Extreme values (outliers) have been removed from the data. Examples are homes that had $0 as the sale price, or homes that increased hundreds of percentage points in price in 1 day. Such data points have a strong effect on the data when taking averages and removing them has the effect of making the graphs smoother. However, you can still see spikes in the graphs and these may be more pronounced when the data set is reduced, such as if you select certain neighborhoods and building classes. It's difficult to remove all of the extreme values, so I would suggest focusing on general trends as opposed to any month to month fluctuations.
Average returns are found by using properties that have at least 2 sales recorded (the time period between 2 sales would be one time span), finding the annualized rate of return, applying that rate of return to each month-year in the respective time span and then averaging across the properties to the selected aggregation (Borough, Neighborhoods and Building Classes).
You may notice extreme values at the endpoints of the graphs (the beginning or the end). This is a result of a small number of sales contributing to these points, so take those with a grain of salt.
At [email protected] removing the name of any flightless birds