Finance & Accounting Research Questions

"How do [robo-advisors] in Germany perform in terms of their (risk-adjusted) performance compared to benchmarks between January 2017 and February 2022?" names the market, the measure, the benchmark and both ends of the window. Once a question is that specific the methods chapter is mostly bookkeeping, because every choice a reader might query has been declared in advance. Set it beside something like the impact of ESG on firm performance, which names nothing you could pull out of a database, and you can see which one already has a data chapter behind it.

Whatever a practitioner would treat as the alternative, named in the question. For returns that is usually an index or a passive portfolio, for a collection or a cost it is the budgeted or projected figure, and for a firm it is often the sector average. Choose it before anything else, because it decides which data you need and it is one of the first things a reader will challenge. A performance question with no benchmark in it can report a number but cannot judge it.

Rarely, and that is where this field gets its precision. ESG pillars, LIFO and FIFO, a securities regulation, a set of governance guidelines with a year in its name: somebody defined the variable before you arrived, so your job is to say which definition you are using and cite it. Where two definitions compete, pick one, say why, and stay with it. Building a measure of your own is a thesis in itself, and usually not the one you meant to write.

Break it into its parts where you can. Asking what each ESG pillar contributes to a firm's default probability, rather than what the headline score does, is what stops a finding being waved away as an artifact of how the index was built. Providers also disagree more than people assume: Berg, Kölbel and Rigobon, writing in the Review of Finance in 2022, found ratings of the same firms correlating between 0.38 and 0.71 across six agencies, most of that gap coming from how each one measures. So say whose score you used and how it is built.

It works the same way with two of the coordinates swapped. Where a finance question names a market and a window, an accounting question names a jurisdiction and a standard, because that is what fixes the rules your data was produced under. Everything else holds: name the population of firms, say which years, say what the comparison is. If your topic sits under a standard that changed mid-period, that change is usually the most interesting thing in the study.

It can, and describing one well is often the first question of a set. Naming three quantities, revenue capacity, projected revenue and actual collections, then asking what the gaps between them are with no attempt to explain them, leaves room for a second question about what causes the gap, answered with interviews or records. Trying to do both in one sentence usually produces a question neither half of your method can answer.

Not for most finance questions worth writing. They run on filings, regulator returns, published governance guidelines, local government revenue records and providers' own published portfolios. Terminals like Bloomberg make some designs possible and are worth asking your library about, but the honest move is to pick a question your actual access can answer rather than writing one around data you hope to get.

Pick the window for a reason and state the reason. A regulation coming into force, a crisis, or the point where a series starts are all defensible; a round number of years is not. A window running from January 2017 to February 2022 reads as arbitrary until you learn it covers the whole life of the products being compared, and that sentence is exactly what belongs in your methods chapter.

Yes. Lenders on how they handle asymmetric information, financial professionals on what generative AI does to their role, investors on whether governance compliance changes a lending decision: none of that sits in a database. If your program expects an econometric model, an interview study can still ride alongside as the second question, explaining what the numbers show.

No. Most finance and accounting questions are bounded to one market and often one sector: insurance claims in Kenya, revenue collection in Tanga, robo-advisors in Germany. That scope is what makes the data reachable and the claim checkable. What you owe the reader is a sentence on why the setting is worth studying, and honesty in the discussion about where the finding may not travel.

Say impact only if your design can carry a causal claim, which usually means a shock, a policy change or a genuine natural experiment. With cross-sectional or panel data and no identification strategy, relationship or association is the accurate word and costs you nothing. Examiners in this field tend to read the verb closely, and an overclaimed impact is one of the easiest things in a finance thesis to attack.