More Info + Better Analytics = Less Tax Evasion: Guest Post by Gaute Solheim

There is little doubt that information is key to effective taxation, and that modern data collection systems are making tax administrators' (and in some ways, honest taxpayers') lives much easier.  I am personally wary of governments (or anyone) amassing all of the data gathering and analytic tools they might like, because I am concerned that life in the panopticon will not be tolerable for lots of reasons. My tolerance for leakage in taxpayer compliance to forestall the complete surveillance of all human activity by government may be higher or lower than others; I am not sure. But the following guest post by Gaute Solheim of the Norwegian Tax Administration gives a nice summary of the tax administrator's more optimistic view. Cross posted from the Surly Subgroup.

Do You Really Need More Information?

By Gaute Solheim, Senior Tax Advisor, Norwegian Tax Administration 

(Mr. Solheim writes in his individual capacity and does not purport to represent the views of the Norwegian Tax Administration.)

In a previous post, I explained with reference to the Cui 2017 paper on Third Party Information Reporting (TPIR) why I expect good quality TPIR, based on a primitive analysis of the human factor in corporate filings. When I started rereading Cui’s paper, I only read a few lines before a chapter heading from the CIA textbook “The Psychology of Intelligence Analysis” popped into my mind: Do you really need more information? That book contains a full chapter (starting on page 51) on how more information sometimes contributed nothing to the quality of the analysis, but was of immense help for the confidence of the analyst.

I will below make my argument for why the answer should be “yes” when it concerns the TPIR data mentioned by Cui in the paper, but it is a qualified yes. TPIR is a bit like cooking. Fantastic raw materials will not end up as gourmet meals on their own. You need a talented and skilled chef and a kitchen with the right tools, as well.

Having spent some time on capacity-building with less developed tax administrations than the Norwegian, I agree with Cui that establishing a TPIR machinery should not be the first priority in these countries. However, TPIR being the wrong starting point for some developing countries is not an argument for abstaining from it in Norway and other jurisdictions with better-resourced administrations. I will below state my case for why I find TPIR useful based on my observations working for the Norwegian Tax Administration (NTA) (which may of course differ from the opinions of the NTA itself).

My knowledge of U.S. or Canadian use of TPIR is limited, but there is a very fascinating story told in a paper by Keith Fogg about the work of the IRS agent Joe West. It is a story starting with an audit in the late eighties continuing into the nineties and ending with the IRS getting a summons for credit card information linking US taxpayers to bank accounts in the Caribbean. That information eliminated credit cards as an easy means of taxpayer access to evaded funds.

Norway was quick to copy and paste the IRS methodology. First as specific requested information. Then, a few years later, TPIR routines were in place, giving us digitally all the credit card transactions from financial institutions. The information was used in a large audit project identifying Norwegians with undeclared financial wealth in other jurisdictions. It did not take long before it was common knowledge among taxpayers that this modus operandi did not fly anymore. The value of this TPIR as a tool to uncover hidden wealth and income in foreign jurisdictions diminished quickly, as measured in volume of reassessments.

Warning: Do not use for tax evasion

Collection of TPIR is not without costs, and after some years, the extra revenue from reassessments probably did not cover the total cost of the collection of this dataset. One could then conclude that this was a big heap of worthless gigabytes. To me that would be flawed logic, like concluding that a fence is pointless if you never find anyone on the wrong side of it. TPIR done right is to a large extent automated audits. Taxpayers knowing that they will be selected for audit on specific issues are inclined to stay compliant or shift to another, more cumbersome, modus operandi like diamonds in a toothpaste tube (see here and here). The toothpaste modus operandi had a shorter life span than the credit card scheme, and then we got FATCA and CRS.

Calculating the value of TPIR as a mental fence keeping more taxpayers inside the compliance triangle would be a hard nut to crack. I will not even try, but I would be surprised if it was insignificant. The story starting with IRS agent Joe West auditing Dorchester Industries and owner Frank Wheaton Jr. in the eighties, and ending in Norway with a lot of TPIR data with few reassessments illustrates a life cycle. It started with tax authorities going out and collecting specific third-party information needed for cracking cases with serious tax evasion; it was later streamlined as part of the TPIR system; and it ended up as mental fences for tax compliance. Today, that specific TPIR has other analytical uses for the NTA as well.

Almost all the numbers needed for producing a correct tax return for the majority of personal taxpayers in Norway already exist digitally on some third party computer somewhere. The number of sources is much smaller than the number of taxpayers. Having a system of distributing digital numbers from a small number of sources to everyone, having everyone manually punch the same numbers into different kinds of software on their home computers, and then collect all these numbers—with all their typing mistakes—does not make sense to me. It did not make sense in 2010, either. For most Norwegian taxpayers, it has not made sense the last two decades.

The main use of TPIR in Norway today is to prepopulate individuals’ tax returns, which are then made available for the taxpayer for correction and amendments. All of this ensures that we may handle millions of tax returns untouched by humans, including auditing every single entry in most of them. The NTA can do this by relying on recycling of existing data established by the business sector as part of its recording of business transactions. Businesses do it on computers with software designed to handle these exchanges of information. It means that, as a taxpayer, I may go skiing this weekend instead of sitting indoors punching numbers into some tax program on a sunny winter weekend. I love it both as a taxpayer and from nine to five as a tax auditor.

A new deduction wanted by the Norwegian government illustrates why TPIR is the way to go. The government wants personal expenses for toll roads to be deductible from income. There is a logic behind that, but that’s not my focus. The public toll roads are 100% digital and all payments are linked to an identified person. To avoid double deductions, the gross toll road payment must be reduced by the amount each individual has reclaimed from his employer or deducted as business expenses. A large part of these toll payments sits as digital information on the servers of the employers.

The NTA’s way of thinking is that TPIR is our friend. The payments on cars registered to non-business people may be reported as TPIR from a few toll road operators. The correction for reclaimed or already deducted payments may be one more item in the existing TPIR filing from employers on each employee. We do not know at this stage what percentage of all the tax returns we may prepopulate with the correct deduction by using TPIR, but we are aiming very high. Designing the compliance program for this deduction without TPIR is not a tempting alternative.

Establishing the infrastructure needed for TPIR of a specific set of data is the main cost. The exchange itself is comparably very low cost if the software doing the job is well designed and integrated in the data systems of the reporting entities. Ask any tax manager, and they will tell you that tax auditors asking for information that the Enterprise Resource Planning (ERP) software (which manages business processes and accounting) is already programmed to produce is a small burden compared to collecting ad hoc sets of data that the ERP is not programmed to report.

We could do targeted audits, waiting three to five years and launch an identical audit program, collecting the same dataset ad hoc, and then spend a lot of resources on sorting out all the mistakes and non-compliance uncovered by the audits. It does not make sense in 2019 to operate like this. It is much more cost efficient to do the work needed to design the software and needed infrastructure that ensures that there will be a steady flow of good quality TPIR.

Within the OECD, there is a program called SAF-T, Standard Audit File for Tax. The core of this is that businesses must run their accounting on software designed for exchange of data in a standard digital format. When this is implemented, my guess is that we will see a new drop in cost per gigabyte TPIR, and it will certainly reduce the taxpayer’s cost of handing over data in response to ad hoc requests. Portugal is at the forefront of using SAF-T, with ten years of experience, generating gigabytes of TPIR monthly on VAT transactions and much more (see here and here).

The constant expansion of what tax administrations receive as TPIR should also influence the more rational taxpayers. They now know that you never know what the next dataset will be. Would you engage in a toothpaste scheme if you knew that FATCA was two years away? What tax administration veil of ignorance will be the next to fall with the help of TPIR? It gets harder and harder for the taxpayer to rationalize away the risk of future exposure of presently hidden activity.

I will restate that TPIR is a superb ingredient in the hands of a skilled chef in a well-equipped kitchen, but add that you really need a close partnership with your suppliers to make it happen without prohibitive costs.

Yes, we really need more information.

What Kinds of Corporate Entities Evade Tax? The Human Factor in Third Party Info Reporting

Under what circumstances should we expect corporate entities to evade tax? The following guest post by Gaute Solheim of the Norwegian Tax Administration posits that it is most likely when they are small, owner-operator gigs, and much less likely when they are giant multinationals. The rationale, drawn from insights about why third party reporting is vital in a self-reporting system, is that the bigger the company, the more people have to work cooperatively to engage in fraud, many of whom will be employees with no direct economic upside. Of course there must be exceptions--Enron comes to mind. But Solheim's observation is that the human self-interest that makes third party reporting an effective compliance tool also minimizes the frequency and scope of Enron-type evasion. Cross posted from the Surly Subgroup.

Gaute Solheim, Senior Tax Advisor, Norwegian Tax Administration

(Mr. Solheim writes in his individual capacity and does not purport to represent the views of the Norwegian Tax Administration.)

I happened to reread a paper by Wei Cui from 2017 on Third Party Information Reporting (TPIR) some weeks ago. Based on my own practical experience working for the Norwegian Tax Administration, I found it hard to agree with him on uselessness of TPIR. In this post, I will explain why I expect good quality in TPIR filing. I plan to write a later post on why TPIR is useful.

I will start with crediting Cui for outlining very well what I will call “the story of the corporate tax revenue collection machine.” He does a good job describing how the government has outsourced the bulk of the revenue collection to corporations. I would love to see papers adding empirical numbers to this description. My Norwegian perspective is a world of VAT, corporations withholding taxes on wages and delivering massive amounts of data used to prepopulate the personal tax filings. The way Cui describe the role of the corporations in the “tax revenue collection machine” is probably even more spot on for me than for a reader with a US perspective.

As I read Cui, he expect legal entities to cooperate sufficiently to ensure fairly good quality TPIR. He does some reasoning on why this is to be expected, but I did not find his arguments very convincing. This may be because for some years I have had a different line of reasoning ending with the same conclusion. What follows is based on material I used in a talk some ten years ago for my colleagues at the Large Taxpayer Office in Norway. The theme of that talk was why we found less evasion in our segment of taxpayers than what other tax auditors found when auditing smaller entities. I called the talk “The Human Factor in Tax Filing”. I believe the reasoning I used there also may explain high quality TPIR.

To make it clear: This is about tax evasion (or deliberately misreporting). Tax avoidance is a different beast.

I started out by observing that legal entities cannot write, and hence are unable to fill out their own tax returns. They need human beings to help them. My next observation was that you can split these “helpers” into two groups. Group one is humans with direct economic upside from evasion, typically the owner(s) of the entities. Group two is humans with no direct economic upside from evasion, typically employees without an owner interest. My third observation was that the influence these two groups had on the filings was strongly linked to the size of the entities. In the smallest one-person owner/manager entities, people with direct personal economic upside from evasion (group one) have a large influence on the numbers filed. In the largest multinational corporations, people with no personal direct economic upside from cheating (group two) do most of the work.

Based on these unsystematic observations of reality, I made the following diagram of my estimate of how the influence of people with economic upside from misreporting (red line) and those without (blue line) varies as a function of entity size. The vertical axis goes from zero control to a hundred percent. On the horizontal axis, each entity is counted as one.I do not have empirical studies to back up any claim that the balance of influence between the two groups shift at a specific gross income or similar figure. If forced to make a guesstimate, I would, for Norway, say somewhere around USD 5 million, with big variations among business segments. I will, for the sake of simplicity, use USD 5 million as the crossing point.

Corporations with less than USD 5 million in revenue are by far the main bulk of the population. Based on this perspective, we would logically start wondering why all these entities represented by humans with incentives to cheat average to a compliance rate of 90% or better. This is the perspective I understand Cui is debating in the last half of his paper, and he formulated some explanations that did not fully convince me.

As a longtime tax auditor, the number of entities to me is an administrative problem. I’ve always measured my own work in money saved: How much of the tax gap did I manage to plug? To get this perspective on reality, I must change the unit of measurement on the horizontal axis from entities to dollars. Each entity claims a space equal to the share of the total government revenue collected or paid by that entity.Something very interesting happens to the perspective when we change from entities to dollars: The main bulk of government revenue is collected, reported and paid by the small group of big entities where the people with no upside is in full control over the numbers reported.

I only have hard numbers for corporate income tax (CIT) in Norway to back up this perspective, but I am confident that it holds true in general. Sixty-five percent of all CIT in Norway is paid by entities with revenue higher than approximately USD 150 million. The corporations with billions in revenue and thousands of employees are few, but they represent the bulk of what matters. When we use this perspective—money—we see that individuals with no personal gain from non-compliance handle the main share of input to the revenue collection machine and TPIR. This is true whether the reports are on wages and taxes withheld from them, VAT, or the entities’ own corporate income tax.

There is another important human factor, the down side. I have observed legal entities with feelings only when someone employed by the entity has made a mistake and the mistake has become public. The public relation department will issue a statement telling the public “The entity is very sorry that a trusted person have failed to live up to the high ethical standards of the entity as explicitly laid out in internal memos, rules, etc.” Then they throw that someone under the bus. Everybody knows this.

Since everybody knows this, it would be strange if that knowledge did not shape the behavior of the people engaged in taking care of compliance tasks on behalf of the huge impersonal corporation. They have no personal gain and they know who the scapegoat will be.

Based on this rather simple analysis of the human factors involved, I am convinced that compliance and quality TPIR is what I in general should expect from large corporations. I do know from experience that in some particular settings this will not hold true, but the big picture is to expect compliance. I will now shift to the smaller entities.

Cui gives a few examples illustrating how the employer and the employee have a win-win scenario in misreporting (cheating). He leaves out some elements that I find very practical in real life. I must confess that a few times in my life it has happened that I was guilty of doing something qualifying at least for a fine. But I’ve never been tempted to do something that relied on recruiting a partner in crime. Somehow human nature realises without thinking too hard about it that conspiracy is on a different scale than a moment of weak ethics. If you do not figure it out on your own, the penal code makes that very clear.

The win-win situation described by Cui differs a lot from an owner/manager deciding to keep some income off the books. While you may do that alone as owner/manager, the Cui win-win may only be established by recruiting someone to take part in the scheme. In addition, the owner knows that employers and employees from time to time end up with conflicting interests. Lay-offs are but one example. From my limited knowledge of the US penal system and tax legislation, I have the impression that an employee deciding to inform the IRS about the misdeeds of a corporation may get much more lenient treatment than the corporation and the owner/manager of that corporation. As a hypothetical owner/manager, I would at least think twice.

This is not saying that this kind of collusion does not happen. It does, but the human factor and a simple risk analysis informs me that it should be more the exception than the rule. Undeclared income is a much more likely scenario in the small corporations than manipulated TPIR. This result is not produced by the integrity of the corporation, which I believe has none, but is what you expect from human beings when acting based on the incentives present in the situation. The sum of these very human choices is what we may call a compliant corporation, sending the tax authorities high-quality TPIR.

As Cui correctly observes, even high-quality TPIR is pointless if it does not contribute to something useful. Tax authorities are collectors of revenue, not of data points. Why I do not agree when Cui concludes that TPIR is not useful will be the theme for a later blog post.

And again: This was about evasion, not avoidance.

Canada "Tax Gap" Study Released

The Government of Canada has released its first study of the "tax gap," which the Government defines as "the difference between the tax that would be paid if all obligations were fully met in all instances, and the tax actually paid and collected." The Canada Revenue Agency (CRA) has not completed a study of the income tax, but has released this paper on the concept and methodology. It has presented a report for GST/HST (Canada's VAT), estimating the tax gap to average about 5.6% per year over the period 2000-2014. For 2014, this produced an estimated tax gap of about $4.9 billion:

This study has been undertaken after many calls from academics and nongovernment organizations, including Canadians for Tax Fairness, which according to the CRA will be involved in consultations regarding the ongoing study. Canadians for Tax Fairness estimate that Canada loses $7.8 billion in income tax revenues to "tax haven" use, a number they constructed using Statistics Canada's foreign direct investment data.

The Government acknowledges that there is no reliable method for measuring the tax gap, and that the exercise is one in speculation based on imperfect information:

There are a number of challenges facing tax administrations undertaking tax gap estimation. The key challenge is access to the comprehensive and good-quality data necessary to produce estimates. A significant proportion of the tax gap involves unreported or under-reported income and assets and economic activity that are deliberately hidden from the government. As a result, many countries that publish tax gap estimates highlight their uncertainty.

Expect more to come from this exercise as the tax gap study is a key component of the Government's pledge to spend $444 million over five years "to enhance [CRA] efforts to crack down on tax evasion and combat tax avoidance."

Understanding the Accidental American: Tina's Story

Tax Analysts has published my talk on taxpayer rights and citizenship-based taxation as enforced via FATCA, which I gave in November at the International Conference on Taxpayer Rights in Washington DC, organized by National Taxpayer Advocate Nina Olson. Tax Analysts' content is normally gated but they have made this column available on their free site.

Samaha and Strahilevitz on Information policy and legal design

Adam M. Samaha and Lior Strahilevitz recently posted a paper called Don't Ask, Must Tell — And Other Combinations, which on its face looks like it has nothing to do with tax but it is relevant to questions about compliance and enforcement, so I thought it worth reading. Here is the abstract:

The military’s defunct Don’t Ask, Don’t Tell policy has been studied and debated for decades. Surprisingly, the question of why a legal regime would combine these particular rules for information flow has received little attention. More surprisingly still, legal scholars have provided no systemic account of why law might prohibit or mandate asking and telling. While there is a large literature on disclosure and a fragmented literature on questioning, considering either part of the information dissemination puzzle in isolation has caused scholars to overlook key considerations. 

This Article tackles foundational questions of information policy and legal design, focusing on instances in which asking and telling are either mandated or prohibited by legal rules, legal incentives, or social norms. Although permissive norms for asking and telling seem pervasive in law, the Article shows that each corner solution exists in the American legal system. “Don’t Ask, Don’t Tell,” “Don’t Ask, Must Tell,” “Must Ask, Must Tell,” and “Must Ask, Don’t Tell” each fill a notable regulatory space. 

After cataloguing examples, the Article gives accounts of why law gravitates toward particular combinations of asking and telling rules in various domains, and offers some normative evaluation of these strategies. The Article emphasizes that asking and telling norms sometimes — but only sometimes — are driven by concerns about how people will use the information obtained. Understanding the connection to use norms, in turn, provides guidance for a rapidly advancing future in which big data analytics and expanding surveillance will make old practices of direct question-and-answer less significant, if not obsolete. In any event, the matrix of rule combinations highlighted here can reveal new pathways for reforming our practices of asking and telling in life and law.

The authors cover taxation under the category Must Ask, Must Tell (MAMT). A highlight:

The personal income tax regime is perhaps the most familiar MAMT regime to many Americans. ... Strikingly, because it collects tax information from third parties like employers, banks, and brokerages, IRS already has much of the most important information that a taxpayer will provide on the applicable 1040. This redundancy has sparked reformers to call for replacement of the current, high-transaction costs MAMT regime with one where the government automatically calculates each taxpayer’s liability (or refund) each year and sends her a bill (or check). Notwithstanding the substantial time savings for taxpayers that such plans may entail, these proposals for reform have not been implemented. What gives?

The authors propose that MAMT might be explained by a need to resolve agency problems, which I don't really buy, and then they suggest that making people make tax declarations themselves is a way to make sure they value their citizenship or participate in democracy or make socially good choices, all notions I have heard before but cannot possibly believe when I read that the vast majority of taxpayers pay a tax prep service to help them get through their tax filing every year. Remember, the tax prep service makes money by making it so the taxpayer doesn't have to understand the form, much less the law. The tax preparation industry would definitely find it a hardship if they could not rent-seek off the complexity of tax filing. Remember California's ready return? TurboTax didn't like it.

Rent-seeking by tax compliance professionals, and the ongoing battle to keep the IRS from being able to serve taxpayers properly, are inter-connected key aspects of tax compliance and enforcement. The more hideously complex the law, the more the tax return preparer can charge for the service (I note that paying premiums to overcome tax complexity and attendant risk of error is but one reason why the US practice of treating certain nonresidents as permanent tax residents cannot possibly be fair).

The authors of this paper seem to understand the interplay between complexity and rent-seeking but they dramatically under-emphasize this in the analysis, and that is a pity. This paper barely scratches the surface of the "must ask, must tell" nature of income tax declarations, and I would have liked to have seen more discussion, especially regarding the global scope of the US tax system. But that is a lot to ask of non-tax experts. The paper concludes with a normative discussion that I am still working through, and I'm not sure if there are lessons there for taxation, or not. In any event, a novel paper that raises some interesting points about mandating the furnishing of information.

Leaving money on the table to avoid extra paperwork

Economist Youssef C. Benzarti has recently posted an article on SSRN, How Taxing is Tax Filing? Leaving Money on the Table Because of Compliance Costs, in which he that "taxpayers forego $800 on average to avoid the cost of itemizing."

Can this really be true? It is hard to imagine given that for resident filers, at least, plenty of cheap tax help is available all around, and itemizing is merely a question of gathering receipts and punching in numbers (If the sample includes a lot of US persons permanently living abroad who don't understand their US tax filing obligations, I could understand the choice to bear higher taxes out of a fear of getting things wrong). It's not like most people are sitting around their kitchen tables calculating things out themselves; but even if they are, it's not really THAT hard, is it? In any event, here is the abstract, worth a closer read. If the analysis is correct, it bolsters the case for worrying about complexity as a second level of tax that is allocated without regard for ability to pay.

I use a quasi-experimental design to estimate the burden of complying with the tax code. Employing a sample of US income tax returns, I observe the preferences of taxpayers over itemizing deductions or claiming the standard deduction. Treated taxpayers forego $800 on average to avoid the cost of itemizing. A revealed preference argument implies that itemizing deductions is as painful as working more than 17 hours at one’s regular job. The amount of foregone benefits is larger for richer households, consistent with the fact that the value of time increases with income. I explore two explanations of the magnitude of the estimates. First, it could be due to an extreme aversion to filing taxes. Such aversion implies that itemizing deductions imposes an aggregate compliance cost of 0.24% of GDP and an extrapolation to filing federal taxes implies that the overall cost of compliance is 1.55% of GDP. Second, if taxpayers are time-inconsistent the revealed preference argument fails, introducing a wedge between foregone benefits and compliance costs. Being present-biased leads taxpayers to forego large benefits even when compliance costs are relatively small. I provide evidence of taxpayers being present-biased. Both explanations - whether driven by preferences or mistakes - suggest that the burden imposed on society by tax compliance is significantly larger than previously estimated. I discuss policy implications of the result.