Super interesting analysis here, and the standout chart is of course the scatterplot showing a pretty obvious positive correlation between revenue beats and growth one year out.
One thing that is hard to tell as visualized is, how dense is that blob of datapoints in the middle? Depending on how many dots overlap, that may account for the vast majority of companies/quarters, and your regression line doesn't fit it so well.
To my eye, there are actually two stories here: Most companies are within 10% +/- of their forecasts, and among those companies, there's not much of a useful correlation, only a rough range of 25% +/- expected growth a year out. However, for companies that fall outside of that range, there's a solid correlation where missing expectations by a lot today suggests poor performance later on, and beating expectations by a lot todays suggests future outperformance as well.
You are correct that if you focus on the blob in the middle, the relationship changes. However, the relationship between performance today and performance over the next year only gets *stronger* if you limit the range (i.e. the regression line gets steeper). However, you are correct that the relationship is *noisier* within that smaller range.
So, stronger relationship, but weaker statistical significance
Super interesting analysis here, and the standout chart is of course the scatterplot showing a pretty obvious positive correlation between revenue beats and growth one year out.
One thing that is hard to tell as visualized is, how dense is that blob of datapoints in the middle? Depending on how many dots overlap, that may account for the vast majority of companies/quarters, and your regression line doesn't fit it so well.
To my eye, there are actually two stories here: Most companies are within 10% +/- of their forecasts, and among those companies, there's not much of a useful correlation, only a rough range of 25% +/- expected growth a year out. However, for companies that fall outside of that range, there's a solid correlation where missing expectations by a lot today suggests poor performance later on, and beating expectations by a lot todays suggests future outperformance as well.
Thanks Aston!
You are correct that if you focus on the blob in the middle, the relationship changes. However, the relationship between performance today and performance over the next year only gets *stronger* if you limit the range (i.e. the regression line gets steeper). However, you are correct that the relationship is *noisier* within that smaller range.
So, stronger relationship, but weaker statistical significance