Quali vs Quanti: How to use them?
Qualitative (more words-focused) and quantitative (more number-focused) work have traditionally been opposed and the fight over the supremacy of one over the other is a never-ending source of quivers in academic circles. However, this fight only really makes sense if you are focusing on the method you want to use. If you are focusing on the results, you should use both. For instance, marketing services and trading are two industries that, given the stakes, mix both techniques to provide the best answer possible to any question.
Therefore, knowing that you need to use both to have the odds on your side, how should you play with quantitative and qualitative together to get the best answer possible?
As we will see in this two-part post, cross-fertilizing qualitative and quantitative work at every stage is your best bet to come up with a good answer. Especially when the question deals with the future or with uncertainty.
Enrich your words with numbers
First, we will explore how your qualitative work can be enhanced by quantitative or algorithmic work.
1. Detect trends in words on media, social media, blogs, academic research, and your own work
Sentiment analysis has come a long way. You can now feed machines with a large number of qualitative work and have them assign a number on a variety of metrics. Usual ones would be sentiment (from positive to negative) and objective (from emotional to neutral) analysis. This can help us to detect useful trends, such as the treatment of thousands of media pieces on a given topic.
Beyond grading sentiment, a modern algorithm can help detect the topics that are being discussed. This can be useful, for example, to understand what people living in a certain place are tweeting about. In turn, we can understand sections of a population better. Nowcasting technologies are all about that.
Finally, another good use of quantitative enhancement of qualitative pieces is to target the source rather than the content. By measuring the predictive power of your sources, you can weigh them and center your qualitative reading around the most promising material.
2. Know where to start your qualitative research
A base rate helps you to anchor your thinking, especially when dealing with anticipation-related questions. Assume you are trying to forecast how much time it is going to take for a bill to be signed into law. If similar laws were implemented in 2 years on average, you would need a serious argument to predict that this one would only take 2 months. Yet, we are all making similar prediction mistakes in our everyday life and job.
Carrying out a Principal Component Analysis helps you to know where to concentrate your effort and attention for your qualitative work. Let’s take sampling as an example. Say that 3 variables (whether the government and the parliament are of the same party; the popularity of the president at the time the bill is on the floor; the effect of the press coverage) explain 80% of the variability of the time it takes for a bill to become law. Thanks to your PCA, you now know that focusing your qualitative work on those variables will be instrumental in your ability to anticipate.
3. Use numbers as very precise words
Numbers can be divided in ways words cannot. They expand the granularity with which we look at things. Our vocabulary, especially when it comes to probability, is often very vague and poor. Words like “certainly, likely, maybe, unlikely” have different meanings to different people. They also divide the probability space into a few large groups. Compare that to the precision of “72% chance, 24% chance, 27% chance”. Numbers, in that sense, are like a hundred small words that help to describe probability better. An anticipation market yields this effect particularly well.
Numbers also allow for precision thinking. Describing a phenomenon quantitatively refines thinking and exposes mental models about such phenomenon. For instance, describing how the economy of a country is doing is one thing: Employment is coming up, stock markets too. Doing the same with numbers is a different thing: Employment is up by how much? Stocks are up by how much? From when? How does it compare historically?