1

Just analyzed the image names of automobiles in an auto auction sale archive and filtered the image names by the automobile's model year and counted the years' frequencies. I came up with the following plot, which I found interesting and have decided to post here for your comments on what the reason for the spike is:

SPIKE

x-axis: time | y-axis:frequency of vehicles with that model year

Ubra
  • 11
  • 2
  • Welcome to our site! The socio-economic aspects of this question are most likely off-topic here - see our [help/on-topic] - but I think the discussion of why the bell-shape occurs in such data in general is pertinent here. – Silverfish Apr 14 '16 at 00:20
  • 5
    On a terminological note, many people use the phrase "bell shaped" as synonymous with the Gaussian or "normal" distribution, but (as seems to be the case in your example) the fact that a distribution is unimodal (single peak) with a smooth "hump" [doesn't necessarily make it normally distributed](http://stats.stackexchange.com/q/129417/22228) – Silverfish Apr 14 '16 at 00:25

2 Answers2

1

People can't sell a car manufactured in the future, so that accounts for the right-hand tail. Moreover, you're unlikely to sell a nearly-new car as an individual, so that accounts for the drop off. And you're unlikely to sell a car that is old/decrepit/at the end of its life due to maintenance costs and the general observation that a very old car is generally only useful for scrap. That accounts for everything left of the peak.

I don't know if this is a snapshot of a single-day's distribution of model years, but if this data is collected over time, it will be necessary to account for the fact that the model year of a car is fixed, and does not change relative to today's date. If you collected these data over a period of several years, you will also be reporting the fact that there are more opportunities for 2003 model year vehicles to appear: every year since 2002.

Sycorax
  • 76,417
  • 20
  • 189
  • 313
1

The "spike" or "bell-shaped curve" just means that intermediate model years, from 1995 to 2000, are more likely than extreme early or later model years. Certainly, more cars will be sold from an intermediate period where they are no longer new but still usable, than from earlier periods when they are old and unusable (low demand), as well as later periods when they are still brand new and no need to sell them (low supply).

Kelvin
  • 1,051
  • 9
  • 18
  • What can be said of the peak-to-present time? For instance, if people buy new in the then present (_i.e._ the past) does this mean that they generally give up their vehicles after 17 years. – Ubra Apr 14 '16 at 01:58
  • On average, roughly yes. In fact if you calculate the mean and standard deviation, you could say that people give up their cars after (present year - mean year) +/- SD years. However it is slightly more complex than that because more recent second hand sales may occur via other channels (e.g., back to the original garage), and some of the sales in the chart shown may be third or fourth hand sales. – Kelvin Apr 14 '16 at 06:16
  • The only thing you could really say without making assumptions, is that most of the sales through that auction are cars from the 1995-2000 period, but whether they are second, third or fourth hand sales we don't know. We also don't know *when* these cars were sold, so it could be that the auction site has been in decline and losing business in recent years! – Kelvin Apr 14 '16 at 06:44