Complexity, Salience, and Income Tax Reporting Behaviour: Evidence from a Natural Experiment

Transcription

1 Kaisa Kotakorpi & Jani-Petri Laamanen Complexity, Salience, and Income Tax Reporting Behaviour: Evidence from a Natural Experiment And Christian Gillitzer and Peer Ebbesen Skov Evidence on Unclaimed Charitable Contributions from the Introduction of Third-Party Information Reporting in Denmark Discussion by: Erzo F.P. Luttmer Dartmouth October

2 Outline 1. Discuss Kotakorpi & Laamanen paper (Finland) Offer suggestions specific to this paper 2. Discuss Gillitzer & Skov paper (Denmark) Offer suggestions specific to this paper 3. What are the wider lessons from both papers? What do we learn about underlying mechanisms? How could we learn more? 2

3 1. Roll-out of partial prefilling in Finland Natural experiment: Roll-out of partially pre-filled tax forms based on (i) observables of tax payers and (i) timing of the adoption by municipalities Experimental groups: Treatment group: Tax payers in early adoption municipalities selected to receive prefilled return Control group: Tax payers in laggard municipalities, who did not receive a prefilled return, but with observable characteristics that would have gotten them a prefilled return in an early adoption municipality. Outcomes: Reporting behavior for: (i) income items vs. deduction items and (ii) prefilled items vs. non-prefilled items 3

4 Great paper! Important topic: evidence on drivers of reporting behavior Great quasi-experimental variation (between 1995 and 1997, the fraction of partially prefilled returns increased from 7% to 57%) Very good data (administrative tax data, long panel) Anything left for a discussant on which to offer constructive comments??? 4

5 Feedback (first try) Perhaps early adopter municipalities differ in unobserved characteristics from laggard municipalities (or treatment and control group don t quite match) Pre-empted by K&L: Inclusion of individual fixed effects (which absorb municipal F.E., except for a couple of movers) 5

6 Feedback (second try) Perhaps individuals in early adopter areas have different trends in reporting than those in laggard areas Because treatment and control group don t quite match on observables and/or because of unobserved differences b/w areas Test the parallel trends assumption E.g., do placebo specifications: are there no differences in changes in reporting behavior between treatment and control in the two years prior to adoption (or after it had been adopted both for treatment and control group). Improve the comparability of treatment and control group Use all observations predicted to receive a partially prefilled return as the treatment group (rather than those actually receiving it). Optional: turn ITT estimates to ATT by using IV. 6

7 Findings 1. Main regression results: Effect of receiving partially prefilled tax form on: Prefilled field Non-prefilled field Any reporting: Avg. amount: Any reporting: Avg. amount: Self-reported Income ~0 ~0 - ~0 Deductions + ~ Further findings The changes in reporting appear to be driven by different reporting for small amounts 7

8 2. Adoption of prefilled char. ded. in Denmark Natural experiment: Starting in 2008, individuals received a tax form in which their charitable deductions were prefilled (3 rd -party reporting of these deductions also started in 2008). Experimental groups: Pre/post 2008: doesn t sound great. But the effect on fraction claiming charitable deductions is so large (doubling!) that it cannot be due to the pre-trend (effectively, it is like doing a visual RD). Outcome: Reporting of charitable deductions on tax form 8

9 Great paper! Important topic: evidence on drivers of reporting behavior No control group, but this not really needed when looking at the fraction reporting charitable deductions Very good data (administrative tax data, long panel) Interesting and creative cuts of the data Behavior by accountants vs. non-accountants Charitable deduction amounts that are strictly dominated Examination of distribution of deduction amounts Differences by primary deficit of individual Anything left for a discussant on which to offer constructive comments??? Actually, no not really (at least not by me; sorry) 9

10 1. Main results: Findings Effect of receiving prefilled tax charitable deductions on: Other line items Charitable Deductions Prefilled field Any reporting: Avg. amount (incl. zeros): N/A ++ ~0 or + Non-prefilled field N/A N/A 2. Further findings: The changes in reporting appear to be driven by different reporting for small amounts Effects are less pronounced for accountants, and only present for deductions under 500 DKK for them. Those who started claiming charitable deductions in 2008, two-thirds claimed deductions in each of the following 3 years, so this concerns regular donors. 2-3% of all donors and about 25% of small donors made charitable contributions in a strictly dominated region. Hint of an uptick in fraction claiming charitable deductions for those with a primary deficit 10

11 3. Lessons from the papers together Easy lessons to draw: 1. Prefilling matters for reporting behavior! 2. Taxpayers leave dollar bills on the sidewalk,.. but primarily bills in small denominations 3. Room for intellectual arbitrage between the papers K&L could look at subgroups (e.g., accountants) with lower decisionmaking costs. (Other such subgroups? Heterogeneity by demographics?) K&L could look at discontinuities around the primary deficit 4. Rich data in K&L allows for further analysis: K&L could examine whether the effects occur for all line items within each of their four cells (e.g., for all non-prefilled deduction line items) (?) Examine effects on difference between prefill and final value 11

12 What explains the totality of the results? To be explained: 1. Effects on reporting: Income 2. Effects concentrated among small amounts (DK and FI) 3. Much smaller, but non-zero, effects among accountants (DK) 4. Reporting of deductions in dominated regions (DK) (Rational) model of line-item by line-item decision-making cost can explain: 1a, 1b, and 2 (record keeping costs outweigh tax benefits for small amounts) 3 and 4 (not worth learning about the rules for small amounts) But how to explain 1d (and 1c, if it is indeed negative)? Deductions Prefilled field [a] 0 (in FI) [b] + (in FI and DK) Non-prefilled field [c] 0 / - (in FI) [d] - - (in FI) Confusion? Do people take an empty (non-prefilled) field to mean it is prefilled at zero? But then why concentrated only for small amounts? Evasion? Do people think that if the item is not prefilled the tax authorities don t know about it? May explain income, but wrong direction for deductions. 12

13 How to explain 1d? (I.e., the negative effect on non-prefilled deductions) Perhaps: A one-time fixed cost for filling out a tax form. When it was not partially prefilled, individuals incur this cost and report most items in each category (perhaps missing just a few) When it is partially prefilled, it is not worth incurring this cost and individuals report fewer items in each non-prefilled category (forgetting many items). For prefilled items, the reported amount goes up because the tax authorities forget nothing, but individuals used to forget a few items even when they incurred the fixed cost. Is the existence of such an overall fixed cost plausible? Further testable implications? No such effects for tax professionals or people using tax preparers In the prefilled category: the fraction of deductions exceeding the prefill amount should decline. If you report a large amount in one non-prefilled category, you incurred the fixed cost and you should then also report small amounts in other non-prefilled categories. Is there a better alternative explanation? 13

14 (end) 14