While a number of researchers of world development examine social change using composite measures as indicators, there is a relative paucity of research on the reliability of these change score composites over time. We construct two development composites based simply on factor analysis, one economic and one social, and then perform reliability analysis on these two development composites at two discrete points in time (i.e. 1970 and 1985) and their change over the 15-year period defined by their beginning and ending points. Despite evidence of reliable beginning and ending points, change in composites over time yield markedly different patterns of reliability. We conclude that if composite indicators of development are used in cross-national research to assess change, the reliabilities of their change should be addressed directly in addition to the reliabilities of their beginning and ending points. The risk of not doing so is faulty inferences with respect to theory.