Well, to be honest, the term "fraudulent" is overly strong. These are also tells for papers that might only be naive, or perhaps just sloppy.
Only bullies pick on the weak. A paper that attacks the "random walk hypothesis" is a paper that picks on a model that no one would defend --- and no one has defended for years. We rejected the normal random walk by normality tests on "day one." A few days later, we found GARCH effects and rejected independence. At any scale where you have 200 observations that can be modeled as stationary you will reject the random walk hypothesis.
So, why would a paper pretend that this is not all wll known?
If a paper written in 2009 is filled with citations to the 1970's or earlier, you have to suspect that the author has not done any review of the current literature, On almost any topic, you can usually find several truly sold review papers in top journals. Picking ups quotes from the 1970's is a silly thing to do.
You can also see some very strange allegations in the literature review. Consider a review that contains a sentence like "...today there are many books teaching how to detect a market trend (Pring, 2003, Murphy 1999)" . Oh, golly, do we only wish that these books really did teach us how to detect a market trend. We're very happy guys with 60-40 bets, so if there was a way to see when we could count on a trend we would be delighted. Stir into the pot that we have read Lo's interview with Murphy, it's hard to think that Muphy knows something that we do not.
Still, even a blind hog picks up the occasional acorn. In a recent article that I regard as completely bogus, I found the reference to Osler (2000) in Economic Policy Review 6(2) pages unknown --- and this could be worth a look. There are a couple of more examples like this.
When the definition of return contains an error --- a logical one, not a typo --- you have to worry.
When notation is overly elaborate or belabored --- you have to worry.
When a paper uses one year of data and the data is sources from Yahoo.com --- you can close the book. Either this is a paper by a student at a place without access to Bloomberg or CRSP or WRD, or you have a crackpot.
Financial data is typically analysed using time series models. Still, there are related model structure that might conceivably be relevant, such as panel data. This is worth thinking about, even if ultimately you decide that it was wrong headed.
Consider the way that the BLS collects much of its data (consumption, employment, use of public resources, etc.). They use a large pannel of consumers, ask them questions over time, aggregate the information, and make their esitmates. They also have to deal with the fact that the pannel itself changes over time --- that is 20% of the pannel is periodically rotated out and a fresh set of consumers at brought into the pannel. This feature of changing the responders over time is what is special about pannel data.
Now, suppose some guy says we should use pannel models to look at financial data. This is original and worth our attention. Could there be something to it?
We'll, if we were looking at a family of mutual funds where individual funds come and go, this could make some sense. But if we are looking at 40 FTSE stocks in 2003, well, the use of a pannel model seems pretty nuts.
If you already owe me an explanation of why pannel models are relevant, and if you then start talking about robust estimation, then I start to think that you are a fraud. This is like talking about "shaken but not strirred" when your pouring from a half-pint of Smirnov's.
An applied statistical paper that does not contain plots --- well that is a candidat for fraud!
A paper that wants to pot shot market efficiency and has no alternative porfolios --- well that is a candidate for fraud!
A paper that talks about "disciples of the random walk hypothesis" --- well, it may not be a fraud, but it certainly does not speak from the understanding that the game has moved on.
A paper that looks at a slew of technical indicators and does not then lead us through the insights that have been obtained --- well, that is surely less informative than one would hope.