Associate Professor of Accounting

McCombs School of Business

University of Texas at Austin

(512) 471-5332

lisa dot desimone at mccombs dot utexas dot edu

Outbound Scores from Using IRS data to identify income shifting to foreign affiliates (with Bridget Stomberg and Lillian Mills, Review of Accounting Studies 24(2): 694-730).

Outbound Scores Datasets

Calculating Outbound Score

Outbound Score is a firm-year score increasing in the magnitude of net intercompany transfers out of the United States. To construct a parsimonious measure useful to researchers, we apply the parameter estimates from our ordered logit to firm-year characteristics .

Outbound Score = 0.6933*RD - 1.8854*AD + 0.4377*SGA + 0.2634*Intangibles + 0.0197*Tobin’s Q + 0.4057*CapEx - 0.1447*Soft Assets - 2.2314*GP% + 0.6527*High Tech - 1.3845*Foreign Sales% - 0.5382*DROS + 1.4334*FROS + 0.0772*FSales Growth + 0.2470*DSales Growth + 0.3329*FTR - 0.2477*Leverage + 1.5965*Interest - 0.0414*Size + 0.0451*Big5 + 0.1615*Non-Crisis + 0.0968*Non-Durables - 0.1232*Durables - 0.2811*Manufacturing - 0.2629*Oil & Gas - 0.5852*Chemicals + 0.0975*Bus. Equip. - 0.1633*Telecom - 0.3829*Shops + 0.2095*Healthcare

where all inputs are from Compustat and:

  • RD = R&D expenditures (XRD) scaled by Sales (SALE). We reset missing RD to zero.

  • AD = advertising expense (XRD) scaled by Sales (SALE). We reset missing SD to zero.

  • SGA = selling, general and administrative expense (XSGA) scaled by Sales (SALE). We reset missing SGA to zero.

  • Intangibles = balance sheet intangible assets (INTAN) scaled by Sales (SALE).

  • Tobin's Q = Assets (AT) plus market value of equity (PRCC_F*CSHO) scaled by Assets (AT).

  • CapEx = capital expenditures (CAPX) scaled by Sales (SALE). We reset missing CapEx to zero.

  • Soft Assets = Assets (AT) less net property, plant, and equipment (PPENT) and cash (CH), scaled by Sales (SALE).

  • GP% = gross profit (GP) scaled by Sales (SALE).

  • High Tech = 1 for firms in the following three-digit SIC codes, following Francis and Schipper (1999) and Core, Guay and Van Buskirk (2003): 283, 357, 360-368, 481, 737, and 873. 0 otherwise.

  • Foreign Sales% = foreign sales reported in the Compustat Segments database scaled by total sales reported in the Compustat Segments database. If Foreign Sales are missing and pre-tax foreign income (PIFO), current foreign tax expense (TXFO) and deferred foreign tax expense (TXDFO) are all zero or missing, we set Foreign Sales% to zero.

  • DROS = domestic pre-tax income (PIDOM) scaled by domestic sales, where domestic sales are obtained from the Compustat Segments database.

  • FROS = foreign pre-tax income (PIFO) scaled by foreign sales reported in the Compustat Segments database.

  • FSales Growth = average annual percent change in foreign sales reported in the Compustat Segments database from t-2 to t.

  • DSales Growth = average annual percent change in domestic sales reported in the Compustat Segments database from t-2 to t.

  • FTR = U.S. statutory tax rate of 35 percent less the firm’s foreign effective tax rate (TXFO plus TXDFO, scaled by PIFO) in year t-1.

  • Leverage = long-term liabilities (DLTT) scaled by Sales (SALE).

  • Interest = interest expense (XINT) scaled by Sales (SALE).

  • Size = natural log of Sales (SALE).

  • Big5 = 1 if the firm is audited by a Big 5 audit firm, 0 otherwise.

  • Non-Crisis = 1 for all years except 2008 to 2010, zero otherwise.

  • Non-Durables, Durables, Manufacturing, Oil & Gas, Chemicals, Bus. Equip., Telecom, Shops, and Healthcare represent indicator variables for Fama-French 12 industry classifications. We adjust these 12 classifications to remove three-digit SIC codes included in High Tech.

This approach is similar to that used by Hadlock and Pierce (2010). For parsimony and consistent with Hadlock and Pierce (2010), we calculate a score instead of a probability. Given that we rank our dependent variable into deciles, calculating a probability as in Wilson (2009) or Lisowsky (2010) would require identifying the maximum of ten probability calculations per firm-year observation.

Recommendations for Sample Selection

  • Corporations incorporated in the USA (FIC=USA)

  • Total assets greater than $10M (AT>10000)

  • In industries other than Financial (SIC = 6000-6900) or Utilities (4900-4999)

  • Non-missing, non-zero pre-tax foreign income (PIFO not in [0,.])

  • Other sample selection criteria required to define input variables as described above

  • -1<=FTR<=1

Recommendations for How to Use Outbound Score

We believe Outbound Score will be useful for researchers examining outbound income shifting by U.S. MNEs. For researchers wishing to compute the score on their own, we recommend restricting the sample to firms meeting the above sample selection criteria. The measure is not appropriate for foreign-incorporated MNEs.

We find the model has similar explanatory power for firms in both high tech and non-high tech industries. We therefore believe it appropriate to use the score in samples that include all industries. We also find, however, that the explanatory power of the model is greatest when estimated using firms with greater than $1B in total assets. Therefore, researchers should consider gauging the sensitivity of results to restricting the sample to the largest firms.

To assist researchers in identifying situations in which our measure might be better suited for their research question than the measure from Collins et al. (1998) and its variants (e.g., Klassen and Laplante 2012), we next compare observations where the measures generate inferences about outbound shifting. Specifically, we examine observations for which our score predicts the greatest amount of net outbound payments but which the Collins et al. (1998) measure does not classify as shifting income, and vice versa. We find that the observations our measure uniquely identifies as engaging in outbound income shifting: (i) are younger and have a smaller foreign footprint evidenced by lower foreign profitability, lower levels of worldwide profitability, and lower gross profit margins, (ii) are more likely to be in high tech industries, (iii) have higher sales growth, and (iv) have higher R&D spending. The observations we uniquely identify as engaging in outbound income shifting have more outbound payments for services, borrowing, and other non-IP-related transactions. These differences could suggest our measure better detects firms in the earlier stages of global expansion; these firms report lower foreign profits than worldwide profits, but these low foreign profits are nonetheless likely inflated because of significant net outbound payments to their foreign affiliates. Indeed, untabulated descriptive tests reveal these firms are younger and have reported foreign income for fewer years. In contrast, although Collins et al. (1998) intend their measure to detect tax-motivated pricing of intercompany transfers, it also identifies firms with large and profitable foreign footprints, which could also stem from (tax and non-tax motivated) investment and location decisions. We conclude that our measure performs well at isolating the portion of outbound income shifting attributable to intercompany payments. In contrast, the Collins et al. (1998) measure performs well at identifying firms that report a larger-than-expected amount of profitability in lower-tax jurisdictions – regardless of whether that profitability stems from intercompany payments. Researchers can use both measures to develop a more comprehensive understanding of how firms achieve greater profitability abroad.