NICHOLAS M. DIDOW, JR. and GEORGE R. FRANKE* Most time-series research fails to consider measurement issues that are addressed regularly in cross-sectional research designs. However, to measure specified con- structs in time-series research, scales may be developed whose reliability and va- lidity may be assessed as rigorously as in cross-sectional research. The authors describe and illustrate a procedure for developing time-series measures using Box- Jenkins ARIMA modeling techniques in an investigation of the effects of national advertising on consumption. Measurement Issues in Time-Series Research: Reliability and Validity Assessment in Modeling the Macroeconomic Effects of Advertising Methods for the analysis of time series have become widely known and are now applied regularly by mar- keters and other social scientists. Time-series analysis has proved useful in such areas as forecasting (e.g., Ka- poor, Madhok, and Wu 1981; Moriarty and Adams 1979), assessing the relationship between advertising and sales (e.g., Aaker, Carman, and Jacobson 1982; Jagpal and Hui 1980), examining market responses to specified in- terventions (e.g., Franke and Didow 1981; Wichern and Jones 1977), and testing hypothesized concomitancies in time-series designs (e.g., Hanssens 1980). The availa- bility of programs for the analysis of time series in such widely distributed statistical packages as BMDP (1981), SAS (SAS/ETS User's Guide 1980), and SPSS (Hull and Nie 1981) is likely to contribute to further examination of time series in a variety of research areas. Developments in time-series analysis have not ad- dressed issues of measure reliability and validity. The reason may be that time-series research typically in- volves such nonpersonal variables as sales, advertising *Nicholas M. Didow, Jr. is Assistant Professor of Marketing, Grad- uate School of Business Administration, University of North Caro- lina. George R. Franke is Assistant Professor of Advertising, College of Communication, University of Texas at Austin. The authors gratefully acknowledge the many helpful suggestions of the two anonymous JMR reviewers. 12 expenditures, and market share, whose measurement is presumably not subject to many human limitations af- fecting measure quality such as inability or unwilling- ness to give information and effects of various other per- sonal and situational factors (Churchill 1979, p. 65). In both time-series and cross-sectional research, though, unless strong a priori evidence is available that a con- struct is captured accurately by a given measure, steps should be taken to develop and assess the best measure possible. The purpose of our article is to suggest a procedure for developing and assessing measures of constructs in time-series research. First the relevance of reliability and validity issues to time-series analysis is discussed briefly. Next, a general procedure is proposed for measure de- velopment and for reliability and validity assessment in time-series designs. The procedure then is illustrated in the marketing research context of examining the macro- economic effects of advertising. MEASURE ASSESSMENT IN TIME-SERIES RESEARCH As Peter points out (1979, p. 6), "Valid measurement is the sine qua non of science." Unless a measure shows evidence of reliability and validity, findings based on it must be regarded with suspicion. Issues in measurement assessment in cross-sectional research are well known to marketers. However, certain aspects of time-series anal- Journal of Marketing Research Vol. XXI (February 1984), 12-19