Wikipedia:Wikipedia Signpost/2023-04-03/Disinformation report



The financial sector of the economy has taken a beating over the last few weeks. Three of the larger US mid-sized banks, Silicon Valley Bank (SVB), Signature Bank and Silvergate Capital collapsed or closed their business in March. They had all been important sources of dollars for cryptocurrency traders. The run on SVB was the "largest bank run in modern U.S. history".

Outside the US, Credit Suisse, one of the thirty most systemically important banks in the world, had to be bailed out by another systemically important Swiss bank, UBS, with the help of the Swiss central bank.

Fortunately, the threat of bank runs seems to be over. The US stock market even went up in March. But the finance sector has one eternal problem – the only product it can sell is trust. Bank depositors need to trust that they can get their money back at a moment's notice. Stock market traders should know that buying stocks is risky, but they need to trust that they will be treated fairly by corporate management and when the time comes to sell the stocks. Without trust, there is no financial sector.

So should we trust banks? Do they lie to us? This article examines this question from a special Wikipedia point of view. On the English-language Wikipedia, do banks follow Wikipedia's rules on paid editing? Or are the articles about banks riddled with falsehoods placed there by sock puppets? I examine the articles about the thirty most systemically important banks in the world, as listed by the Financial Stability Board.

The answers are not uniform for all thirty banks. On average, each of the bank articles examined has been edited by 17.5 socks. The range is very wide however. The article on France's Groupe BPCE was edited by only one blocked sock, the lowest number. The article on America's Goldman Sachs was edited by 55 blocked socks, the highest number. So it appears that at least some of the world's systemically important banks might have been breaking our rules.

Method
This is not an academic paper – simply a quick overview of an important recent question – but the method used should be explained so that the reader can better follow the evidence presented in the following table and the explanations below it. I've used the same method before in articles about the The oligarchs' socks last year and The "largest con in corporate history"? last month. It's just about counting the number of banned or blocked editors who have edited an article and determining how many of those editors were blocked as sock puppets (users who use multiple accounts to deceive) and how many of those socks were blocked or banned for reasons likely to indicate paid editing. In this case I looked for wording in the block summary or in the sockpuppet investigation such as "undeclared paid editing" or UPE, "terms of use" or ToU, "conflict of interest" or COI. I also included those socks who were noted as being blocked by checkusers which indicates a serious case of socking, where the sock can only be unblocked by another checkuser. The first step – finding the editors who have been banned or blocked on each article – is easily automated. The remaining steps sometimes involve the use of judgement.

The bank articles examined are about the thirty most systemically important banks in the world. These banks are all active internationally, are generally large and well known, and are involved in some of the more complex and controversial areas of the banking business, such as derivatives trading. They are selected by the Financial Stability Board because they present the largest risk to the global economy if they were to fail. They should also be the most trustworthy banks in the world, if only because they are the most tightly regulated, both domestically and internationally.

The table below starts with Risk Class in the first column (officially, the Financial Stability Board calls this Buckets). JP Morgan Chase should be viewed as the most important of the 30 systemically important banks listed, should it fail. The fourth column gives the total number of editors blocked or banned who have edited the article. This number includes blocked bots, editors blocked for vandalism, and for many other wiki-sins likely unrelated to paid editing. The fifth column is the most important. It includes sock puppets blocked for paid editing or conflict of interest violations, plus official checkuser blocks. The sixth column includes other socks who may or may not be suspected of paid editing.

Many sock puppets and sock farms have edited several of these articles, suggesting that there might be a network of socks operating in banking sector articles. The seventh column lists several socks who have attracted my attention, mostly for the number of these articles they have edited. But others are also listed, such as, who have become infamous on Wikipedia for their editing or other actions. Similarly I have listed some of the sock farms (large collections of sock puppets apparently working together) in the final column.

Conclusions
The table shows a range of sock puppet editing among these articles on systemically important banks. Some banks such as the French banks Groupe BPCE and Crédit Agricole, Italy's UniCredit, and the China Construction Bank show no sock puppeting by the most likely paid editing socks. On the other hand, three banks, America's Goldman Sachs, the Bank of America, and Germany's Deutsche Bank, have had over 20 socks of this type editing their articles. In general, with some exceptions, American banks have had the most edits by the most likely paid editing socks. Chinese and Japanese banks, along with the above French and Italian banks have the fewest socks of this type editing the articles about them. The two Swiss banks in the news, Credit Suisse and UBS, are in the broad middle ground with only six and nine socks of this type editing the articles about them.

The articles on banks in the highest two risk classes tend to have higher indications of socking than other classes and the articles about the lowest risk class have the lowest indications of socking, with the possible exception of Wells Fargo.

The seventh column shows a selected group of editors who have edited the articles and been blocked for socking. – whose block summary reads "UPE – appeal is only to the Arbitration Committee" has edited two of these articles. who edited several articles on the Adani group, edited at least 23 of these articles. Several other now blocked socks edited multiple articles in this list.

Perhaps the most interesting column in the table is the final one showing the sock farms who edited multiple articles. The sock farm edited at least 11 of these articles. They are known for editing business articles, especially finance articles, as well as political articles. The sock farms edited almost as many.

We remind our readers that no examination based purely on Wikipedia's edit history can prove or disprove whether an editor has been paid to edit articles. Nevertheless, we can say that we have little or no reason to suspect those twelve banks which have fewer than five editors listed in column 5 of paying for Wikipedia editing. Similarly, we might say that the three banks with more than twenty editors listed in column 5 are the most likely among these thirty banks to have paid for Wikipedia editing.

None of this evidence can be taken as final proof of any rules being broken, but there certainly is some interesting evidence.