Een aangepaste versie verscheen in het NRC van 20 Juni 2012

The media storm

The ESRI is both a think tank and a research institute. Reports that are deemed relevant to the public debate are released to the press and public. Academic work, including early versions in the form of working papers, are put on the web and announced through dedicated lists such as IDEAS/RePEc.

That is the theory. In practice, there is one journalist who keeps track of the ESRI’s academic output and she occasionally writes a piece — small articles somewhere deep in the Sunday Times. She faithfully represents both the research and its early stage.

These pieces are widely ignored. Last Sunday’s, however, was picked by a journalist of the Independent. He tried to talk to me, but I was on holiday. So, a derivative piece was written. The working paper that was a proof of concept became a research report with a policy conclusion.

The ESRI should have explained that the paper was a working paper, trying to establish a method to answer an important question rather than answering the question.

Instead, the ESRI withdrew the paper because of “methodological flaws” (that remain hidden from me) and claimed that the paper was revised (which it is not). The paper is being revised, but the previous version stands until it is replaced, which will not happen before I fully understand the additional results.

Journalists then started asking questions about relationships and political pressure.

The background

Why am I involved in this in the first place? My research combines strategy with opportunism. The strategic part of the research is about projecting future emissions and resource use, and predicting the impact of energy and environmental policy. For instance, we wanted to know the effect of the carbon tax across the income distribution, but found that there was no baseline for either indirect taxes and income distribution or price inflation and income distribution. The latter paper never got published, because referees objected (rightly) to our focus on price changes while ignoring income changes, at a time when large portions of the population experienced downward social mobility.

So, strategically, we needed a model that would tell us what happens to your consumption pattern (and hence your pressure on the environment) when you lose your job. Opportunistically, the first application of this model was to the pecuniary incentives to work.

Previous research on incentives to work only compared benefits to earnings. (This is not trivial because both benefits and taxes are fiendishly complicated.) Outside of Ireland, there is some research on childcare and work incentives, but none on other costs of working. This is wrong. Working is expensive: Besides childcare, there are commuting, convenience food, cleaning services, representative clothing, and grooming (for those outside academia). The costs of working are substantial, and can affect the pecuniary returns to employment.

So, our purpose was to write a paper for the international literature in economics.

There are no immediate policy implications. You might be content with the pecuniary rewards to working as they are. You might argue for lower benefits, for lower taxes for the low paid, for X-taxes, for subsidized child care, or for a shake up of the markets that keep the prices of food, energy and transport so high.

The research

The working paper is here. There are two steps to the research.

First, we need to estimate how being unemployed affects the pattern of consumption. The Household Budget Survey (HBS) has the best data on that for Ireland. The new HBS will be there any day now, so we used the previous one from 2005. This is unfortunate for a policy study, but fine for an academic proof of concept.

So, you want to estimate a model that gives expenditure on a group of goods and services as a function of all those things that might explain such expenditure as well as unemployment.

There are some things that you need to consider, such as whether to include pensioners. There are methodological issues with censoring of data, infrequent purchase, and selection. You need to think about the functional form: Is being employed a fixed cost, or is it a cost that varies with, say, income?

The paper has all the decisions that we made. You can quibble with each of them.

The HBS has one major problem for this application: The number of children living with their parent(s) is known, but not their age. There is a step function at 4: Child-care is very expensive for young children, and fairly cheap for older children. The results for the cost of childcare are thus an average of those with younger and older kids. In Dublin, childcare is about 1000 euro per child per month. Three kids in quick succession, with both parents working, would come to a peak childcare expenditure of 36,000 euro per year. (Oddly, university fees attract more discussion.)

Michael Collins published a paper on the cost of working (after we did, as far as I know) based on data that are much richer but less representative.

In the second part of the research, you need to compare the cost of working to the income difference due to working. For this, we used the Statistics on Income and Living Standards (SILC) data, which is the Irish and European standard. The data are the actuals, so we fitted a model to estimate the counterfactuals (what would your income have been, had you been unemployed) and the difference.

Again, we can quibble about the model specification.

We could have used the income data in the HBS, as that would avoid the complications of comparing data from one representative sample of the Irish population to data from another representative sample. Because there are grumblings about the quality of the income data in the HBS, we did not. Perhaps, for a proof of concept paper, we should have.