Abstract:
|
For various reasons, survey respondents fail to provide data after periods of cooperation that vary in length. Likewise, after a period of nonresponse, respondents may begin providing data again. The purpose of this paper is to provide statistical estimates of response length, defined as consecutive months with a price reported, and its converse, nonresponse length, for items reported to the Producer Price Index (PPI), a principal federal economic indicator. The estimates are produced via the use of Cox Proportional Hazard models and their derived Survival Curves, which are modified to predict the occurrence of both nonresponse and return to cooperation from nonresponse over different, finite time horizons. Subsequently, estimates of response and non-response length are produced for a set of two-digit NAICS sectors representing up to 564 six-digit NAICS industries, which are then plotted in the x-y plane for comparison. The results suggest modest variation in response and nonresponse across NAICS sectors, but also the existence of seasonal variation in the occurrence of nonresponse. The estimates produced may provide guidance in handling nonresponse.
|