481276e45.
2. Yermack, D. “Corporate Governance and Blockchain” http://www.nber.org/papers/w21802.pdf.
3. Although real-time accounting traditionally means that the company's books are updated monthly or quarterly, in the future blockchain world envisioned in this article, the term actually means near-instant (daily) updates of accounting information. ;
4. In this article, when referring to blockchains, I usually refer to the public blockchain, not the private blockchain. Of course, although many companies may not voluntarily disclose all inspection data on public accounts, one can think of a situation where the most important figures (such as sales, leverage, etc.) are public and the rest are confidential. Or distributed to a small number of selected participants through a private blockchain. In other words, even if the future blockchain environment will consist of many private (licensed) networks that are hidden from the public eye, some censored entities (such as regulators and credit rating agencies) can be updated at any time through distributed ledgers. Balance sheet information.
The release of all or part of the balance sheet information may be driven by regulatory or market forces. Another possible approach is to use a so-called side chain, ie the company uses a private blockchain that is periodically (partially) connected to the main (public) blockchain.
5. Privacy issues are important to most companies, and the power to minimize the number of participants in real-time financial statements may be there forever. For example, one can think of a situation where only shareholders holding a certain number of shares can access the books (ie, the books are not completely public);
6. As for the scalability of the blockchain ledger, this article does not discuss in detail;
7. Duffie, D., Lando. D, “The structure of credit spreads with incomplete accounting information” https://doi.org/10.1111/1468-0262.00208.
8. Of course, the operational risks associated with blockchain management should not be ignored. There is no doubt that a fully publicized book will have errors and implications, but this (potential) risk will not be discussed further in this article.
9. Nakamoto Satoshi, "Bitcoin: A Peer-to-Peer Cash System (2008)" https://bitcoin.org/bitcoin.pdf.
10. Antonopoulos, A. M “Proficient in Bitcoin” (2014);
11. Swan, M. “Blueprint for Blockchain in the New Economy” (2015);
12. The fastest-moving miners receive some bitcoin rewards for this service. A new block will be created every 10 minutes, and miners who lose in the competition will not get anything. Bitcoin mining is sometimes called For "competition accounting." See Harvey, CR "Encryption Finance" https://papers.ssrn.com/sol3/Papers.cfm?abstract _id=2438299.
13. Lazanis, R. “How does the technology behind Bitcoin change the accounting we know” https://techvibes.com/2015/01/22/how-technology-behind-bitcoin-could-transform-accounting-as- We-know-it-2015-01-22;
14. Along the way, some trials are underway. For example, NASDAQ is experimenting with the use of “stained coin” technology as a way to record stock trading using blockchains, see: Hern, A. “Nasdaq bets on the bitcoin blockchain to become the future of finance” https://www.the Guardian.com/technology/2015/May/13/Nasdaq-Bitcoin-Blockchain.
15 Altman, E. “Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy.” Journal of Finance 23.4 589-609 (1968) https://doi.org/10.1111/j.1540-6261.1968.tb00843.x. 16 Merton, R. “On the Pricing of Corporate Debt: The Risk Structure of Interest Rates.” Journal of Finance 29.2 449–470 (1974) https://doi.org/10.1111/j.1540-6261.1974 .tb03058.x.
17. By excluding asset values and volatility from stock price and balance sheet information, the model produces an estimate of the company's default probability. The Merton model uses the Black and Scholes framework to solve the asset value and volatility implied by stock prices and volatility. Asset value and asset volatility can then be combined into a risk metric called the default distance (DD), which is inversely proportional to the company's default probability, see Black, F., Scholes, M. “Price pricing for options and corporate liabilities. Https://doi.org/10.1086/260062.
18. The necessary accounting variables are total assets, total liabilities, working capital, retained earnings, EBITDA, Z-score sales, and the total liabilities of the Merton model. The data was downloaded from Yahoo Finance.
19. In order to separate the dynamic impact of real-time accounting on the two credit risk measures, I conduct a sample analysis of the stock price every quarter. Although this may be typical in Z-score applications, it is more common when using the Merton model. It is to update the stock price every day.
20. In the Merton model, the stock return volatility is calculated based on the daily data of the previous quarter, and the risk-free rate is set to 10 bp;
21. Through improvements, risk metrics are more up-to-date or timely due to the use of more updated input information.
22. The reason is the highly nonlinear relationship between PD and DD. The standard Merton model is given by the mathematical properties of the normal distribution, while the Moody KMV model is given by the internal database.