Transit/parking related discussion from the General Developments thread that Steely deemed too off-topic despite having to do with proper amounts of parking in downtown Chicago, which seems like a "General Development" issue to me, but OK...
Quote:
Originally Posted by dagobert
Association is not causation, or a more statistical term Correlation is not Causation. In any introductory statistics or econometric analysis course you would learn that there are a number of dependent variables that impact the independent variable.
To say that their is a relationship between amount of parking downtown and public ridership you would have to do a bit of cross-sectional time series data analysis (also known as panel data or longitudal data analysis) which also looks at other contributing factors such as price of oil (we saw a big decline in price of oil in mid 1980s through late 1990s), amount of jobs downtown vs. suburbs, amount of people living downtown vs. suburbs, amount of crime recorded in the city vs. suburbs, amount of crime on CTA trains, cost of driving one mile in a car vs. cost of taking CTA train one mile, cost of parking, etc. Maybe a bigger contributing factor to a decline in ridership after 1984 was decline in price of oil and thus cost of drving and not building of parking in the loop. I might poke around some databases to see if such a study had ever been done for any major American city.
We also have to worry about omitted variable bias (or confounding) since we aren't talking about a controlled experiment but an observational study (looking at historical data). For example we can't measure perception of how safe people feel taking CTA trains as opposed to a car.
Also social attitudes (towards driving, commuting downtown, living in the city) in mid-70s might have been similar to those in early 80s since social attitudes are fairly similar from one year to the next, but they may vary considerably over longer period of time. So if this is true that social attitudes in late 90s are different than in 70s and assumption of independent error terms across observations in a time series is violated. The reason why this is important is because under the classical econometric model error terms for each observation need to be independent of one another. Otherwise error terms reflect omitted variables that influence the demand for parking or public transit ridership. This could also lead to autocorrelation and other problems.
Hope this helps you understand the sheer complexity of analyzing such complex problems as this one, those results are scientific and unbiased.
|
If you actually read the post of mine that you cited, you'd note that I
specifically stated that my stats didn't constitute causal proof, but whatever. I'd consider posting stats on historical transit ridership, gas prices (remember when transit ridership plummeted in the late 90s when oil was $15/barrel? Me neither.) but I hesitate to bother...