Examination of Witnesses (Questions 240
- 259)
TUESDAY 20 NOVEMBER 2007
Ms Rebecca Riley, Mr Simon Kirby and Dr James Mitchell
Q240 Lord Layard:
The original number here, the 3.1%, assumes no response of capital?
Ms Riley: Yes.
Q241 Lord Layard:
Do you have any estimate of how capital might have responded?
Ms Riley: We have some empirical evidence on
how the capital stock might have changed given A8 migration. This
migration is a larger segment of the population and it is over
a longer period of time.[6]
We do not have estimates of how capital has changed. On the other
hand, it would be reasonable to assume that, if migrants account
for 4.3% of employment, some of the capital that has accumulated
in the last six years or so is also being used by these migrants
and that it is there because the population is there. The migrants
are there, so it seems reasonable to assume that there is some
capital accumulation because of that additional population group.
Q242 Lord Turner of Ecchinswell:
I think I understand. I assumed originally that you had modelled
the impact of immigration versus a counterfactual of no immigration,
in which case I would have thought you would logically take into
account whatever your model suggested was the capital response.
If that is not so, if one worked on the assumption that this is
an open economy where change in the supply of labour would have
no necessary impact on our savings rate and we would simply bring
in savings from the rest of the world through an international
capital flow, it would not be 3.1% because to the extent that
there was any capital it would all come from overseas. People
have to have a return on their capital and therefore the GDP might
be higher but at the GNP level 3.1% cent would still be the correct
number.
Ms Riley: If it was completely the case that
all the capital was financed from abroad, that probably would
be the case. That is the reason for drawing attention to national
income per head or consumption per head which probably has not
risen as much as GDP per head.
Q243 Lord Skidelsky:
I just wonder why anyone should be in the least bit interested
in the impact of migration on GDP. Surely the important thing
is its impact on GDP per head and one is only interested in its
impact on GDP in so far as that is an input number into working
out the latter; yet everyone always starts with its impact on
GDP.
Ms Riley: I think that is probably right. We
have two sets of estimates here that you are asking about. It
is difficult to explain. One is a set of model-driven simulations
where we do take into account these capital feedbacks etc. The
other is more like a growth accounting exercise where you are
saying how much of the population is migrant population and therefore
how much are they contributing to our GDP. That is a different
exercise than the model simulations. We have in our model simulations
tried to look at the impact on GDP per capita of the increase
in A8 migration. There we find different effects in the short
and long run. In the long run we find an increase in GDP per capita
which is driven by the fact that migrants tend to be of working
age. Therefore, we have more people working for the same number
of people above and, below working age, and therefore GDP per
head rises in the long term.
Q244 Lord Layard:
In some ways one might say that the real issue is the impact on
the incomes of the resident population, which is not necessarily
well reflected by looking at anything on a per head basis. For
example, if we come back to the capital argument, would I be right
in saying that it makes a lot of difference whether there is an
increase in capital, regardless of who it is owned by, in terms
of its impact on the wages of the resident population? If the
capital comes in, then the existing capital has to be spread over
so many more workers and that sustains the incomes of the existing
population whoever the capital is owned by.
Ms Riley: Yes. That is the type of effect we
find in the short versus the long run. In the short run for example,
if we have a migration change or a change in the population and
the capital stock has not adjusted to that change, we might expect
to see a dampening in wage growth in comparison to the case without
that population change. In the longer term when capital accumulates,
productivity and wages should reflect that accumulation and we
would return to a situation where wages would go back to where
they were without the migration change, so yes, that capital effect
does affect the existing population, and migrants as well, as
it accumulates, regardless of who owns it.
Q245 Lord Lamont of Lerwick:
Do any of these calculations take into account lifetime costs?
Ms Riley: No, they do not. I am not sure I understand.
Q246 Lord Lamont of Lerwick:
If you are taking the benefit that immigration brings to the economy,
obviously in the early years when they are actively working that
is a plus. I was asking: does this take into account consumption
when they are no longer part of the labour force?
Ms Riley: No. We have a dynamic model from which
to calculate these estimates but we are not looking at the effects
in 20 years' time and discounting them through to today. That
is, of course, relevant particularly when you are thinking about
the effects on public finances etc. The estimates we have calculated
do not take these longer term factors into account. When we think
about the public finances for instance, if migrants arrive during
their working age and stay for the rest of their life, they will
have a period where they are perhaps not contributing actively
to the economy. They will still have made a net contribution because
they have missed out their childhood during which typically individuals
are recipients of public funds. We do not have estimates of the
size of these effects.
Q247 Lord Best:
Your recent modelling of the economic effects of migration from
the A8 countries for the period 2005 to 2015 found a negative
impact of immigration on productivity. The impact on GDP per head
is slightly negative for the first few years before turning slightly
positive. What explains those effects and how confident are you
in the assumptions made in the model? How accurate do you think
the results will prove to be against the actual economic impacts?
Ms Riley: I will try to explain what is driving
those results. The reason there is a negative impact in these
estimates on productivity is largely a short run phenomenon. It
is because it takes a while for the capital stock to adjust. This
is the point that we were making earlier, that wages and productivity
in the longer term, as capital accrues, will rise because of that.
In the short run, if you have an unanticipated migration shock
or population shock of any kind, you might expect the capital
these workers are meant to work with not to be in place. Therefore,
that productivity is slightly lower. That does not mean that migrants
in a particular job are less productive than other workers; it
just means that the capital stock is lower relative to the number
of workers are working with a lower capital stock per head.
Q248 Lord Best:
Might they also acquire English as a better language than when
they start speeding up the process, making them more productive?
Ms Riley: Those types of effects have not been
taken into account in these calculations. They would mean that
the short-run effect on productivity would be probably lower than
we have estimated here. You would see a time profile where you
would increase productivity more quickly as migrants assimilate
to the new country. That is not what is driving these results
but clearly that would be important to take into account. The
effects on GDP per head are negative in the short run and positive
in the long run in these estimates. The reason for that is two
fold. First of all, for the reason I have just explained, productivity
is a bit weaker in the short run because of the capital stock.
Secondly, we have this increase in the population which, if wages
are not entirely flexible in the very short term, we would expect
maybe to see some period of economic adjustment where you would
have a small rise in unemployment. We have a change in the population
and not everybody finds a job immediately. Therefore, they are
not contributing to GDP immediately. For those two reasons, (the
productivity effects and the unemployment effects) GDP per head
is a bit lower in the short term. The unemployment effect goes
away relatively quickly and people are working, so that starts
to raise GDP per head. As the capital stock accumulates, productivity
starts to rise and that also affects GDP per capita. There are
these two effects that mean that in the longer term we have in
any case not negative GDP per capita effects. The reason in these
estimates we have a positive GDP per capita effect in the long
run is driven by the fact that migrants are of working age, actively
contributing to the economy as opposed to being children or above
working age.
Q249 Lord Turner of Ecchinswell:
How have you defined working age and with increases in longevity
should we not make working age in itself an endogenous variable
rather than a fixed assumption?
Mr Kirby: In this model we have taken the working
age defined as it is now.
Q250 Lord Turner of Ecchinswell:
Which is 20 to SPA?
Mr Kirby: Yes.[7]
Q251 Lord Turner of Ecchinswell:
End of story. So you have the first five years of the rise in
the women's SPA but you stop in 2015?
Mr Kirby: That is right, yes.
Q252 Lord Moonie:
Are you testing the model all the time in practice to see how
reality conforms or is anybody else?
Mr Kirby: This model is tested by ourselves
and our users. It is a model that is sold to the Bank of England
and many financial industries. We do run tests. We look at the
properties of the model compared to others that are similar to
see if the results are as we would expect. We also look at it
in terms of how it performs in forecast mode as well, so we do
a number of different tests to see how it does perform.
Q253 Lord Macdonald of Tradeston:
We are told that a lot of A8 migrants are currently doing low-paid
jobs for which they are over qualified. How, in your view, can
we expect the economic impacts to change if these A8 migrants
do find higher-paid jobs better suited to their skills?
Ms Riley: You asked in a previous question how
confident we were about whether our models reflect what is actually
going on. Our results are sensitive to the assumptions we make.
One of the key assumptions is the productivity of migrants versus
the existing population. Given the occupational distribution of
A8 migrants, which is quite different from migrants that we have
seen in the last decade, relatively concentrated in low skill
occupations, we might expect to see a slightly negative impact
on productivity certainly in the short term. I should stress that
does not mean that a migrant is less productive in a given job.
An A8 migrant may in fact be more productive and much of the evidence
suggests that is the case in a given job because maybe they are
highly skilled or over qualified for the job they have taken.
If you are looking at population averages, if you suddenly have
a change in the occupational structure of employment, you would
expect to see productivity changes. If we look at the wages of
A8 migrants as an indicator of their productivity relative to
the rest of the population, that would suggest that A8 migrants
in employment are about 74% as productive as the average person
in employment in the UK. That is based on just taking wages as
an indicator of productivity. Those numbers are consistent with
the research that we are aware of but have not conducted that
tried to look at the returns to education for migrants versus
other people. It is clear that there seems to be a penalty associated
with being a migrant of the magnitude I have just described. We
have conducted some estimates where we look at how the macro-economic
impacts differ if we take into account that 25 percentage points
difference in productivity levels. It translates quite straightforwardly
into GDP. If we have an estimate of the GDP effect of 0.9%, say,
then we take roughly a quarter of that off for difference in productivity.
We can adjust the estimates to take into account those types of
factors. I suppose what we would need to know to answer your question
would be, if migrants then move into jobs commensurate to their
skill levels, exactly what the migrant skill levels were. That
is not something we have done research on. However, we could say
if they were to assimilate to the average of the existing population,
then we would expect to see the actual GDP effects observed at
the moment to rise by about a third, say. There is a separate
issue relating to A8 migrants. There is evidence about migrants
generally that there is this short term wage or productivity penalty
which disappears over time. It is not obvious that A8 migrants
will go into more skilled occupations and therefore that this
penalty will disappear. They may very well do but it is not obvious
that they will do because the nature of this migration is very
temporary. It seems that A8 migrants come, stay for a couple of
years and then leave again. They would never achieve that assimilation
process. They may never move into those higher skilled occupations.
If they stay, of course you would expect them to.
Q254 Lord Kingsdown:
Have the rates of profit and capital accumulation changed in response
to immigration? What empirical evidence is there of that?
Dr Mitchell: We are not aware of any direct
empirical evidence on the response of the rate of return to increased
immigration. Economic theory suggests that the rate of return
on capital should rise, with a lag, and the capital stock would
rise as well. Certainly since the mid-1990s, we have seen the
rate of return on capital rise in the United Kingdom, but we are
not aware of any work that tries to identify the direct causes
of that rise. As well as the migration into the UK, we have had
the latest shock of China's increased globalisation and interaction
in the world economy. That is widely believed also to have put
up rates of return. We are not able to offer any direct evidence
on what effect migration has had but we are not aware of any other
work that has looked at it either.[8]
Q255 Lord Lamont of Lerwick:
The scale of immigration from eastern Europe was vastly under-estimated.
Do you think, if there had been a more accurate estimate or more
planning, this could have made any difference to the economic
impact of immigration in any way?
Dr Mitchell: Yes, we do. Rebecca I am sure in
due course will talk about this. Basic economic theory again suggests
and our simulations also suggest that if the migration had been
better anticipated we would have expected some of the capital
stock to have increased in advance of the migration inflow. As
a result, productivity would have increased more quickly. We also
have some more general views relating to the issue of why the
migration flows from the A8 were unanticipated. Certainly, if
one looks back to the sole piece of evidence we have been able
to see, which provided forecasts of the likely impact of the A8
accession in 2003, a forecast of 13,000 a year inflow, obviously
we can now see that that was a considerable under-estimate. We
notice from the Home Office's joint submission to this review
that they continue to make projections of future migration flows.
They seem to deny that these are forecasts. They say they are
based solely on looking at past trends. To us, looking at past
trends seems like a very restricted forecasting model—an
autoregressive model. Why they were doing that in 2003 and why
they continue to do that does surprise us because we would expect
the migration inflow to be determined not just by past trends
but economic factors since they do to a degree influence that
inflow. Our research has indeed found that to be the case. For
example, we find that the relative per capita income of the UK
relative to the source country is a key driver of migration inflows
into the United Kingdom, as are the stock of migrants of that
nationality already present in the United Kingdom. We certainly
do not dispute the view that the Home Office were under-anticipating
the likely inflow in 2003. But we are not sure if everyone else
was under-anticipating it to the same degree as the Home Office.
We hope a lot of people put confidence intervals around their
forecast—an charts—and we would certainly encourage
the Home Office increasingly to go down the route of quantifying
the uncertainties.
Q256 Lord Lamont of Lerwick:
Maybe the private sector had a better idea all along.
Dr Mitchell: Yes.
Q257 Lord Lamont of Lerwick:
The effect on capital would not have been different from how it
turned out anyway.
Dr Mitchell: Yes. To the extent that the public
sector forecasts are a guide to the general public, we would encourage
further work to be done on modelling to identify the determinants
of the inflow with an aim to producing better forecasts so that
the likely economic benefits of migration can be maximised.
Ms Riley: We are talking about economic adjustment
happening more smoothly. Economic adjustment is typically less
costly if it happens less abruptly. In our calculations we would
probably not have seen much of an unemployment rise if there had
been some anticipation.
Q258 Lord Skidelsky:
Do you have any doubts about using models that depend on rational
expectations to the extent that yours do? All your forecasts on
impacts and so on depend on this and on that depends your distinction
between short and long run impacts and your view that in the long
run you get a certain type of equilibrium. Basically, they are
equilibrium models that depend on these things. Has that any real
connection with what goes on in the real world?
Ms Riley: First of all, our models do rely on
rational expectations.[9]
However, they also are estimated models and include adjustment
costs. We do have empirically determined adjustment to the long
run equilibrium. The model is constructed from a set of estimated
equations, where we take these short-term factors into account.
We do also sometimes look at how impacts differ if we do not assume
forward-looking behaviour. Typically what we find is that adjustment
is more drawn out and more oscillatory as well.[10]
Q259 Lord Skidelsky:
You mean up and down?
Ms Riley: Yes. We get greater fluctuations in
adjustment. We have not tested the sensitivity to the assumption
of forward-looking behaviour in this particular exercise but that
is what we have found when we have undertaken similar analysis.
Dr Mitchell: As in any modelling exercise, we
are aware of the limitations of the model and we certainly are.
Our model embodies state of the art theory and allows frictions
to explain short run deviations from the long run. That is consistent
with the modern paradigm. Of course, paradigms change as we all
know, making it particularly important, as one can do in modelling
exercises of our sort with estimated models, to quantify the uncertainty
around estimates. A couple of times today we have tried to emphasise
this. The migration data themselves are uncertain. The forecasts
of migration flows are uncertain as are the likely macro-economic
effects. When people adopt quite strident views based on point
estimates, it is always useful to bear in mind, as any sensible
person does, there is a considerable range of uncertainty around
those estimates. We always try to give users an explicit guide
to how uncertain we are with our estimates, to reflect our ignorance.
6 Note by witness: This migration refers to
migration from all counries outside the UK, 1998-2005. Back
7
Note by witness: UK working age population is actually
defined as 16 to State Pension Age is NIESR's global econometric
model NiGEM. Back
8
Note by witness: Barwell (2007) "The macroeconomic
impact of international migration" Bank of England Quarterly
Bulletin Vol 47, No 1, pp 48-59, refers to two studies on Israel
that find evidence that an increase is net inward migration led
to a rise in the rate of return to capital and stimulated investment. Back
9
Note by witness: The results presented in NIESR's submission
are not forecasts. The simulations are a guide to the likely economic
impacts for a given set of assumptions. Back
10
Note by witness: NIESR's model is set in what is essentially
a neo-keynesian framework where agents are forward looking, but
nominal rigidities, namely sticky prices and adjustment costs,
slow down the adjustment to the long run equilibrium. Back
| 
