Abstract
Estimates from studies of income underreporting (the proportion of undeclared to true income) by the self-employed using the ‘Engel curve’ approach of Pissarides and Weber (PW; J Public Econ 39(1):17–32, 1989) have been based predominantly on survey data on incomes and expenditures. This paper uses a unique dataset, from New Zealand, that matches survey data on household incomes with administrative tax register data for the same households. This allows us to measure evasion under different incentives for misreporting—official tax returns and an independent statistical survey—and to quantify the impact of measurement error in survey-reported incomes on underreporting estimates. We find that using tax return data leads to robust estimates of income underreporting by the self-employed of around 20% on average. By contrast, estimates are only around half as large when based on survey data. This result reflects both measurement error in, and attenuation biases arising from, survey-reported incomes. The former appear to account for much of the difference. If self-employed survey reporting in other countries demonstrates similar differences from equivalent tax records—as seems likely a priori—then many previous estimates of self-employment income underreporting based on the PW approach may be biased downwards.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
An example where direct measurement has been attempted is the US Taxpayer Compliance Measurement Programme (TCMP); See Feldman and Slemrod (2007) for discussion.
Below, we use the terms employee/self-employed rather than employment/self-employment. Employees are those who are hired by a third-party. The self-employed are those employed on their own account. They might also receive a shareholder salary, but they still retain discretion over the amount paid (to themselves). The main distinction for our purpose is that self-employment income is not third-party reported and there is discretion over the income declared as opposed to employees where withholding and third-party reporting makes income visible.
The Household Economic Survey in New Zealand is the equivalent of the Living Costs and Food Survey in the UK and the Family Expenditure Survey in the US.
With access to confidential administrative taxpayer data, a number of tax administrations have sought to identify the extent of underreporting of earnings by the self-employed and/or the employed. For the US, for example, the IRS reports underreporting of non-farm incomes by as much as 63% in 2008–2010, while in the UK, HMRC (2016, p.50) report a 2014–2015 tax gap of around 14% from self-employed sources There is no equivalent evidence on the extent of evasion responses for this group in New Zealand using either survey-based or register data.
Some countries’ household income-expenditure surveys use administrative sources for income data. In this case, potential biases from survey-sourced, relative to tax return-sourced, data would not be relevant. Statistics New Zealand moved to this approach in 2018/19 (Statistics New Zealand, 2020). In addition, whereas some studies have used disposable income as the relevant income definition others have used a gross income measure. The HES in New Zealand reports a gross income measure.
Note that this survey-based estimate is much lower than the 63% IRS underreporting estimate quoted above for the US. It is unclear how far this is due to the different methods and data used but, as discussed further below, it could be related to a tendency for survey-based estimates to be biased downwards relative to register-based estimates.
While both these papers attempt to resolve the same question as this paper, i.e. the degree of underreporting by the self-employed, the methodology, the observational unit of analysis and consequently the empirical strategy used are different. Papers using the PW method rely on household-level observations, while these papers address individual-level data.
However, the Paulus (2015a) study is restricted to housing expenditures such as utilities as the PW expenditure category which, in our New Zealand case, are shown to be strongly affected by the ability to report these under business expenses instead of as private household spending. See Paulus (2015a, pp. 15-16) for arguments regarding the Estonia case. Waseem (2019) provides an interesting alternative approach to identifying self-employment tax evasion responses in Pakistan, comparing responses associated with ‘to-zero’ tax rate cuts with responses to similar sized but ‘not-to-zero’ tax cuts.
The income-gap measured by \(\kappa\) captures the deviation of reported income by the self-employed from their ‘true’ income. These deviations from true income could be via legal tax avoidance mechanisms allowed by the tax system, or illegal tax evasion. The share of legal tax avoidance captured in the estimate will therefore depend on the extent to which tax minimisation opportunities are available specifically to the self-employed or when legal avoidance schemes are more readily exploited by them. Underreported income therefore does not necessarily imply ‘underreported tax’.
See Hurst et al. (2014) for a recent empirical application using a similar specification.
Individuals are classified into white-collar if they occupy the positions of managers or supervisors and blue-collar otherwise.
Kukk and Staehr (2017) and Nygard et al. (2018) pursue alternative approaches to estimation of permanent income, based on a survey measure of ‘regular income’ (Kukk and Staehr) or an average of 7 years income data (Nygard et al.). Nygard et al. (2018, p.1905) show that underreporting estimates are larger when this permanent income measure is used compared to estimates based on annual data. We find that traditional instruments for permanent income (education and occupation variables) perform well, but we also include our controls for income stability and volatility in all regressions.
Ideally, we would like to combine as many characteristics as possible to ascertain which combination of characteristics is differentially associated with underreporting. However, the low number of observations for self-employed households means that dividing them into smaller cells for each defining characteristic results in very low observations per cell making results unrepresentative and/or large regression standard errors.
Diaries, on the other hand, suffer from ‘diary fatigue’ which might affect reports if they pose a high burden on respondents, while short period diaries do not deal well with infrequently purchased items. A mixture of both types of technique is used in the HES to record expenditure data—the approach pursued by Statistical Agencies in a number of countries.
Browning and Leth-Petersen (2003), comparing recall and diary recording of expenditure on food at home for the US, suggest that individuals do a ‘remarkably good job’ when recording food at home as opposed to total expenditure.
The coverage ratio of tobacco and alcohol in the Living Costs and Food Survey in the UK (the equivalent of HES in New Zealand) with respect to the National Accounts is 40%.
The coverage ratio in the Living Costs and Food Survey is variable and ranges from 55 to 80%.
Additionally, durables may be used differently between employees and the self-employed as a source of saving or consumption smoothing. For example, the greater volatility of income for the self-employed may encourage greater durable purchases in years of unusually high income.
Paulus (2015a) finds a larger income-gap when comparing self-employment incomes to public sector employees rather than private sector employees.
The types of income and expenditure surveyed in the HES are discussed in more detail in later sub-sections.
For further discussion of the IDI, see Appendix A.
This definition is applied to register income as it is thought to be a more accurate measure of reported income that can be obtained from a recall question.
A concern with this definition, introduced by PW (1989), is that despite having been used as a strategy to measure the importance of the self-employment source, the rule is applied on the proportion of reported self-employment income to total reported income which differs from ‘true’ income. This definition might therefore entail misclassification of households that underreport their self-employed sources heavily and who, using this rule, are classified as employed, thus mis-measuring the benchmark ‘employees’ group. For this reason, we also explore the alternative ‘opportunity’-based definition.
Some sources of taxable income such as dividends and interests are not observable within the IDI. Interests and dividends are subject to withholding taxes, and therefore, there is no obligation to file a tax return as long as they are withheld at the right tax rate.
The main types of income that are non-comparable are investment income, non-taxable benefits and family tax credits; see Appendix B.
With only 3 years of data to construct our income growth variable, we do not use this income data as a permanent income proxy, unlike previous studies that have had access to a longer time-series dimension, or data on ‘regular’ income (Kukk and Staehr 2017).
Summary statistics of expenditure and income are presented in Appendix C. All income and expenditure variables are deflated to the year 2006 using the quarterly CPI. An alternative specification used the food CPI to deflate food expenditure; results are comparable.
We comment further on the assumption of accurate reporting of food expenditure in Sect. 5.
In addition, first stage regression F-statistics and Kleibergen–Paap Wald F-statistics in Appendix Tables C2 and C3 are all strongly supportive of the instruments used: all F-statistics exceed 30.
The US Internal Revenue Service (IRS) in their tax gap reports document that underreporting is concentrated in categories of income with limited information reporting, and underreporting generally decreases across income categories with greater information reporting.
PW (1989) instrument the self-employment dummy variable in order to correct for the misclassification. However, there are no clear instrumental variables to correct for this bias such that weak instruments can introduce a larger bias than not instrumenting.
In the New Zealand (HES) dataset, the reported gender is for the ‘household representative person’ (HRP) who completes the survey. This is typically, but need not be, the primary earner in the household. Clearly, households with two or more adults may contain different genders. Nevertheless, we find that defining gender by that of the HRP there are significant differences between genders in income underreporting.
This age split yielded approximately equal-sized age groups; similar results were obtained when we allowed for alternative age group thresholds.
Further disaggregation into regions is not possible due to small cells. This result (higher underreporting in urban areas) is in contrast to some presumptions that rural areas are more prone to underreporting—for example, because self-employed farming activity dominates rural areas and the personal/business boundary can be hard to monitor. However, unban self-employment occupations—such as taxis, construction, professional services—may provide similar or greater opportunities for underreporting.
See Cabral and Gemmell (2018) for classification details. Where a household receives self-employment income from more than one source, e.g. sole trader and director shareholder; we calculate the primary source of self-employment income and classify the household accordingly.
Of course, \(E_{i}^{\text{S}}\) may also be measured with error but, as noted above, this reduces the efficiency of estimates of \(\beta\) but does not induce bias.
The use of self-reported employment status to classify self-employed may be subject to larger measurement error due to for example the misperception or misunderstanding of the difference between self-employment and third-party employee status. Our second definition of self-employment defined as the availability of a self-employment source is less likely to be subject to measurement error, but not free from it. Failure to report a self-employment income source—for evasion or failure to take reasonable care—would lead to classification errors in this variable. The definition of self-employment used throughout this exercise has relied on the latter. Given that we cannot fully assess the extent of misclassification in this variable, the two alternative specifications of the self-employed variable were used to estimate underreporting for the purpose of robustness.
Consistently, column (7) shows that the equivalent regression of the measurement error on register (‘true’) income yields a negative correlation \((b_{{vY^{\text{R}} }} )\) between the two. The variance ratio in column (5) is estimated as the variance of the error over the total variance.
In an earlier exercise, Bound and Krueger (1991) compared late-1970 s earnings data in the US Current Population Survey (CPS) and Social Security payroll tax records, finding reliability ratios of 0.82 (men) and 0.92 (women). Interestingly, these reliability ratios fell substantially when the income data were assessed in first differences (over 2 years)—to 0.65 (men) and 0.81 (women). Bingley and Martinello (2017), however, find measurement errors for income to be classical in a Danish study (but not for a ‘length of schooling’ variable).
They asked respondents to report their gross annual income in 2009, including earnings such as pension contributions and payments, unemployment insurance payments, cash benefits and other forms of transfer income.
Note that mean reversion is higher in self-employment than employment income; this reduces the bias when income is the explanatory variable.
Note that in this regression, the benchmark variable for measuring underreporting, namely expenditure, is omitted. The fact that self-employed and employed with similar characteristics report similar values of income to the survey while lower to the register simply illustrates the upward bias in the survey. To measure underreporting, however, the relevant measure is relative to expenditure capacity, not simply differences between income reports.
Table 5 column 4 shows that self-employment (log) labour income is 0.079 higher in the survey compared to the register, while the equivalent value for employees is − 0.10. For comparable income, the values are 0.085 and − 0.04, respectively.
References
Aguiar, M., & Hurst, E. (2005). Consumption versus expenditure. Journal of Political Economy, 113(5), 919–948.
Artavanis, N., Morse, A., & Tsoutsoura, M. (2016). Measuring income tax evasion using bank credit: Evidence from Greece. The Quarterly Journal of Economics, 131(2), 739–798.
Bingley, P., & Martinello, A. (2017). Measurement error in income and schooling and the bias of linear estimators. Journal of Labor Economics, 35, 1117–1148.
Bound, J., Brown, C., Duncan, G. J., & Rodgers, W. L. (1994). Evidence on the validity of cross-sectional and longitudinal labor market data. Journal of Labor Economics, 12(3), 345–368.
Bound, J., Brown, C., & Mathiowetz, N. (2001). Measurement error in survey data. In J. Heckman & E. Leamer (Eds.), Handbook of econometrics (Vol. 5, pp. 3705–3830). Amsterdam: North Holland.
Bound, J., & Krueger, A. B. (1991). The extent of measurement error in longitudinal earnings data: Do two wrongs make a right? Journal of Labor Economics, 9, 1–24.
Breusch, T. (2005). Estimating the underground economy using MIMIC models. Working Paper, National University of Australia, Canberra, Australia.
Brewer, M., & O’Dea, C. (2012). Measuring living standards with income and consumption: Evidence from the UK. ISER Working Paper Series, No. 2012–05.
Browning, M., & Leth-Petersen, S. (2003). Imputing consumption from income and wealth information. Economic Journal, 113, F282–F301.
Cabral, A. C. G., & Gemmell, N. (2018) Estimating self-employment income gaps from survey and register data. Evidence from New Zealand. Working papers in public finance, 07/2018. Victoria Business School, Victoria University of Wellington, Wellington, New Zealand.
Cabral, A. C. G., Kotsogiannis, C., & Myles, G. (2019). Self-employment underreporting in Great Britain: How Much and Who? CESifo Economic Studies, 65(1), 84–107.
di Baldini, M., Bosi, P., & Lalla, M. (2009). Tax evasion and misreporting in income tax returns and household income surveys. Politica Economica, 25, 333–348.
Dilnot, A., & Morris, C. N. (1981). What do we know about the black economy? Fiscal Studies, 2(1), 58–73.
Engstrom, P., & Hagen, J. (2017). Income under-reporting among the self-employed. A permanent income approach. European Economic Review, 92, 92–109.
Engström, P., & Holmlund, B. (2009). Tax evasion and self–employment in a high–tax country: Evidence from Sweden. Applied Economics, 41(19), 2419–2430.
Feige, E. L. (2016). Reflections on the meaning and measurement of unobserved economies: What do we really know about the ‘shadow economy’? Journal of Tax Administration, 2, 5–41.
Feinstein, J. S. (1991). An econometric analysis of income tax evasion and its detection. The RAND Journal of Economics, 14–35.
Feldman, N. E., & Slemrod, J. (2007). Estimating tax noncompliance with evidence from unaudited tax returns. Economic Journal, 117(518), 327–352.
Gemmell, N., & Hasseldine, J. (2012). The tax gap: A methodological review. Advances in Taxation, 20, 203–231.
Gemmell, N., & Hasseldine, J. (2014). Taxpayers’ Behavioural responses and measures of tax compliance ‘gaps’: A critique and a new measure. Fiscal Studies, 35(3), 275–296.
Giles, D. E. (1999). Modelling the hidden economy and the tax-gap in New Zealand. Empirical Economics, 24, 71–90.
Gray, P. G. (1955). The memory factor in social surveys. Journal of the American Statistical Association, 50(270), 344–363.
Hurst, E., Li, G., & Pugsley, B. (2014). Are household surveys like tax forms? Evidence from income underreporting of the self–employed. Review of Economics and Statistics, 96(1), 19–33.
ISWGNA. (2006). Estimates of unrecorded economy and national accounts. Retrieved from https://www.oecd.org/ctp/administration/36333066.pdf.
Johansson, E. (2005). An estimate of self-employment income underreporting in Finland. Nordic Journal of Political Economy, 31(1), 99–109.
Kleven, H., Knudsen, J., Kreiner, M. B., Thustrup, C., Pedersen, S., & Saez, E. (2011). Unwilling or unable to cheat? Evidence from a tax audit experiment in Denmark. Econometrica, 79(3), 651–692.
Kreiner, C. T., Lassen, D. D., & Leth-Petersen, S. (2014) Measuring the accuracy of survey responses using administrative register data: Evidence from Denmark. In Improving the measurement of consumer expenditures (pp. 289–307). Chicago: University of Chicago Press.
Kukk, M., Paulus, A., & Staehr, K. (2019). Cheating in Europe: Underreporting of self-employment income in comparative perspective. International Tax and Public Finance. https://doi.org/10.1007/s10797-019-09562-9.
Kukk, M., & Staehr, K. (2014). Income underreporting by households with business income: Evidence from Estonia. Post-Communist Economies, 26, 257–276.
Kukk, M., & Staehr, K. (2017). Identification of households prone to income underreporting: Employments status or reported business income? Public Finance Review, 45, 599–627.
Lyssiotou, P., Pashardes, P., & Stengos, T. (2004). Estimates of the black economy based on consumer demand approaches. Economic Journal, 114(497), 622–640.
Meyer, B. D., & Sullivan, J. X. (2009). Five decades of consumption and income poverty. National Bureau of Economic Research, Paper No. w14827.
Neter, J., & Waksberg, J. (1964). A study of response errors in expenditures data from household interviews. Journal of the American Statistical Association, 59(305), 18–55.
Nygard, O. E., Slemrod, J., & Thoresen, T. O. (2018). Distributional implications of joint tax evasion. Economic Journal, 129, 1894–1923.
Paulus, A. (2015a). Income underreporting based on income-expenditure gaps: Survey vs tax records. University of Essex, Institute for Social and Economic Research. Working Paper No. 2015-15.
Paulus, A. (2015b). Tax evasion and measurement error: An econometric analysis of survey data linked with tax records. University of Essex, Institute for Social and Economic Research. Working Paper No. 2015-10.
Pischke, S. (2007). Lecture notes on measurement error. Unpublished paper, London School of Economics. Available at http://econ.lse.ac.uk/staff/spischke/ec524/Merr_new.pdf.
Pissarides, C. A., & Weber, G. (1989). An expenditure–based estimate of Britain’s black economy. Journal of Public Economics, 39(1), 17–32.
Schuetze, H. J. (2002). Profiles of tax non-compliance among the self-employed in Canada: 1969 to 1992. Canadian Public Policy/Analyse de Politiques (pp. 219–238).
Scott, C., & Amenuvegbe, B. (1991). Recall loss and recall duration: An experimental study in Ghana. Inter-Stat, 4(1), 31–55.
Slemrod, J. (2018) Tax compliance and enforcement. NBER Working Paper No. 24799. Cambridge, MA: National Bureau of Economic Research.
Slemrod, J., & Weber, C. (2012). Evidence of the invisible: Toward a credibility revolution in the empirical analysis of tax evasion and the informal economy. International Tax and Public Finance, 19(1), 25–53.
Statistics New Zealand. (2020). Changes to the household economic survey 2018/19. Available at https://www.stats.govt.nz/methods/changes-to-the-household-economic-survey-201819.
Sudman, S., & Bradburn, N. M. (1973). Effects of time and memory factors on response in surveys. Journal of the American Statistical Association, 68(344), 805–815.
Torgler, B., Schaffner, M., & Macintyre, A. (2007) Tax compliance, tax morale, and governance quality. International Studies Program Working Paper, Andrew Young School of Policy Studies, Georgia State University.
Waseem, M. (2019). Does cutting then tax rate to zero induce behavior different from other tax cuts? Evidence from Pakistan. Review of Economics and Statistics, forthcoming. Available at https://doi.org/10.1162/rest_a_00832.
Acknowledgements
We thank participants at the International Institute of Public Finance Annual Conference, Glasgow, August 2019, and especially our discussant, Amin Mawani (York University, Toronto) and Mazhar Waseem (University of Manchester, UK) for helpful comments on an earlier draft. We are also grateful for helpful comments from two anonymous referees, Matt Benge, Sean Comber and John Creedy on previous versions of the paper.
Disclaimer
The results presented in this study are the work of the author, not Statistics NZ; they are not official statistics. They have been created for research purposes from the Integrated Data Infrastructure (IDI), managed by Statistics New Zealand. The opinions, findings, recommendations and conclusions expressed in this paper are those of the authors, not Statistics NZ, Inland Revenue nor the OECD or its Member Countries. Access to the anonymised data used in this study was provided by Statistics NZ under the security and confidentiality provisions of the Statistics Act 1975. Only people authorised by the Statistics Act 1975 are allowed to see data about a particular person, household, business, or organisation, and the results in this paper have been confidentialised to protect these groups from identification and to keep their data safe. Careful consideration has been given to the privacy, security and confidentiality issues associated with using administrative and survey data in the IDI. Further detail can be found in the Privacy impact assessment for the Integrated Data Infrastructure available from www.stats.govt.nz. The matching of different data sources on the IDI spine is done by Statistics NZ. These datasets are anonymised thereafter and made available to researchers. Further information on the IDI is provided in Appendix A. The results are based in part on tax data supplied by Inland Revenue to Statistics NZ under the Tax Administration Act 1994. This tax data must be used only for statistical purposes, and no individual information may be published or disclosed in any other form, or provided to Inland Revenue for administrative or regulatory purposes. Any person who has had access to the unit record data has certified that they have been shown, have read, and have understood section 81 of the Tax Administration Act 1994, which relates to secrecy. Any discussion of data limitations or weaknesses is in the context of using the IDI for statistical purposes and is not related to the data’s ability to support Inland Revenue’s core operational requirements.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Cabral, A.C.G., Gemmell, N. & Alinaghi, N. Are survey-based self-employment income underreporting estimates biased? New evidence from matched register and survey data. Int Tax Public Finance 28, 284–322 (2021). https://doi.org/10.1007/s10797-020-09611-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10797-020-09611-8