The Microsoft paper is theoretical
I suspect that this is the epicenter of the claim. I did not find any evidence of Nigerians only pretending to have bad English, nor the central idea of the paper.
The paper is titled "Why do Nigerian Scammers Say They are from Nigeria?" and says "this approach suggests an answer to the question in the title." It is not giving a definitive answer but a hypothesis. No Nigerians were consulted, "language" is only mentioned in passing, and it makes assumptions that scamming is a completely rational enterprise built on values like "[f]ailure to repel all but a tiny fraction of non-viable users will make the scheme unprofitable". A highlight from the paper:
Why so little imagination? Why don’t Nigerian scammers claim to be from Turkey, or Portugal or Switzerland or New Jersey? Stupidity is an unsatisfactory answer: the scam requires skill in manipulation, considerable inventiveness and mastery of a language that is non-native for a majority of Nigerians. It would seem odd that after lying about his gender, stolen millions, corrupt officials, wicked in laws, near-death escapes and secret safety deposit boxes that it would fail to occur to the scammer to lie also about his location.
(Of course, some Nigerian scammers do lie about their location.)
Smith's visits to Nigerian cybercafes
Daniel Smith's 2007 book A Culture of Corruption: Everyday Deception and Popular Discontent in Nigeria details his observations observing and interviewing some low level Nigerian scammers and the people around them. He gives this explanation for why scam emails follow themes like the Nigerian Prince type email:
The fact that millions of 419 e-mails in in-boxes around the world fall into a few basic genres is a product of how the scam enterprises are organized. The people sending the e-mails usually begin with copies of other e-mails that have been used in the past. Their themes are directly related to the structure of real fraud and corruption in Nigeria, to the ways in which the biggest forms of Nigerian corruption occur at the nexus between the state and the global economy, and to the unfolding of current events. Those working directly within the loose organizational hierarchy send most of their e-mails based on the previously successful prototypes. Those working freelance also begin with the prototypes, but are more likely to try stories that could generate a direct transfer of money, since the freelancers obviously don't have the resources to concoct the elaborate scenarios that could produce the biggest payoffs.
The Nigerian Prince scam is actually very old, predating even the internet. Wikipedia cites a NYT article from 1992 mentioning such a successful scam. Smith gives an example of an elaborate heist from 1989 (also successful). The name "Nigerian National Petroleum Corporation" comes up in both stories, though it is no more the real perpetrator than a "Nigerian prince".
Why would someone fall for an obvious scam in the internet age? There may not be an answer that would satisfy the type of person who wouldn't fall for it. More recent data from the Better Business Bureau shows that even five years ago (2018), "Nigerian/Foreign Money Exchange" scams were still going strong, costing Americans at least $700k. (However, it is not clear from that how many of the scams were based on "outlandish" premises or had terrible grammar.)
Some scammers work alone, but others are part of a larger organization. According to Smith, those working closely with an organization are expected to forward all the responses up the chain (but there is not always honor among thieves). Freelancers, on the other hand, try to get the money for themselves first, but if they can't, they pass the information on to organizations in exchange for a small fee:
None of my informants had ever actually received any money directly from a dupe, but they had all received some money from people higher up the hierarchy to whom they passed recipient responses. The two young men who worked directly in the loose organization received fairly regular money. None said they could really support themselves from the effort, though they all repeated stories they had heard about people who had received money directly from a scam victim. The dream of a big payoff seemed to be as much of a motive as the small income they actually earned.
Scamming can be like a lottery: For some the scheme is unprofitable, therefore it's not rational.
Many scammers are not native English speakers and do not have proofreaders
I found high level data for Nigeria on EF and Statista, but the paper The Language of Scam Spams: Linguistic Features of "Nigerian Fraud" E-mails gets to the heart of the issue best. The emails analyzed were from 2002-2003 and found that the vast majority had mistakes that were typical of English learners and "only 20% demonstrated real fluency". Ironically, the majority of emails also tried to employ "learned or specialized vocabulary", such as fancy Latin words for effect. The paper cites something interesting from another paper that I wasn't able to access:
If they are aware of their limitations in English—and Blommaert and Omoniyi suggest otherwise, claiming that at least some of these writers "appear to assume that their English is 'good' enough to pass as native
speakers [sic]" (2006, p. 602; italics original)—I suspect they count on the content of their mailings to prove irresistible to recipients, with greed
winning out over skepticism.
The paper also discusses how $250 was enough to pay for almost an entire year's bills in some areas, while the cost to get started sending spam was trivial due to free email accounts and cheap cybercafes, with the conclusion being that "probably invest minimal time in writing and sending the mailings and have very little to lose and potentially a great deal to gain by casting their nets as widely as possible."
Sometime misspellings have a purpose
In other types of scam emails, misspellings have a clearly defined purpose.
Many email filters will block messages containing spammy keywords, thence l33t spellings and unicode confusables to bypass them. (I am not sure if any "Nigerian" emails try to avoid the filters like that.)
Other attacks may use typosquatting (where scammers may purchase "example.co" to trick users who want to access the established site "example.com"). This is always deliberate, and if there are any other misspellings in other parts of the campaign, it's purely a coincidence. For phishing (and especially spear phishing), the English may actually be quite good.