Source of Data Incentives and Disincentives of Profit S

51 affiliations in different location high and low tax environment. 250 However, since that HL approach is more relevant with theoretical and analytical framework of this thesis, Dharmapala and Riedel’s approach will beyond scope of analysis.

4.4. Source of Data

As mentioned before, I will focus on micro level framework particularly on individual firm data. Analytical framework on profit shifting, explained either by HR or HL approach also stressed out the importance of profitability financial accounts of multinational enterprise. Therefore, the database for analysing profit shifting strategies should be the database contains information on profitability and financial account with uniform format. Based on literature review, this information is published both by government institution and private sector. Today, there are at least three government institutions which routinely maintain data on activities and financial report of multinational enterprises, namely: Bureau of Economic Analysis BEA Operations of Multinational Companies Database in U.S., Deutsche Bundesbank Micro- database on Direct Investment MIDI, and UK Office of National Statistics Annual Inquiry into Foreign Direct Investment AFDI. Dozens of empirical studies have utilized these databases, for instance Buettner and Ruf 2007 has been used firm-level MIDI database to analyse location decision of German’s multinational enterprises 251 ; or Egger et al. who used the same data to analyse internal borrowing of multinationals. 252 Unfortunately, although these databases are great in detail, they cannot be accessed by public and only focused on information of multinational enterprises originated from their countries. On the other hand, private institution such as Bureau van Dijk produces various databases about activities and financial performance of firms, for instance: ORBIS, ORIANA, AMADEUS, COMPUSTAT, and others. 253 These databases contain rich information on size, profit and loss account, activities, industry, balance of payment, performance indicator, legal status, or even information of ownership and their affiliations. Concerning to their advantages, these databases has been used in many empirical studies on profit shifting. Karkinsky and Riedel 2012 or Dischinger and Riedel 2008 used Amadeus database, 254 while ORBIS database used by Loretz and Moore 250 See their methodology at Dhammika Dharmapala and Nadine Riedel, “Earnings Shocks and Tax-motivated Income- shifting: Evidence from European Multinationals,” Journal of Public Economics, Vol. 97, No. 1 2013: 97-98. 251 Thiess Buett ner and Martin Ruf, “Tax Incentives and the Location of FDI: Evidence from a Panel of German Multinationals,” International Tax and Public Finance, Vol. 14, Issue 2 2007: 151-164. 252 Peter Egger, Christian Keuschnigg, Valeria Merlo, and Georg Wamser, “Corporate Taxes and Internal Borrowing within Multinational Firms,” NBER Working Paper, No 18415 2012. 253 Database published by Bureau van Dijk can be distinguished according to territory region and types of firms. For instance, OSIRIS only contains information for public firms in Asia Pacific region. See http:www.bvdinfo.comen- gbhome for further information. 254 Tom Karkinsky, and Nadine Riedel, “Corporate Taxation and The Choice of Patent Location within Multinational Firms,” Journal of International Economics, Vol. 88, Issue 1 2012: 176-185; or Matthias Dischinger and Nadine 52 2013. 255 In this thesis, I choose ORBIS as my main source of data. With regards to its coverage, ORBIS could considered as the most complete database released by Bureau van Dijk as it provide information for more than 140 million worldwide firms. ORBIS, as well as many products from Bureau van Dijk have several technical drawbacks. First, the maximum periods of available information is only for ten years. Therefore, this database is not suitable for long-term analysis and could not capture impact of tax policy changes to firms’ performance assume that tax policy is rarely to change in short-term period. Second, since ORBIS only provides information of ownership status on last available year, therefore, any changes in ownership beforehand cannot be captured. Possibility of misclassification between parent and subsidiary could occur especially in the case of econometric study with panel longitudinal data. 256 However, I ignore these weaknesses since any business restructuring or merger and acquisition activities are assumed likely not frequent; and going into detail misclassification will introduce noise to empirical estimation. The same reason also argued by Budd et al. 2005. 257 In order to find accurate firms that are consistent with the unit and scope of analysis, I made filtering process in ORBIS database as showed on Table 4.1. The periods of analysis are 9 years, starting from 2005 to 2013. Table 4.1 – Filtering Process with ORBIS database Filtering Stages Result 1 Firms located in developing countries 28,250,666 2 Active firms 24,736,722 3 Non-independent firms 2,452,741 4 Foreign shareholder min. 50.01 759,875 5 Firms do not have subsidiaries 717,466 6 Industrial firms non-financial sector 624,280 7 Types of shareholders business types 171,053 8 Availability of financial account 7 - 9 years 28,679 Total first screening 28,679 Note: data as available during 12 -13 March 2015 At the first stage of filtering process, firms should be located in developing countries. Based on World Bank classification of income level and Table 2.1 list of tax haven countries, I include 111 countries on my search. Moreover, the firms should be active which means that they still actively operate and not in liquidation period, bankruptcy, or dissolve. Next, I focus to various filtering process regarding Riedel, “Corporate Taxes and the Location of Intangible Assets within Multinational Firms,” Munich Discussion Papers in Economics, 5294 2008. 255 Simon Loretz and Padraig J. Moore, “Corporate Tax Competition between Firms,” International Tax and Public Finance, Vol. 20, Issue 5 2013: 725-752. 256 Matthias Dischinger, Bodo Knoll and Nadine Riedel, “The Role of Headquarters in Multinational Profit Shifting Strategies,” International Tax and Public Finance, Vol. 21, Issue 2 2014: 253. 257 John W. Budd, Jozef Konings and Matthew J. Slaught, “Wages and International Rent Sharing in Multinational Firms,” The Review of Economics and Statistics, Vol. 87, No. 1 2005: 73-84. 53 to the ownership and subsidiary-parent relationships. One of the advantages of ORBIS database is the availability of independence indicators which can filtrate whether a firm have ownership relationship with other firms or not independent. I exclude all independent firms from my potential observations. The firms should also own by foreigner in majority 50.01 258 and do not have subsidiaries. These stages are to ensure that they are fully controlled by their parent direct shareholder and do not have any possibility to control other firms. I also control the shareholders’ types to ensure that they are not public authorities or governments, individual or families, and employeesmanagersdirectors on particular firm. I also exclude all firms that are not industrial firms, namely: banks, hedge funds, foundation, insurance, public authorities, trustees, venture capital, and others. These types of firms or entities are likely having different treatment of taxation, different objective besides to maximize profit, or different financial accounting format. Finally, in order to eliminate samples with incomplete information, I also entail firms with availability of financial account at least for seven years. Through all the filtering process, I found 28,679 firms in 66 countries for my potential sample. However, this is not the final sample, concerning the possibility that not all firms have complete information for variables in regression analysis. I will discuss this later.

4.5. Estimation Strategy