TESTING THE INCOME POOLING HYPOTHESIS AND ITS LINK TO THE TAXATION OF COUPLE HOUSEHOLDS: EVIDENCE FROM DEMAND SYSTEM ESTIMATION FOR GERMANY

Whether couples pool their resources and behave like a unit or spend their income individually is crucial for social and tax policy. In this paper, I provide a test of the income pooling hypothesis using administrative cross-sectional survey data on expenditures and individual incomes of couple households in Germany. The test is performed within the quadratic almost ideal demand system framework, which allows for an endogenous expenditure budget and endogenous individual income contribution shares in an instrumental variables approach. Although perfect income pooling is broadly rejected, there are significant differences regarding the marital status, the presence of at least one child in the household and whether the household is located in a former West or East German federal state. Married and unmarried couples with children are closer to the acceptance of the hypothesis than couples without children. The approach allows to calculate justifiable differentials of the marginal tax rates within the household if income pooling is rejected.


Notes
also uses the QUAIDS to test the income pooling hypothesis. The indirect 1.
utility function of PIGLOG preferences can be found for instance in Poi (2012).
(1999) estimator, is used in this paper to estimate the QUAIDS model.
A collective household labor supply model that also incorporates consumer demand is 3.
examined theoretically for example in Blundell et al. (2005) and in Cherchye et al. (2012), who extend the former and apply it empirically. Applications within the QUAIDS framework that focus on environmentally relevant consumer goods can be found in West & Williams (2004) and Beznoska (2014).
Maitra & Ray (2006) also use the highest level of education to instrument the incomes of the 4. household members to evaluate household expenditure allocation in South Africa.
See Appendix A for detailed results of the estimated Heckman model and the distributions of 5.
I use the prices for men's trousers and women's trousers as proxies for men's clothing and 6.
women's clothing, respectively. Revenue statistics for 2013 from data provider Statista suggest that expenditures for trousers are the most relevant expenditure group within clothing expenditures for men and women. For the price of men's footwear, I use the price category "classic or casual shoes for men ". The respective price for women's footwear is called "pumps Testing the Income Pooling Hypothesis and its Link to the Taxation of Couple Households: Evidence from Demand System Estimation for Germany https://science-direct.net/testing-the-income-pooling-hypothesis-and-its-link-to-the-taxation-of-couple-households-evidence-from-demand-system-estimation-for-germany/ Published by ElSevier BV Copyright © 2022 (PDF Created by Saudi MEP) Page: 3 or casual shoes for women ".
The problem should therefore be rather minor, which is also confirmed by comparing the 7.
results without correction with the estimator for censored demand systems proposed by Shonkwiler & Yen (1999) in Appendix D. This also adds confidence to the assumption that the demand for clothing and footwear behaves more or less like other non-durable consumption, even though it can be seen as a semi-durable good.
The expenditures for clothing and footwear may depend on the occupational status in case of 8. specific jobs. Therefore, the sample is also restricted on households with a more similar income share of both partners limiting the variation from couple households with larger income differences and heterogenous jobs.
See Appendix B for the complete estimation results. 9.
The formulas for the elasticities in the QUAIDS can be derived according to Banks et al. (1997)  See Appendix C for the complete estimation results. 11. Symmetry cannot be rejected at the 10 percent level in this model and is therefore imposed. 12.
A different first-stage SUR model is only estimated if the control variables are changed. For 13. example, the dummies for the federal states had to be excluded in the interaction model with a dummy for East Germany. Therefore, this was also done in the first stage to have an equivalent specification.
The interaction models could also be estimated with sub-sample analyses. However, the 14.
coefficients do not vary much.
This holds under the assumptions that savings and durable consumption are separable and the 15.
derivative of the average tax rate with respect to income is zero. The latter is obviously not true with progressive marginal tax rates, but the effect is expected to be neglectable and therefore left out for simplification.
This perspective could alter if family policy is concerned, as expenditures for children are 16.