Misguided criticism

The "discrepancy" in the unemployment statistics showed up in at least three different places in the GWU's daily newspaper during the week. Its editorial expressed consternation about the "gross discrepancy" in the country's jobless statistics, even though the leader writer seemed perfectly aware of at least one reason why our two national measures produce different results. The piece then went on to express concern about those whose unemployment shows up in one measure but not in the other.

Back when the Labour Force Survey first came out, with its broader definition of unemployment compared to what we were used to in the ETC statistics, the Opposition's media made it sound as if unemployment had "risen" from the lower ETC number to the higher LFS number. Now they show a grudging recognition that the two methods are different, but they also keep trying to squeeze some mileage from the higher LFS findings and from the "discrepancy" between the two.

The editorial lamented: "Within an area of a few miles, surrounded by the sea, and with a population of almost 400,000, it is hard to imagine that there can be such crass discrepancies in the administration of the same government."

Other countries offer unemployment statistics that are far more intricate than ours. The Americans publish six different unemployment rates, though the method of gathering data is uniform. Three of them restrict coverage to those who are actually looking for work, while another two include also those who are currently not seeking work.

Within this last group, they make a distinction between "discouraged workers" and the broader "marginally attached" workers, and they publish unemployment rates that are progressively broader in coverage. Their widest definition includes with unemployed also those part timers who would rather be full-timers but are prevented by economic conditions.

We are not the only country with dual methods of measuring unemployment, and in common with the other countries, our survey method differs in its findings from the administrative-records method.

Taking a cheap stab at the ETC, the editorial went on: "We understand that it is in the ETC's interest to count only those who register with it... while the LFS includes also those who for some reason do not register... with it. So why shouldn't we add the 4,000 to the list of those who are unemployed and are also looking for the work that they deserve?... The probability is that they prefer to try hard to find employment, instead of going to the trouble of registering. And is this not a sign of a lack of trust in the operations of the ETC?"

Those searching for work but choosing not to register deserve attention, and the paper should devote some of its investigative resources to find out more about, for example, any workers made redundant by restructuring or foreign competition, who are looking for work but stay away from the ETC, foregoing the financial and other benefits that registrants are eligible for.

The LFS, which came out last week, indicated an increase of 996 in total unemployment between June 2002 and June 2003. The LFS's unemployment rate edged upwards from 6.9 per cent in June 2002 to 7.5 per cent a year later. The ETC reported that June 2003 unemployment was up by 117 from the previous June. Later ETC numbers showed a substantial increase in July, when school leavers coincided with other factors, but the figures stabilised in subsequent months.

The two methods

The most substantial difference is that the ETC draws on its own full records, while the LFS uses a sampling method. The ETC's jobless figures are based on its records of unemployment registrants, and the ETC's employment figures are based on reports from all employers on their hiring and firing. For every one person that the ETC records as employed or unemployed, the ETC has a corresponding name and identification number.

On the other hand, LFS samples a few thousand households. The random selection of respondents is designed to keep the luck of the draw from skewing the sample results, which are then expanded into conclusions about the whole population. A predictable but inevitable error creeps into the extrapolation of the sample findings to the population.

On top of the inherent fluctuations in the labour markets and the inevitable measurement errors, there is the LFS's sampling error. Combined with the definitional difference, these produced a gap between the ETC and the LFS. The gap dropped from 4,201 in June 2001 to 3,928 in June 2002 and went up to 4,807 a year later.

The ETC's administrative-records method cannot be error-free and at certain points in time a revision is often necessary; a cleaning-up process intended to minimise errors that could creep in over time. Measurement of unemployment registrants is an in-house task for the ETC, and information on public sector payrolls is readily available.

On the other hand, notification from other employers on their recruitment and terminations may be slow in arriving. The ETC waits a prudent length of time to allow tardy reports to reach it, which is why it publishes the unemployment figures sooner than the employment figures.

This brings up the second major difference in the unemployment figures. The ETC's unemployed all register with the ETC. On the other hand, the Labour Force Survey reports all self-declared unemployed. In line with international standards, the surveyor asks questions relating to employment status. If the respondents are unemployed, actively seeking work and currently available for work, they are classified as unemployed, whether or not they are registering for work.

The international standards that call for labour market surveys rather than registration methods are understandable. Countries have different rules governing unemployment registration and related matter of eligibility for unemployment benefits or insurance. In some countries, counting only registrants will not do at all because there are limits on the length of time the unemployed person can stay on the register, and there may be other restrictions for eligibility.

In our case, the registration method is a good measure of registrants. It is equally true that the act of registration and the requirements of registration follow a number of steps that are readily monitored, while the surveyor relies on the declarations of the respondents.

One advantage of the sampling survey method is that it can track discouraged workers, those who are no longer looking for work. This is the information that the unemployment registration cannot monitor. A sample survey is the instrument for this purpose. By definition, discouraged workers are not currently looking for work and therefore, in the LFS, they are classified with the inactive, not as unemployed.

These are not the ones who are of concern to the GWU's paper. Their point of contention has to do with those who are actively looking for work but are not registering with the ETC. This is a matter for technical analysis, not a political bone of contention. Malta has made big strides in its statistical system, and it was probably inevitable that the progress would tempt some political opportunism. The temptation should be resisted.

The irony in all this is that had the Labour Party been in government, Malta would not be meeting Eurostat standards and probably there would be no Labour Force Survey. There would also be no gap between the LFS and the ETC unemployment data. Such a world would presumably meet with the approval of l-Orizzont.

It would contribute to a mature public debate if l-Orizzont gives up this game of trying to ridicule two sets of different results. What we need is a sensible assessment and not shallow arguments that smack of misguided political opportunism.

www.josefbonnici.com