While we should view database-input information suspiciously, insights may nonetheless emerge. Joel Anderson recently compared the reported property damage costs with the reported release volumes in PHMSA data 2010 to 2018.
This is just pipeline spills, excluding tanks, etc. Note this plot is log-log scale so small changes can equate to significant $$.
Again, ‘user beware’! Take this only as one possible, and suspect, starting point in your examination of pipeline spill history.