Villalonga (2004) Lang and Stulz (1994), Berger and Ofek (1995), and Servaes (1996) find that diversified firms trade at an average discount relative to single-segment firms; this suggests diversification destroys value Villalonga (1999) and Campa and Kedia (2002) show that the discount is only the product of sample selection biases; diversified firms trade at a discount prior to diversifying, and when selection bias is corrected for the diversification discount disappears This paper questions the “discount” further by investigating the possibility that the discount is an artifact of Compustat’s segment data Compustat provides disaggregated financial information for business segments that represent at least 10% of a firm’s sales, assets, or profits, and prior studies have mostly relied on this data to determine which firms are diversified and how much The author uses a new census database instead
Problems with Compustat 1.The extent of disaggregation in financial reporting is much lower than the true extent of a firm’s industrial diversification; firms that appear to be “single-segment” may not be 2.4 digit SIC Codes are too broad in many cases to pin down exactly what activities the firm is undertaking 3.Financial accounting standards allow managers considerable flexibility in how segments are constructed; the aggregation of activities into any given segment differs from firm to firm 4.Segments are self-reported and Compustat assigns them into SIC codes 5.Firms may change the segments they report even when there is no real change in their operations; what appears to be diversification or focusing may just be reporting changes
The Longitudinal Research Database (LRD) Some previous studies use the LRD The LRD covers only U.S. manufacturing establishments (plants) The new census database includes non-manufacturing establishments This is important; less than 20% of all multi-segment firms in Compustat’s segment files are manufacturing-only; 56% are non- manufacturing only; only 24% are diversified across both sectors The majority diversified across both sectors have most of their assets in non-manufacturing
The Business Information Tracking Series (BITS) BITS is a new census database that covers the whole U.S. economy at the establishment level These data allow the author to construct business units The author uses a common sample of firms and Lang and Stulz’s method to compare the value estimates obtained using BITS to those obtained using Compustat The author finds a diversification discount using Compustat segments, but a diversification premium using BITS business units The results call into question the adequacy of segment data for research in corporate finance, strategy, etc.
Explanations The author offers two explanations for the results: 1.Relatedness: Compustat segments measure mainly unrelated diversification (because similar activities get grouped into one segment), whereas BITS allows the author to measure all types of diversification. Thus, the results suggest that related diversification is associated with a premium and unrelated diversification is associated with a discount 2.Strategic Accounting: Diversified firms aggregate their activities into segments in ways that make them appear to be poor performers
Data BITS provides establishment level panel data between 1989 and 1996 for all U.S. private-sector establishments with positive payroll in any of these years, from both private and public firms BITS includes over 50 million establishment-year observations from over 40 million firm-years The basic unit of analysis in BITS is the business establishment, defined as “a single physical location where business is conducted or where services or industrial operations are performed” For each establishment-year observation, BITS contains information on its employment, annual payroll, primary four-digit SIC code, location, start year, the firm and legal entity to which the establishment belongs, and the firm’s total employment A “legal entity” is a corporation, partnership, or sole proprietorship A “firm” in BITS is “the largest aggregation of business legal entities under common ownership and control”
Coverage BITS contains data on the entire population of U.S. establishments from all sectors of the economy, excluding farms, railroads, the Postal Service, private households, and large pension, health, and welfare funds BITS contains no performance data; the author links BITS with Compustat to overcome this problem This limits the sample to publicly traded firms The author limits the sample further by following the previous literature and excluding firms that operate in the financial sector, agriculture, the government, etc. The author also eliminates observations with missing variables and outliers (firms whose imputed q is higher than four times or lower than a fourth of their actual q)
Sample The resulting sample has 1,555,371 establishment-year observations from 12,708 different firm-years The author aggregates all of a firm’s establishments with a common SIC code into “business units” This is the BITS-based analytical equivalent of the Compustat segments, but the number of business units is not limited to 10 The maximum number of business units within a firm is 133 On average, there are 122 establishments per firm, 15.5 establishments per business unit, and 7.9 business units per firm The average number of segments per firm is 1.7 (4.6 business units per segment) 43% of the sample switches from undiversified to diversified when one moves from a segment analysis to a business unit analysis 21 percent of the firms have more than 10 business units
Limitations of BITS BITS includes only U.S. establishments and their employees because it is U.S. Census data Compustat includes foreign operations, employees, etc. The author uses Compustat’s geographic segment files to determine the percentage of U.S. vs. non-U.S. operations for each firm Tests suggest that the extent of coverage in BITS is uncorrelated with the variables the author uses in this study, so the author uses the full sample for most of the results The author checks the robustness of the main results by running everything after excluding firms with non-U.S. operations Another limitation is that BITS still relies on 4 digit SIC codes, which aggregate some unrelated activities and also distinguish between some related activities (related in terms of production technology or markets or both); SIC codes need not correspond to a firm’s actual business units
Excess Value Measures Excess Value is measured two ways: 1.The difference between a firm’s Tobin’s q and its imputed q 2.The natural logarithm of the ratio of Tobin’s q to its imputed q Tobin’s q is measured using the ratio of the market value of common equity plus the book value of preferred stock and debt to total assets; Imputed q’s are measured two ways: 1.In segment data, the imputed q is the asset weighted average of the hypothetical q’s of the firm’s segments, where a segment’s hypothetical q is the average of the single-segment firms in the industry-year 2.In BITS data, the imputed q is the employment weighted average (assets are not measured by BITS) of the hypothetical q’s of the firm’s business units, where the hypothetical q is the average of the single-business unit firms
Industry q Measures How closely do the SIC codes in Compustat match those in BITS? The author checks this using the subsample of firms that are classified as single-segment in Compustat 51% of the single-segment, single business unit firms have the same 4 digit SIC code in both BITS and Compustat 9% match at the 3 digit level; 8% at the 2 digit level; 8% at the 1 digit The remaining firms (about 25%) do not match at all Among single segment but multibusiness firms, 84% choose their single segment’s SIC from among the set of 4 digit codes they operate in according to BITS However, only 14% choose to report the 4 digit SIC code of their largest business unit (measured using employment, not assets) Overall, in 81% of the cases, firms choose a segment SIC code that does not match that of its largest business
Results The author compares the average excess value of multi-segment firms to that of single segment firms and confirms the finding in the previous literature: multi-segment firms trade at a statistically significant discount relative to single-segment firms When the author performs a similar analysis comparing multi-business unit firms to single business unit firms, there is a statistically significant premium The author performs several robustness checks, including using alternative measures of industry q and more complex measures of diversification, checking whether weighting segment q’s by employment instead of assets is sufficient to remove the discount, restricting the sample to firms with 100% of their operations in the U.S., reconstructing business units while imposing a 10% materiality condition similar to Compustat’s (using employment instead of assets), and combining vertically related activities into one business unit; the results hold up