de.mpg.escidoc.pubman.appbase.FacesBean

Post

 
 Vis
  Elasticiteten af skattepligtig indkomst
Item is

Ophav

 Ophav:
Holme, Ask Michael1, Forfatter
Schjerning, Bertel2, Vejleder
Tilknytninger:
1Det Samfundsvidenskabelige Fakultet, Københavns Universitet, København, Danmark, diskurs:7001              
2Økonomisk Institut, Det Samfundsvidenskabelige Fakultet, Københavns Universitet, København, Danmark, diskurs:7014              
skjul Ophav
Vis Ophav

Indhold

Ukontrollerede emneord: skat, skattepligtig indkomst, elasticitet
 Abstract: Most of the modern economies in Europe and North America use income taxes as the primary source for government funds with quite high marginal tax rates as the result. Therefore the effect coming from income taxation on the behaviour of the consumers/agents is central for tax policy debate seen from both an economical and a political point of view. The size of the behavioural effect -, today almost always measured as the elasticity of taxable income (ETI) - is crucial for determining the final revenue and welfare effects from a tax reform. However there is currently no clear consensus among economists about the size of ETI. Therefore both scientists and government officials have no valid data to rely on when they are calculating tax reform consequences; which is a factor that heightens the estimation uncertainty and makes it harder for politicians to make the correct policy decisions. This thesis explores the question of how to correct estimate ETI, which since Feldstein (1999) proved it a sufficient statistic for calculation of deadweight loss has been the parameter of interest. In the first section the existing literature regarding the ETI estimation is surveyed starting with Gruber og Saez (2002); Saez (1999) which has been the foundation for most of the empirical estimations of ETI. The section describes their theoretical income model and the resulting identification strategy. Subse-quently a number of later additions to this approach are reviewed combined with some of the latest empirical introductions which include ETI estimation using error correction models. Additionally a quite new structural estimation approached introduced by Saez (2010); Chetty et al. (2011) is examined together with its results when used on Danish income data. The section concludes that their currently is no clear consensus regarding the size of ETI with estimates going from being insignificant in some studies to well in excess of one in others. This range of estimates have very different implications for the choice of optimal tax rates. An elasticity around zero will allow policy makers to set extremely high tax rates while an elasticity exceeding 0,18 will place the Danish “top tax’’ above the Laffercurve top-point. It is also concluded that the new error correction based estimation approached completely lacks a theoretical basis for the identification strategy. A fact which makes parameter interpretation extremely difficult. It is therefore recommended that use if this new approach is done with extreme caution until it has been subject to a more critical review. Finally optimistic views are expressed about the structural estimation approach however more knowledge about the performance of this estimator is needed before its results can be seen as robust. In order to conduct a more explicit study of the properties of the different estimators a theoretical model of agents minimizingtheir tax expenditure is developed in the next section. The model is based on a quasi-linear, iso-elastic utility function with a structural elasticity parameter. The agents in the model are exposed to a kinked tax system which gives rise to bunching in the sense that some agents will locate their income exactly at the tax kink. Additionally the model to allows for intertemporal in-come switching, dynamic response to tax reforms and rising inequality in the income distribution. The theoretical model is used to conduct a number of Monte Carlo experiments testing the performance of the different model specifications used in the literature. The results shows that especially new introduced error-correction models and the heavily instrumented models inspired by Weber (2010) returns highly biased estimates when subject to even mild amounts of “structural noise’’. However also the more widely used approach based on Gruber og Saez (2002) suffers from bias when subjected to dynamic response and intertemporal income switching. The section concludes that the overall performance of the error correction based model is unsatisfactory given the quite high bias under quite low amounts of noise combined with extremely large tax changes. The model can therefore not be recommended as basis for for empirical work and estimates from such work should not be seen as valid by politicians and government officials. The structural estimator is also subjected to Monte Carlo experiments but due to certain technical decision in the model setup it is not possible to test its performance under dynamic response and intertemporal income switching. Seen from a theoretical point of view these types of “noise” should not affect the structural estimates as long as data around a tax change is discarded before doing any estimation. However the experiments shows that the structural estimates is extremely sensitive to choice of the "bunching-bandwidth" parameter and that the graphical analysis approach used by Saez (2010); Chetty et al. (2011) to choose this parameter isn’t a plausible approach to guess the correct value. To improve the performance of the approach from Gruber og Saez (2002) an extension is constructed which enables the model to control for dynamic response and most types of intertemporal income switching. This new estimator shows superior performance in the Monte Carlo experiments and is therefore recommended for use in empirical work to remove bias from the traditional estimator. Finally the new extended estimator is used along with the traditional approach to estimate the ETI on Danish income data for the period 1994-2008. The new estimator is used in a version which corrects for dynamic response but not for income switching due to lack of sufficient large tax jumps in the esti-mated period. The estimation points towards an elasticity in wage income around 0,11 and one in taxable income around 0,08 with the latter only being slightly significant at a 5 percent level. These esti-mates is an order of magnitude larger than those found by both this and other studies using the traditional approach from Gruber og Saez (2002). A difference which is consistent with the theoretical expectation since the control for dynamic response should remove a negative bias. This theses concludes that none of the currently available ETI estimators can be seen as robust considering the large amount of “structural noise” found in income data. However the estimates from the new extended estimator can be seen as an upper-bound of the elasticity given that they corrects for the largest negative bias considered. Further research around ETI estimation should therefore focus on gaining a better understanding of the different ways individual behaviour is influenced by income taxation and on improvement of the new structural estimation approach. Improved knowledge regarding individual behaviour will make it possible to make a model concerning this behaviour and thereby construct control variables for use in empirical estimation. Improvement of the structural estimator should focus on gaining a more robust way of determining the correct the “bunching-bandwidth" parameter since its value have large implications for the ETI estimates. Until such improvement is done estimates from the structural approach should be used with caution and rigorous sensitivity analysis should be conducted when used in empirical work.
skjul Indhold
Vis Indhold

Filer

Navn:
specialetilweb.pdf (Hovedtekst)
Bemærkninger:
-
Tilgængelighed:
Offentlig
Mime-type / størrelse:
application/pdf / 846KB
Copyright dato:
2012-06-15
Copyright information:
De fulde rettigheder til dette materiale tilhører forfatteren.
skjul Filer
Vis Filer

Basal

Bogmærk denne post: https://diskurs.kb.dk/item/diskurs:31184:2
 Type: Speciale
Alternativ titel: En kritisk gennemgang af estimationsmetoder
Alternativ titel: The elasticity of taxable income
Alternativ titel: A critical review of the estimation methods
skjul Basal
Vis Basal

Links

Vis Links

Detaljer

Sprog: Danish - dan
 Datoer: 2011-09-11
 Sider: -
 Publiceringsinfo: København : Københavns Universitet
 Indholdsfortegnelse: Indholdsfortegnelse 1
Afsnit 1 Indledning 3
Afsnit 2 Gennemgang af den eksisterende litteratur 4
2.1 De tidlige bidrag 5
2.2 Estimater på paneldata 5
2.3 Kombinering af reformer og brug af instrumenter 7
2.3.1 Den grundlæggende empiriske strategi 8
2.3.2 Endogenitetsproblemet og andre estimationsændringer . 9
2.3.3 Estimation og resultater 11
2.4 Kort- og langsigtede effekter samt andre nyere bidrag 15
2.4.1 En fejlkorrektions specifikation - Holmlund og Söderström (2008) . 15
2.4.2 Mere uformelle lags - Giertz (2008b) og Heim (2009) 17
2.4.3 Fejlkorrektion og individspecifikke effekter - Bækgaard (2010) . 18
2.4.4 Overlappende skatteændringer og stød persistens - Weber (2010) 20
2.5 En strukturel estimationstilgang: Bundtning . 23
2.6 Opsummering . 26
Afsnit 3 Teoretisk model for indkomstændringer 28
3.1 Den grundlæggende model 29
3.2 Knæk i skattesystemet . 29
3.3 Ændringer over tid 31
3.4 Eksogene indkomststød (“Mean reversion”) . 31
3.5 Indkomstforskydning over år . 33
3.6 Begrænsede adfærdsvirkninger 35
3.7 Heterogen indkomstvækst baseret på indkomstniveau 35
Afsnit 4 Monte Carlo forsøg med estimatorerne 37
4.1 Modelspecifikationer . 38
4.2 Parametervalg . 42
4.3 Indledende kørsler . 43
4.4 Forsøg med persistens i stød . 47
4.5 Effekt af begrænset tilpasning . 50
4.6 Effekt af flere på hinanden følgende skatteændringer . 52
4.7 Effekt af indkomstforskydning 54
4.8 Skæv lønvækst . 55
4.9 En udvidet modelspecifikation 58
Afsnit 5 Estimation på danske registerdata 63
5.1 Empirisk strategi . 63
5.2 Datagrundlaget 64
5.2.1 Stikprøvekriterier . 65
5.3 Skattesimulatoren . 66
5.4 Beskrivende statistik . 68
5.5 Estimationsresultater . 70
Afsnit 6 Afsluttende bemærkninger 74
Referencer 77
Appendiks A Udregning af Lafferkurvens toppunkt 81
Appendiks B Figurer 81
Appendiks C Data 84
 Note: -
 Type: Speciale
skjul Detaljer
Vis Detaljer

Kilde

Vis Kilde