Regression Analyses / New Italian Regulations
Reducing Errors in Regression Analyses
In an article in “The Value Examiner”, Keith Sellers, Michael Huang, and Jacob Sellers examine the application of data transformation within the framework of ordinary least squares (“OLS”) regression for the purpose of reducing prediction errors. As background for their explanations, they include a discussion of major assumptions that underly OLS regressions, specifically the absence of multicollinearity, of heteroscedasticity of residuals, normal distribution of the residuals, and linear relationships between the dependent variable and each independent variable. Following this introduction, they introduce a data set used to demonstrate the effect of the transformation proposed. This is a set of publicly traded US banks as of 31 December 2020 with total assets between 150 million and 5 billion USD. This set consisted of 252 banks which was divided randomly into a training sample of 100 and a test sample of 152 banks. For the regression analysis, the market value was used as the dependent variable and a variety of financial measures – such as total assets, total loans, net income, or average daily trading volume – were considered as independent variables. As most of these possible independent variables were primarily capturing size, the authors by transforming them into size-independent ratios and were then running a reverse stepwise regression, removing independent variables that showed with a P-value above 0.1 low significance. The results, derived from the training sample, where then back-tested with both samples to derive the percentage error. The regression was then repeated after additionally transforming the data with the natural logarithm, to consider that financial data is typically distributed lognormal. The reversal of this transformation, to derive a market value in dollars. As expected, the back-testing with the training sample – as biased – resulted in better predictions, making the test sample a better measure for the reliability of the prediction. The lognormal transformation showed here substantially better results, also allowing for the use of five independent variables, instead of four with the non-transformed data.
The article “Regression Analysis and Business Valuation: A Transformative Approach for Reducing Error” by Sellers, Huang, and Sellers was published in the January/February 2022 issue of “The Value Examiner” on pages 4 through 13.
New Italian Documentation Guidelines
On 26 November 2021, the Italian revenue service (agencia entrate, here: “AE”) published a new circular on transfer pricing (“TP”) documentation, aimed at providing clarification on suitable documentation that allows verification of compliance with the arm’s-length principle. It fully replaces the previous circular on this topic. Besides explaining the regulatory environment, the circular focuses on different types of documentation, specifically the Master File, the National Documentation, documentation for permanent establishments, and for small and medium-sized businesses. The AE also include a section that addresses the simplified TP approach for low-value-adding services (“LVAS”). This is an approach introduced by the OECD for certain “routine-type” support activities that allows for a simplified pricing, using the costs incurred for the provision of the services plus a markup of 5 percent. The benefit of this approach for the taxpayer and the tax administration is that it allows to forego certain aspects of the comparability analysis, especially the search for comparable transactions. The AE provides detailed information on how to document the application of the LVAS approach, including detailed descriptions of the services, the written contracts (which should be prepared in advance), the calculation of the overall costs of the service provider, and the allocation to the service recipients.
From our perspective, the AE deserves consideration in all circumstances where TP analyses involving Italian entities are involved. Due to the strict approach of the local tax administration, the final documentation should always involve a local advisor. Nevertheless, clarity about the documentation requirements already during the actual analysis will already be helpful in reducing efforts later in the documentation phase. The LVAS approach is a good example. This approach is already widely applied, since pioneered by the OECD. Nevertheless, it can be observed that this is often done in a pragmatic, cost-efficient approach which may forego the level of detail that even the OECD Guidelines indicate. The explicit mentioning in the AE reinforces that the approach is also acceptable for Italy. By mirroring the documentation requirements in the OECD Guidelines, the AE emphasizes that the application of the LVAS-approach is still subject to detailed documentation. Finally, considering the documentation requirements early is especially important for Italian entities, as the local regulations set a clear deadline when the TP documentation must be at hand – the time of filing of the tax return which needs to indicate the existence of the document. As evidence, the TP documentation in electronic form needs to be signed with a certificate – a ‘marca temporale’. Globally, this is a comparatively strict approach and TP analysts are well advised to consider this fix deadline in their project, to avoid unnecessary delays, as TP documentation not prepared by the necessary date will not be able to provide penalty protection for the client. Circular 15/E on TP documentation is available online on the website of the AE here: https://www.agenziaentrate.gov.it/portale/documents/20143/3930122/Circolare+oneri+documentali+TP+26+11+21+h.+13.20+-+REV.pdf/5f23fab7-0853-68c6-cb5d-868fdfed9490