It is difficult to compare time series. Indexing can help. Can be calculated in one way or the other. Equal with all: Everything is compared with the same point in time. It  has to be chosen carefully. Equal with all: You don’t see the differences at the beginning any more. You can’t help that. Therefore: Watch out!

Have a look.

Good, Not Great: Cumulative change in number of jobs. Quelle: Wall Street Journal, 06.01.2012.Engine and Caboose: Percentage change in the ISM manufacturing index vs. NFIB. Quelle: Wall Street Journal, 01.12.2010.
Source: “Wall Street Journal”, left 2012–01–06, rechts 2010–12–01.

Here, it was calculated like that: How far away are all the values with respect to January 2010 (left) and January 2005 (right)? In percent. The beginning is 0 percent. How big were the differences then? This you don’t see.

Commodity prices, January 6th 2009=100; coffee, orange juice, wheat. Quelle: The Economist, 04.08.2010.

Source: The Economist, 2010–08–04.

Here, it was calculated like that: All values have been divided by January 2009. And multiplied by 100. The beginning is 100. How big were the differences then? This you don’t see.

Slowest Recovery Since the 1960s. Quelle: Heritage Foundation, 30.08.2012.
Source: Heritage Foundation, 2012–08–30.

Here, things were cut out and glued together. Actually, every chart had another starting point: in the 60s, 70s, 80s and so on. Now, in the graphic every chart has the same starting point. How big were the differences then? This you don’t see. Has the starting point been carefully chosen? Invictus says no.

Even more can disappear with an index: have a look at Paresh. Better than disappindex is logarithm.