My data, pulled from nycopendata.socrata.com, focuses on the rental value of co-ops in Queens from 2008-2011. I wanted to discover which buildings were the best values, highlighting the differences between neighborhoods. To do this, I created a few visualizations that explored the market value for buildings of different sizes, number of units, and date of construction. My initial data suffered from unevenness as certain neighborhoods had significantly more addresses included than others. I tried to combat this by clearly showing the number of distinct addresses included for each neighborhood through scatterplots where appropriate. In all cases, I also filtered the neighborhoods to include only those that had data collected for at least 20 distinct addresses.
One tricky thing about the data included is that in some instances, the same address was checked for multiple years. I therefore had to be careful not to total certain values, which would skew the data. For example, the Total Units could not be summed as data that looks like this:
Would end up like this:
For this reason, Total Units had to remain a dimension and Gross Sq. Ft., Full Market Value, and Market Value per Sq. Ft. had to either be a dimension or, in certain cases, an averaged measure.
With these things in mind, I first looked at how the average market value per sq. ft. changed over time. By using a line graph, the gold standard for time-series analysis, I hoped to see if there were any significant changes between years or differences between neighborhoods.
Overall, values stayed relatively steady, with the largest variance occurring in Bayside between 2010-2011 with a difference of 16.08. Interestingly, while all other neighborhood co-ops decreased in value in 2010, Briarwood alone saw an increase. Briarwood deviated again in 2011, when values declined from the previous year at the same time as the values in other neighborhoods rose. This graph implies that market values in Queens are recovering from the worst of the recession, but as the number of years evaluated are small, this trend may be deceiving. If additional data were available, we would ideally be able to put this graph into a larger context to see if this trend proved true.
As this kind of time analysis proved somewhat disappointing, I moved onto comparing the market value against other variables. My next visualization attempted to see whether the year the co-op was built would affect the market value. Even though year built is another time variable, I didn’t want to use a line graph as I didn’t have data for each and every year. Instead, I opted for a scatterplot to get a truer depiction. This time, I also separated the neighborhoods into small multiples to clearly see the differences in each.
Unfortunately, few clear trends were discerned, despite some dramatic trend lines. Most co-ops do appear to have been built around the midcentury with very few recent constructions and none at all after the year 2000. Unfortunately, whether this is only true for the collected data set or indicative of a wider trend remains unclear. There did not appear to be any true correlation between the year of construction and the average market value in any neighborhood. The vertical concentration of the data points shows that most years contained a spread of values. Likewise, no true correlation was seen between the number of units in a building and the market value per sq.ft. Other factors must be at work in the value of a Queens co-op rental, including general upkeep, whether or not the units have been refurbished, and location of the building within the neighborhood, as only a few.
Finally, I wanted to look at whether the number of units in a co-op rental may result in more or less gross income for the owner. I again turned to a scatterplot, mapping my Average Gross Income per Sq. Ft on the y-axis and the number of Total Units on the x-axis. Again, I made small multiples of the data to see the distinctions between neighborhoods. I also sized the data points according to the Average Gross sq. ft of the given address in order to check for any over or undersized units that might affect profitability.
This result was a little more salient. Interestingly, the graphs varied pretty widely between neighborhoods. The neighborhoods with a greater range in building sizes proved to show stronger and more reliable trends. In both Rego Park and Briarwood, the market value appears to favor smaller buildings, while in Astoria, Elmhurst, and Flushing you would be better off buying big. It would be interesting to look at the same graphs with more addresses mapped and other types of building rentals included such as traditional apartments, condos, and perhaps even houses to see if the correlations stayed true.
Ultimately, my visualizations proved to be a little disappointing and would have benefited from a wider range of data both in terms of the number of co-ops and the years over which they were evaluated. It can be both difficult and deceiving to try to spot broad trends with a small amount of data and I’m afraid that’s where I ended up.
Dynamic versions of these visualizations can be viewed at:
https://public.tableau.com/profile/bgavlin#!/
Data Sources: