#We have not had much time to study polynomial regression, but let's try a little bit here. Take this US census data.
#We'll fit a 2nd-order (quadratic) and 3rd-order (cubic) model to the data up to 1960, then use it to predict into the "future" of 1970-2010 to see which one gives better results.
#As a side note, the coefficients are nicer if we recenter the x values so x=0 is the year 1900, so that's already done here:
Year,YearMinus1900,Population
1790,-110,3929214
1800,-100,5236631
1810,-90,7239881
1820,-80,9638453
1830,-70,12866020
1840,-60,17069453
1850,-50,23191876
1860,-40,31443321
1870,-30,38558371
1880,-20,49371340
1890,-10,62979766
1900,0,76212168
1910,10,92228496
1920,20,106021537
1930,30,123202624
1940,40,132164569
1950,50,151325798
1960,60,179323175
1970,70,203211926
1980,80,226545805
1990,90,248709873
2000,100,281421906
2010,110,308745538