An Input-Output Analysis of the Relationships Between Communications and Travel for Industry

Numerous public policies have been promulgated on the assumption that telecommunications will be a useful trip reduction instrument. However, many scholars have suggested that the predominant effect of telecommunications may be complementarity – increasing travel. Although short-term, disaggregate studies of single applications such as telecommuting have tended to find a substitution effect, more comprehensive studies, on the aggregate scale, are needed. One of the few such studies used input-output analysis to examine relationships between transportation and communication input intensities across 44 industry classes in Europe for 1980, and found strong evidence of complementarity. The present study has applied a similar methodology to the input-output accounts for the US across multiple points in time. Specifically, this study applied the input-output analysis technique developed by Leontief in 1936 to analyze the relationship between transportation and communications as industrial inputs in the U.S. Generally, input-output analysis offers a static view of the structural relationships, expressed purely in monetary terms, among the different sectors of an economy for a certain period of time. We analyzed correlations between transportation and communications using the input coefficients of transportation and communications in the input-output table (direct coefficients matrix). Positive correlation coefficients indicate complementarity: industries that require a lot of transportation inputs also tend to require a lot of communications inputs, and conversely. Negative correlation coefficients imply substitution. Ten benchmark I-O accounts (between 1947 and 1997, inclusive) were collected, which are prepared using the most detailed data sources available, generally the economic censuses. Trying to find the best balance between highly disaggregated industry classifications (which may exhibit a lot of random noise that would obscure the pattern of interest) and highly aggregated ones (which contain such a small number of cases that it may also be hard to identify underlying relationships), we created four scenarios reflecting different levels of aggregation across sub-industries. Scenario 1 is the most disaggregate level (containing 79-131 categories, depending on year), while Scenario 4 is the most aggregate categorization, containing just the nine top-level industries. We analyzed correlations for five selected pairs of transportation and communications industry categories: the manufacturing pair (i.e., transportation manufacturing correlated with communications manufacturing), the utilities pair, the two manufacturing-utilities pairs, and the overall pair (all transportation manufacturing and utilities correlated with all communications manufacturing and utilities). In this study the Spearman correlation is used, which is a nonparametric correlation measure. Since the input-output coefficients are not normally distributed, the Pearson correlation coefficients are not strictly appropriate. Using Spearman correlations, we conducted a cross-sectional analysis for each time period, and compared results across time based on the five sets of correlations between transportation and communications. Thus, 200 correlation coefficients in all are computed in this study (five sets for each of four scenarios for each of 10 benchmark years).