What we can learn from Electoral Science
Introduction:
Last week was a whirlwind. So many things had happened, and I did not have much time to put pen to paper (or rather, fingers to keyboard). I had originally wanted to write about the idea of manufactured urgency, but then I turned on the TV and thought of something better to engage with: politics.
It’s election counting day today in 4 of the 5 states that had elections last week. Madhya Pradesh, Rajasthan, Chhattisgarh, and Telangana have their counting done today, while Mizoram will have counting done tomorrow. It is a week when ballot boxes become guns and votes become bullets. By their own description, the states where the elections are going on have become battlefronts with an active war raging on. Turn on any news channel, and you’ll see anchors sitting together and converting this rather important event into something of a sport. At the moment, the live feed is from India Today is on another tab, and the fine art of psephology has been being used to its fullest levels to engage with eager audiences who want to know how the next five years will turn out to be for them. Therefore, I thought, let me write about psychology from a marketing perspective and how we can use that in our day-to-day lives.
What is Psephology?
Psephology, at its core, is about understanding voter behavior. It’s a blend of many things: statistics, sociology, and psychology. Over the years, with the advent of big data and fast computing, it’s become increasingly possible for psephology to improve phenomenally. Psephologists now use sophisticated models that can analyze past voting patterns, demographic changes, and current political trends to forecast election outcomes.
Well, this much we all know. But let’s take some time to go and understand some of the basic ideas that help build up Psephology.
Where it came from:
Greece was home to many ideas, and psephology was one of them. In fact, the word has its roots in the Greek "psephos" - which means pebble. It reflects the ancient Greek practice of casting votes using pebbles or stones when making major decisions. Back in Tamil Naduw here I come from, we used to have a system called “Kuda olai”. Kudam meaning pot, and olai meaning written palm leaf. Coming to think of it, kudaolology sounds rather impressive too.
However, the modern avatar of Psephology is rather new. It was only in the mid 20th century that it began to take shape in the form that we currently understand. Statistics, opinion pooling, rigorous data collections and the whole shabang. It is understandable that this largely coincides with the rise of democracy as the primary form of governance across the globe.
Scholars such as R.B. McCallum, Alison Readman, and many others laid the foundation for the systematic study of electoral behavior, using statistical methods and historical data. They were able to predict voting patterns based on trends, and sample information that was collected on the ground.
The foundational principles
Now. This all sounds simple at the onset. But there is a lot that we are still learning. Remember when I say there is politics in everything? It so happens this is also the case in academics. Let’s talk a little bit about the idea of statistics here.
Remember the statistics courses that you took in your first year? That is basically not statistics. To be more specific, that’s just part of statistics. It’s called ‘frequentist’ statistics. There exists another significant part, called Bayesian statistics that we do not really discuss in regular college classes. They are taught in machine learning and AI courses (Bayes pretty much invented these fields with his theorem).
Unlike frequentist statistics, which analyze data from repeated experiments to make inferences, Bayesian statistics revolve around the use of probability to represent uncertainty about the world. This approach is particularly relevant in ever-changing landscapes, such as politics.
Bayesian methods begin with a 'prior'—an initial assumption based on past knowledge or data. For instance, a psephologist might start with a hypothesis about a political party's chances based on historical election performance. As new data comes in, like recent poll results or demographic shifts, the Bayesian approach updates this hypothesis, refining the predictions. This 'posterior' probability becomes a more informed, nuanced view of the potential election outcomes.
The beauty of Bayesian statistics in psephology lies in its flexibility and adaptability. It acknowledges that our understanding of voter behavior is not static; it evolves as more information becomes available (which is what rapid technological changes have made possible). Especially in close races where every new piece of data can sway the prediction, Bayesian models excel by incorporating this data in real-time, offering a dynamic, constantly updating picture of the electoral landscape.
Let’s take this one step further. Bayesian statistics allow for a more intuitive understanding of uncertainty. In the context of elections, where numerous variables can influence outcomes, this approach provides a probabilistic assessment of different scenarios. For instance, rather than just predicting a win or lose, it can offer probabilities of different margins of victory, considering various influencing factors. This is definitely more useful when the decisions you take are not just based on victory or loss, but on the quantum or magnitude of the victory and where you are coming from.
Taking it one step forward
Some of you may be familiar with the constant “approval ratings” that US presidents get from time to time. Coupled with Bayesian methods, this can become a powerful tool to steer how policy is crafted. Politicians understand votes. Remember the adage, “what gets measured gets done”. Taking the logic, if votes can be measured in real time (as is the case with Psephology) you could “get done” policy interventions. This is not just once every election cycle, but on a constant basis. In other words, these tools help in firmly ensuring that the flag of democracy flies high.
From Elections to marketing
I would be lying if I told you things like this are not becoming a reality in marketing too. Bayesian methods are used to understand customer behavior more deeply. By analyzing customer data, companies can predict how likely a customer is to make a purchase, their preferred products, and even when they might buy next. Large firms like amazon have the kind of data to make this possible. This is precisely what people are talking about when they say using “advanced analytics” to drive marketing. This is the “data-based decision making” that managers keep talking about.
Bayesian models are also used to predict market demand for products or services these days. Marketers can integrate various factors like historical sales data, economic indicators, seasonal trends, and promotional activities to forecast future demand, helping in inventory management and pricing strategies.
One place in which the methods are used rampantly is in RTB - or real-time bidding in the context of ads. Firms today use Bayesian algorithms to optimize their resource allocation on content personalization, optimizing marketing spend across channels and purchasing ads.
Bayesian statistics provide marketers with a powerful toolkit for understanding and predicting customer behavior, optimizing marketing efforts, and making data-driven decisions in an environment of uncertainty.
Closing remarks
As we draw to a close, it's clear that the principles of psephology and the nuanced applications of Bayesian statistics extend far beyond the realm of politics and into the very fabric of our daily lives, especially in the dynamic world of marketing. The transition from analyzing voter behavior to understanding consumer preferences underlines a broader theme: the power of data and predictive analytics in decision-making processes.
Right now, it looks like there we’re getting some definitive pointers from the data on TV. I’m signing off now. Until next week!