Exit, Tweets and Loyalty

That is the title of a new paper by Avi Goldfarb, Mara Lederman and myself.

In 1970, Albert Hirschman wrote a widely read book, Exit, Voice and Loyalty, whereby he noted that economists relied solely on a particular mechanism — exit — to think about how organisations that aren’t performing well are disciplined. Don’t like what you are buying, leave. He noted that political scientists seemed to be more predisposed to an alternative mechanism — voice –to discipline organisations. Don’t like what you are buying, complain. But while those notions were talked about following Hirschman, suffice it to say, economics at least has gone on regardless, continuing to pay scant attention to voice even if they themselves seem quite disposed to complaining pretty much constantly (well, amongst my circles anyway).

One of the constraints has been that it has been hard to measure voice. Exit can kind of be captured by market share or consumer churn data. But voice tends to be more obscure. To be sure, some people did study consumer boycotts but compared with the amount of voice actually being exercised, this was surely the tip of the iceberg. Fortunately for us, big data has come along — and by “big” I mean more data than can be held in a Excel spreadsheet.

Our big data comes from Twitter. We noted that people seemed to use Twitter to complain about airlines. This is certainly a big activity amongst economists but as it turned out if you looked in a year you could find millions of tweets directed at airlines in the US. We looked and wondered if this could be a measure of voice.

First of all, we wanted to make sure it wasn’t just noise. We could look at tweets and how they were phrased but our big task was to match the tweets with location and time data from airlines to see what particular airline markets the tweets were likely about. That was challenging but thanks to the data being big, we were able to do that. Moreover, we found that tweet volume was associated with data on airline delays so it seemed that it wasn’t all hot air.

Second, we were interested in how tweet volume related to competition. Hirschman had thought that voice was more likely to be exercised when consumers couldn’t really exit but then had wondered if that made sense because why would a firm who knew they had locked in consumers care about voice. We took a different approach and noted that as markets became less competitive, consumers could still exit but more importantly, if they did exit, that was a bigger loss for a firm precisely because they could earn more in the face of less competition. In other words, if a firm who had market power was actually going to care more about their relationship with their consumer and so would like to talk before separation. Consumers, knowing this, would exercise voice more as a result.

Our paper determined that was in fact the case. Controlling for delays, the tweet volume (including negative tweets) went up as concentration in the market rose. Moreover, we found evidence that airline responsiveness (i.e., replies to complaints) tracked this as well. Finally, we also determined that this was stronger when tweets gave us indications of a long-term relationship (i.e., “a platinum fantasic member I …”).

So we think Hirschman is back and is back because digitisation now provides us with a window into complaints. US airlines are one matter and we don’t know if the exercise of voice actually helps in terms of performance. But we think that voice might finally get the attention it deserves from economists.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s