Accruals and future performance: Can it be attributed to risk?

Abstract

We decompose broad-based measures of accruals into firm-specific and related-firm components. We find that the negative relation between accruals and future firm performance is almost entirely attributable to the firm-specific component. Standard risk-based explanations are hard to reconcile with this fact. To the extent expected returns have a common component spanning related firms, a risk-based explanation would suggest a stronger negative relation between accruals and future firm performance when related firms are also growing. Instead, the attenuation we document is more likely attributable to suboptimal investment decisions, which the stock market and analysts do not incorporate in a timely manner.

Appendices

Appendix 1

1.1 Bureau of economic analysis survey tables

We use the Benchmark Input–Output Surveys of the Bureau of Economic Analysis (BEA Surveys) as the basis for identification of economically linked industries. These data allow us to cleanly identify linkages across customer and supplier industries. The BEA surveys provide a detailed view into the interdependencies across industries based on the production and consumption of various goods and services. The BEA Surveys are updated every 5 years and are dated with a “look-back,” and we are careful to ensure that we use this data when it was publicly available. For example, the tables identified as “2002” were released in September 2007 to cover the industry output from 2002 through 2007. We only use this data after the public release of the “2002” table during the 2007 calendar year.

The BEA Surveys contain a variety of tabulated information. We are most interested in the MAKE and USE tables. The MAKE table is a I × C matrix populated with the dollar production of each commodity, c, by each industry, i. Thus the sum of the cells in each row (industry) of the MAKE table reflects the total production of commodities for that industry. The USE table is a C × I matrix populated with the dollar consumption of each commodity, c, by each industry, i. Thus the sum of the cells in each row of the USE table reflects the total consumption of a given commodity across all industries.

We need to make several research design choices when using the BEA Surveys. First, we need to decide on the granularity of industry definition. The BEA Surveys are provided at a detailed, summary and sector level. For the 2002 BEA Surveys, the dimensionality of the MAKE and USE tables across these three levels are as follows: (1) detailed (430 industry codes), (2) summary (133 industry codes), and (3) sector (15 industry codes). We use the summary level BEA Surveys in our empirical analysis. Second, we need to combine some intermediary industry codes to allow mapping back to standard industry classification schema such as SIC and GICS. These are performed manually for a small number of industry codes (see Menzly and Ozbas 2010 for details). Third, we need to combine the MAKE and USE tables to create a balanced I × I matrix reflecting the proportional use of commodities that are produced and then used across industries within the US economy. To do this, we convert the MAKE table to reflect the proportion of a given commodity that is produced by a given industry. The dollar amounts in the cells of the I × C MAKE table are therefore scaled by the respective sum of each row (i.e., the total amount of a given commodity that is produced by a given industry, relative to the total amount of commodities produced by that industry). Likewise, we convert the USE table to reflect the proportion of a given commodity that is consumed by a given industry. The dollar amounts in the cells of the C × I USE table are therefore scaled by the respective sum of each row (i.e., the total amount of a given commodity that is consumed by a given industry, relative to the total amount of that commodity that is consumed across all industries in the US economy). We then take the matrix multiplication across the modified MAKE and USE tables to create an I × I industry level input–output table.

Appendix 2 shows the final input–output table for the sector level (15 industry codes) using the 2002 BEA Survey tables. For example, the agriculture, forestry, fishing and hunting sector (labelled as AGRIC) consumes 31 % of the commodities that it produces and the bulk of the rest is consumed by the manufacturing sector (labelled as MANUF). It is clear from this visualization that there is a concentration of economic activity along the main diagonal. Thus our input–output matrix reflects the combined effect of related firms in the same industry and related firms in different industries. Not surprisingly, there is a strong within-industry economic interdependence between firms in the US economy. In our empirical analysis, we separately examine the two types of related firms. Peer firms are those in the same industry grouping (i.e., diagonal elements of the input–output table), and nonpeer firms are those along the supply chain (i.e., non-diagonal elements of the input–output table).

Appendix 2

See Table 7.

Table 7 Visualization of the 2002 sector level input–output table

Appendix 3

See Table 8.

Table 8 Variable definitions