Corporate taxes and the location of intellectual property

Transcription

1 Corporate taxes and the location of intellectual property Rachel Griffith Helen Miller Martin O Connell

2 Why are Intellectual Property decisions interesting? Intangible capital accounts for growing share of inputs The income from IP is highly mobile Intellectual property has been in policy spot light

3 Why are Intellectual Property decisions interesting? Intangible capital accounts for growing share of inputs since the mid 1990s UK investment in intangible (knowledge) assets has been greater than that in fixed capital OECD described growing significance intellectual property & its simultaneous use by many different parts of a firm as one of the most timportant tcommercial ldevelopments in recent tdecades. d The income from IP is highly mobile Intellectual property has been in policy spot light

4 Why are Intellectual Property decisions interesting? Intangible capital accounts for growing share of inputs The income from IP is highly mobile Firms can and do separate income from real activity offshore holdings of IP can be used to reduce tax A tax lawyer quoted in the New York Times noted: most of the assets that are going to be reallocated as part of a global repositioning are intellectual property that is where most of the profit is. Intellectual property has been in policy spot light

5 Why are Intellectual Property decisions interesting? Intangible capital accounts for growing share of inputs The income from IP is highly mobile Intellectual property has been in policy spot light Many governments want to encourage production and use of intangible capital to boost productivity; tax policy is one instrument Mobility of IP income raises questions over governments ability to tax it treatment of foreign earnings of multinational firms (CFC rules) Recent introduction of Patent Boxes which sharply reduce the rate of corp. tax applied to patent income

6 Taxation of foreign earnings of multinationals Credit vs Exemption system for taxation of foreign source income Credit: tax worldwide income with credit for foreign tax (UK until 09) Exemption: foreign source income can be remitted tax free (UK now) Opportunities to avoid tax in both anti avoidance rules encompassed in Controlled Foreign Company (CFC) rules Income in low tax country can be additionally taxed in home country We consider how CFC rules affect the location of patents

7 "Patent Box" Belgium, Netherlands and Luxembourg currently operate Patent Box income from patents is taxed at substantially lower rate Belgium 6.8% (full rate, 34%); Netherlands 10% (full rate, 25%); Luxembourg 5.9% (full rate, 39%) UK propose to introduce one in 2013 reduce tax rate on the income from patents from 28% (24%) to 10% HM Treasury estimate revenue cost at 1.3bn

8 Location and taxes Multinational Headquarter location Treatment of foreign source income Corporate income tax CFC regime Location of innovation Corporate income tax R&D tax credits Location of IPR holdings Corporate income tax Patent box Royalty treatment (Withholding rates) Location of production Corporate income tax

9 Location and taxes Multinational Headquarter location Treatment of foreign source income Corporate income tax CFC regime Location of innovation Corporate income tax R&D tax credits Patents associated with number of countries: where e technology ogy was created where legal protection sought which patent office patent filed at where company that legally holds the patent is based Location of IPR holdings Corporate income tax Patent box Royalty treatment (Withholding rates) Location of production Corporate income tax

10 Our contribution Model the impact of taxes on the location of patent holdings Allow firms to have heterogeneous responses to taxes Simulate the effect of Patent Box on the location of patents, under different assumptions about CFC rules

11 Model of firm behaviour Firm has a successful discovery, decides which subsidiary should apply for the patent the location of the subsidiary determines how the income will be taxed Discrete choice model; firms choose location where value is highest Value to firm of holding patent in a location depends on: revenue from patent Costs of holding patent in that subsidiary: those related to earning income there, (taxes) fixed costs of locating any benefits that arise from holding patent in that location

12 Expected value to firm of a patent i: firm t: year p: patent application j: location V: value c: variable cost PQ: revenue F: fixed cost

13 Empirical specification & heterogeneity Unobserved heterogeneity random coefficient allows heterogeneity in response to tax across firms Observed heterogeneity 6i industry-size i categories - estimate t 6 separate regressions allow coefficients (and distribution) to vary across industry and firm size

14 Heterogeneity in firm behaviour We allow for heterogeneity in firm behaviour along three important dimensions: 1. we allow all coefficients to vary across three broad industry classifications 2. we allow all coefficients to vary with the size of the firm 3. we allow the income earned by a patent to vary with unobserved patent (firm) characteristics through including a random coefficient the distribution of the random coefficients varies across industry and firm size

15 Data: Firms and Patents Location of Intellectual Property data on Patent applications PATSTAT: EPO patent applications; address of subsidiary that made application measures the location of the IPR Matched to multinational firm ownership structure Amadeus and other sources

16 Patent example: Swedish company Astra Zeneca AB holds application Subsidiary of UK MNE Astra Zeneca Invented in multiple countries (income split from activity)

17 Data: Firms and Patents Location of Intellectual Property data on Patent applications PATSTAT: EPO patent applications; address of subsidiary that made application measures the location of the IPR Matched to multinational firm ownership structure Amadeus and other sources Result: European parent firms and their patent applications held in European and US subsidiaries Firms headquartered in 13 European countries Belgium, Denmark, Finland, France, Ireland, Italy, Luxembourg, Netherlands, Norway, Spain, Sweden, Switzerland and UK Choose from 15 locations above plus Germany, and US

18 Data: Firms and Patents Data on 639 firms with 233,471 patent applications filed Assign to 3 industries; within this spilt by firm size Large Medium Chemicals Pt Patentst 81,268 23, Firms Electrical Patents 93,473 93,473 Firms Engineering Patents 64,811 64,811 Firms

19 Taxes Statutory corporate rate in source country vary across countries and over time (Table 3) Important interactions between jurisdictions : CFC regime CFC regimes, (roughly) apply when: (Table 4 gives details) a great enough proportion of the subsidiaries income arises from passive sources the subsidiary is located in a country deemed to be a low tax country Use home country tax rate for pairs of countries where CFC regime at home and source defined as low tax Patent Boxes rates used in simulations

20 Table 6: Location of Intellectual Property in the Chemical industry Dependent variable=1 if patent t application in location Large firms Medium firms (1) (2) (3) (4) (5) (6) logit logit RC logit logit RC tax rate ( ) (0.04) 04) (0.09) 09)** (0.14) 14)** (0.08) 08)** (0.18) 18)** (0.22) 22)** Variance of ( ) (3.43)** - (3.17)** country effects yes yes yes yes firms patent applications 81,268 81,268 81,268 23,307 23,307 23,307

21 Table 6: Location of Intellectual Property in the Chemical industry Dependent variable=1 if patent t application in location Large firms Medium firms (1) (2) (3) (4) (5) (6) logit logit RC logit logit RC tax rate ( ) (0.04) 04) (0.09) 09)** (0.14) 14)** (0.08) 08)** (0.18) 18)** (0.22) 22)** Variance of ( ) (3.43)** - (3.17)** country effects yes yes yes yes firms patent applications 81,268 81,268 81,268 23,307 23,307 23,307

22 Table 7: Location of Intellectual Property in the Electrical industry Dependent variable=1 if patent t application in location Large firms Medium firms (1) (2) (3) (4) (5) (6) logit logit RC logit logit RC tax rate ( ) (0.04) 04)** (0.09) 09)** (0.12) 12)** (.08) 08)** (0.19) 19)** (0.27) 27)** Variance of ( ) (2.21)** - (3.46)** firms patent applications 93,473 93,473 93,473 22,335 22,335 22,335

23 Table 8: Location of Intellectual Property in the Engineering industry Dependent variable=1 if patent t application in location Large firms Medium firms (1) (2) (3) (4) (5) (6) logit logit RC logit logit RC tax rate ( ) (0.05) 05)** (0.11) 11)** (0.13) 13)** (0.07) 07)* (0.16) 16)** (0.21) 21)** Variance of ( ) (2.15)** - (3.27)** firms patent applications 64,811 64,811 64,811 26,063 26,063 26,063

24 Elasticities From these coefficient estimates we can construct market own and cross tax-elasticities Aggregated over patent specific elasticities (therefore over firms, industries etc.)

25 Table 9 market elasticities Country changing tax rate Location country Belgiu um Denm mark Finlan nd e Franc any Germ Irelan nd Italy Luxem mbourg Nethe erlands ay Norw Spain Swede en erland Switz UK US Belgium Denmark Finland France Germany Ireland Italy Luxembourg Netherlands Norway Spain Sweden Switzerland UK Institute US for Fiscal Studies

26 Table 9 market elasticities (subset of countries) Country changing tax rate Location country Belgium France Ireland Luxembou urg Netherlan nds Sweden UK Belgium France Ireland Luxembourg Netherlands Sweden UK

27 Table 9 market elasticities; Ignore CFC Location country Country changing tax rate Belgium France Ireland Luxembou urg Netherlan nds Ireland looks more price elastic when don t account for CFC Sweden UK Belgium France Ireland Luxembourg Netherlands Sweden UK

28 Table 9 market elasticities; standard logit model Country changing tax rate Less elastic demand; unrealistic substitution patterns Location country Belgium France Ireland Luxembou urg Netherlan nds Sweden UK Belgium France Ireland Luxembourg Netherlands Sweden UK

29 Elasticities in random coefficient model standard logit: cross tax elasticities are the same across countries (within each column) cross-tax elasticity is a function of only the tax rate in the country that is adjusting its tax rate, the share of that country and the coefficient i on the tax rate, which h is constant t across countries random coefficients model: cross tax elasticities vary across countries they depend on the countries' characteristics and how close they are to each other; countries that have more similar characteristics will be seen as closer substitutes t by firms, and therefore the cross tax elasticity it will be higher

30 Summary of estimation results Firms respond to taxes when choosing where to hold IPR Allowing for observed and unobserved heterogeneity in responsiveness is important gives more realistic substitution patterns than standard multinomial logit model Controlling for unobserved country effects important the attractiveness of a country is positively correlated with tax, failing to control lfor this makes it look klike tax doesn t matter

31 Counterfactual policy analysis We use estimated elasticities to simulate potential policy reforms Table 10 introducing Patent Boxes Benelux countries: Belgium (6.8%) Netherlands (10%) Luxembourg (5.9%) UK (10%) consider what if CFC regimes capture patent boxes Include UK removing CFC regime for patent income first consider impact on location of income from patents then consider impact on tax revenue

32 Initial shares (Table 10) Predicte ed share of patent applica ations 30% 25% 20% 15% 10% 5% 0% No Patent Boxes Benelux Patent Boxes UK Patent Box

33 Benelux Patent Boxes Increase share in Benelux Predicte ed share of patent applica ations 30% 25% 20% 15% 10% 5% 0% countries Fall in UK share (12% -8%), and elsewhere No Patent Boxes Benelux Patent Boxes UK Patent Box

34 UK Patent Box F ll i B l t i h Predicte ed share of patent applica ations 30% 25% 20% 15% 10% 5% 0% Fall in Benelux countries share Increase in UK share (to 17%) No Patent Boxes Benelux Patent Boxes UK Patent Box

35 Add CFC interactions Dampened effect Predict ted share of patent applic cations 30% 25% 20% 15% 10% 5% 0% Smaller gains for Benelux Frances loses less No Patent Boxes Benelux Patent Boxes UK Patent Box Remove UK CFC regime

36 UK loses less when Benelux PB UK shares But gains less when UK PB 18% Predicted shar re of patent applications in UK 16% 14% 12% 10% 8% 6% 4% 2% 0% No Patent Boxes Ignore CFC CFC Ignore CFC CFC no UK CFC Benelux Patent UK Patent Box Boxes

37 UK shares Smaller gain in UK share when no CFC regime 18% Predicted shar re of patent applications in UK 16% 14% 12% 10% 8% 6% 4% 2% 0% No Patent Boxes Ignore CFC CFC Ignore CFC CFC no UK CFC Benelux Patent UK Patent Box Boxes

38 Remove UK CFC regime Predict ted share of patent applic cations 30% 25% 20% 15% 10% 5% UK share higher than initially but lower than when the UK operates CFC regime 0% No Patent Boxes Benelux Patent Boxes UK Patent Box Remove UK CFC regime

39 Table 11: Tax revenue government revenue from patent income, function of the share of patent income held in a country and the relevant tax rate. For illustration, assume that CFC regimes continue to operate based on the statutory tax rates Expect fall in revenue in non-patent Box countries In others depends on whether increase in share of income outweighs lower tax rate applied to income

40 Table 11: Tax revenue enue, indexe ed to inital re evenue vernment rev Gov No Patent Boxes Benelux Patent Boxes UK Patent Box

41 Tax revenues What if other countries followed suit and also introduced Patent Boxes? We simulate the effect of France of Sweden introducing a 10% Patent Box UK share of patent income falls further and with it govt. revenues

42 Extensions Interactions between governments in tax setting we showed indicative evidence that tax competition could erode any benefits model government behaviour; write down government s maximisation problem and find optimal policy More heterogeneity in the model? we assume fixed benefits and costs of choosing a location are common across firms likely to depend on (unobserved) firm characteristics allow more random coefficients

43 General conclusions Important to account for heterogeneity in firm location choice industry size variation in responses to taxes with unobserved characteristics Tax does affect location of patent holding Patent boxes attract patent holdings but lead to reductions in revenue