大數據挖掘是投資者的好幫手
Fraud is a word to strike fear into the hearts of any investor, who tend to take a company’s financial figures at face value. But time after time they find themselves burnt when overly aggressive or even fraudulent accounting leads to calamity.
欺詐是一個所有投資者都害怕的詞——他們往往完全相信一個公司的財務數據。但一次又一次，他們發現自己在過於激進乃至欺詐性會計手法導致災難的時候蒙受了損失。
Enron is the classic case of a seemingly rock-solid corporate titan that was in reality a fragile edifice of dodgy numbers and accounting subterfuge. More recently, Valeant, the Canadian pharma group, has lost nearly $80bn of its value over accounting concerns.
安然(Enron)就是一個典型案例：一個表面看上去堅如磐石的龐大企業，實際上是用虛假數據和假賬堆砌起來的脆弱大廈。最近，加拿大製藥集團Valeant因外界對其會計手法的擔憂而損失了近800億美元的市值。
The company recently said that its internal accounting review had found nothing that would force it to restate its earnings, helping its shares regain their footing, but many big-name investors are still nursing huge losses.
Valeant最近表示，經過內部會計審覈，該公司不必重報利潤數據，這幫助其股價重新站穩了腳跟，但許多知名投資者仍然蒙受巨大損失。
Can mining reams of data about companies help investors detect issues early? Deutsche Bank’s financial scientists believe so, and have developed a model that scans for potential problems.
挖掘公司的海量數據可以幫助投資者及早發現問題嗎？德意志銀行(Deutsche Bank)的金融科學家們相信可以，並開發出了一個查找潛在問題的模型。
It mines the Securities and Exchange Commission’s database of companies censured for accounting problems — highlighting how banks, trading firms and regulators are increasingly turning to novel technological solutions to uncover market abuses.
這個模型挖掘美國證交會(SEC)的數據庫，該數據庫包含被指責存在會計問題的公司的數據。德意志銀行的這一舉動突顯出，銀行、交易公司和監管機構都越來越多地求助於新技術解決方案來發現市場濫用行爲。
“Accounting numbers are like volcanoes. When they lie dormant, people forget how dangerous they can be,” Deutsche Bank said in a recent note.
德意志銀行在最近發佈的一份報告中表示：“會計數據就像是火山。當它們休眠的時候，人們忘記了它們可以多麼危險。”
The German bank’s model used “Benford’s Law” to identify possible irregularities. In 1938, physicist Frank Benford noticed that in a random selection of numbers digit 1 tends to appear more often at the beginning of a number than 2, and 2 more often than 3.
德意志銀行的模型使用了“本福德定律”來查找可能的不合規行爲。1938年，物理學家弗蘭克•本福德(Frank Benford)注意到，隨機選擇一些數字時，首位數爲1的數字往往比首位數爲2的數字更常見，首位數爲2的數字比首位數爲3的數字更常見。
This curious law is used to analyse everything from weather patterns to election fraud.
這種奇特法則被用來分析一切事物——從天氣模式到選舉欺詐。
“The natural extension of this hypothesis,” Deutsche Bank’s Javed Jussa wrote, “is that companies that do not conform to Benford’s law may exhibit some sort of accounting irregularity.”
德意志銀行的賈韋德•尤薩(Javed Jussa)寫道：“從這種假設可以自然地推導出，不符合本福德法則的公司可能從事了某種會計不合規行爲。”
Deutsche Bank’s quantitative analysts are not the only ones looking to harness modern technology and data mining to unearth potential problems.
並非只有德意志銀行的定量分析師在尋求利用現代技術和數據挖掘來揭示潛在問題。
Regulators are also looking to capitalise on recent advances in computing and “machine-learning” algorithms to autonomously scan financial markets and company reports for signs of fraud or abuse.
監管機構也在尋求利用計算和“機器學習”算法上最近的進展來自動掃描金融市場和公司財報，以尋找欺詐或市場濫用行爲的蛛絲馬跡。
This is the future of fraud detection, says Steven Blum, a managing director at Control Risks’s compliance and forensic accounting division. “It’s a tool, but an increasingly powerful tool. And the more data you get into the mix the more powerful it becomes.”
化險諮詢(Control Risks)合規和法務會計部門的董事總經理史蒂文•布盧姆(Steven Blum)表示，這是欺詐偵查的未來。“它是一個工具，但這種工具越來越強大。你把越多數據輸入進去，它就越強大。”
For example, a few years ago the SEC rolled out what has been dubbed its “Robocop”, a computer program more formally called the “Accounting Quality Model”.
例如，美國證交會幾年前推出了一個綽號爲“機械戰警”(Robocop)的計算機程序，它的正式名稱是“會計質量模型”。
This uses the regulator’s treasure trove of financial data to detect patterns in corporate earnings reports that might hint at strange behaviour — whether just aggressive accounting practices or outright deceit.
該程序使用證交會的金融數據庫來發現企業利潤報告中的模式，從中發現可能隱含的古怪行爲——無論是激進的會計手法還是赤裸裸的欺詐。
The details on the SEC’s Robocop are sparse, but Mr Jussa found several important indicators that could hint at potential accounting mishaps.
關於美國證交會“機械戰警”的具體情況，人們知之甚少，但尤薩發現了多個可能暗示着潛在會計問題的重要指標。
Some are technical market factors. Sizeable bets against a company — as measured by the “short interest” — are a classic sign that some hedge fund managers smell something fishy. Unusually high trading volumes and prices far away from their one-year lows are also warning signs.
一些指標是技術性的市場因素。對一家公司的大規模做空（以“賣空數量”衡量）是一些對衝基金公司發現疑點的典型跡象。異常高的交易量和遠高於一年低點的價格也是值得警惕的跡象。
But there are more fundamental warning signs, such as the reported cash flow that a company’s assets throw off, the expected dividend yield and the turnover of the company’s assets.
但還有更爲基本的預警信號，比如公司財報中資產產生的現金流、預計股息收益以及資產週轉率。
The most common problem is the misstatement of revenues, for example aggressively booking in orders as actual sales — something some tech companies have a tendency to do — or recording sales in one period when the goods are actually shipped in the next quarter. Another is accounting for some costs as assets.
最普遍的問題是虛報收入，比如將訂單記爲實際銷售的激進做法——某些科技公司傾向於這麼做——或者將下一季度才真正發貨的商品記入本期銷售。另一個問題是將一些費用記爲資產。
Early identification of potential problems can help investors avoid some dud bets. Deutsche Bank ran its accounting fraud model against 30 years of financial data, and found that a portfolio of companies it found as “potentially vulnerable to accounting irregularities” underperformed the broader stock market by a third.
及早發現潛在問題可以幫助投資者避免一些無用押注。德意志銀行用其會計欺詐模型處理了30年的金融數據，結果發現，模型認定“可能存在會計不合規風險”的那些公司的股價表現比大盤低三分之一。
Mr Blum points out that “big data” and evermore sophisticated algorithms will become increasingly common among a variety of organisations in the finance industry.
布盧姆指出，“大數據”以及越來越先進的算法將在金融行業中的衆多機構中變得越來越普遍。
Even some repentant sinners are now using technology to detect market abuses. After being censured and fined for “spoofing” — submitting a blizzard of orders and cancellations only to move the price of a security — in 2010, trading outfit Trillium developed a system to spot and prevent spoofing internally, which it is now marketing to other firms.
就連一些犯過錯但痛改前非的機構，現在也在使用技術來發現市場濫用行爲。在2010年因“幌騙”（spoofing，一種市場操縱手法，先提交大量訂單然後取消，目的只是爲了改變證券價格）而遭受譴責和罰款之後，Trillium開發了一套系統，在內部發現並阻止“幌騙”。現在Trillium正向其他公司推銷這套系統。
“The controls we built have turned out to be a next-generation solution to spoofing,” says Michael Friedman, Trillium’s general counsel.
Trillium的總法律顧問邁克爾•弗裏德曼(Michael Friedman)表示：“事實證明，我們建立的控制措施是杜絕‘幌騙’的下一代方案。”
The “Surveyor” program autonomously scans for the market order book for potential manipulation, such as spoofing, insider trading, two-sided trading — where a company pushes a stock around by trading with two separate accounts, either intentionally or by accident — and “wash trades”, the simultaneous selling and buying of a stock through different brokers.
“監督員”(Surveyor)程序自動掃描市場訂單簿，以尋找潛在的操縱行爲，比如“幌騙”、內幕交易、兩面交易（指一家公司通過兩個獨立賬戶交易來操縱股票，無論是有意還是無意）以及“虛買虛賣”（wash trades，指通過不同經紀商同時買進和賣出一支股票）。
In other words, technological advances and increasingly complex markets may make an investor or regulator’s job harder and more complex, but may also offer some potentially powerful solutions to keeping financial markets clean.
換言之，技術進步和日益複雜的市場可能讓投資者或者監管機構的工作變得更加艱難和複雜，但可能也提供了一些潛力巨大的解決方案來保持金融市場的純淨。
