Correlation and/or Causation? Trend or no trend?

Question

http://courses.dcs.its.utexas.edu/speedway-files/college/eco304k/L2Q8.gif

I can't tell. too many outliers? Is there Correlation? Is there Causation? Is there a trend!?

Answer 1

Ok.  First off, what are you looking for?  Correlation between what?  Causation between what?  You have no hypothesis, so I'm confused.

Causation between household expenditures on autos and household income?  Okay... so then your data can either be saying that as HHI increases, HH auto expenses also increase... but it can also say the opposite - as HH auto expenses increase, HHI must also increase.  (That would indicate causation, btw.)  But I wouldn't draw any conclusions with regards to causation because you haven't proved that increasing HHI causes people to spend more on autos.  All you can say is that as one increases, so does the other - i.e. you have direct relationship between HHI and auto expense (as one increases, so does the other).  Causation can only be determined once you have looked at other possible causes of the relationship.  In this case, that would mean that you would have to control for, among other things, inflation, rising cost of autos in general (note that in the time period you're analyzing expensive, gas-guzzling SUVs became popular), and lots of other things which I can't mention now, because I don't know what you're trying to prove - causation between what?

As for outliers, imo you only have one - '95, and even that isn't so far off.  And that means you do have good correlation between the variables - which means that HHI is a good predictor of how much money a household spends on autos per year.

Answer 2

there seems to be a loose correlation, since it does have a generally upward sloping trend.  you can't tell causation from the graph, though, since correlation does not imply causation