In this article I am going to look at the narrative side to econometric analysis, and what we gain from understanding bundles of data through an instrumental variable. There are various limitations to this type of analysis, in my opinion the most major one being the requirement of a natural experiment, and all the factors, which may go wrong alongside that. This is going to be based on a lecture by Matt Dickson a researcher who is currently looking at the raising of the school leaving age and consequent effects on the labour force survey.
Policing and Crime
This is based on the academic paper “Panic on the Streets of London: Police, Crime, and the July 2005 terror attacks”
The government has a specific interest in policing and crime through an economic perspective due to the associated costs and its effect on the general population. The statistics used for this analysis are the total amount of recorded crimes per 1000 of population, and the amount of police officers per 1000 of the population, for a respective year in the United Kingdom.
We can see that the crime rate and police rate increased alongside each other from the 1960s onwards, but from 1990 onwards there is a decline. This applied for violent crimes, robbery, etc. In this we can conclude that more police officers does not facilitate more crime and vice-versa. In this there is a positive correlation, which does not tell us about causation in the relationship.
The academic paper mentioned above uses a fixed variable in order to examine the correlation, in that the influx of police was rare due to the terrorist threat. This allows a natural experiment to be created, as the increase in police had nothing to do with “conventional” crime.
The paper specifies a treatment group that consists of the London boroughs of Westminster, Kensington, Camden, Islington, and Tower Hamlet, while the controlled group is every other borough in London. There was hardly any difference in police deployment throughout those areas before the attacks. During the operation of increased policing due to the attacks where there was effectively police on every corner. Now considering the crime rate we see that before the period crime in the treatment areas was a bit higher, but then with expanded police force we see that the crime rate drops after and during the attack period for the treatment group. This had shown a definite decrease in crime rate, but it required a massive spike in the number of police per 1000. The data allows us to establish that there was an approximate elasticity of how changing policy changes the crime rate. Resulting in an elasticity of 0.38, so a 10% increase in police numbers reduces crime by approximately 4%.
The critique that we may employ here is that due to the very nature of the natural experiment, it may have had an unaccounted for effect on the operation of crime. There is also no control comparison so a terrorist attack without a changed police response. Making it difficult to immediately take this statistical analysis at face value, clarifying the limitation of natural experiment.
However, the results are still interesting. Allowing us to expand our analysis into what kind of policing would be required to keep crime at a desired level, accepting the fact that we cannot eradicate crime. The cost of developing this kind of police force can be brought into consideration or whether other methods can be introduced which would be as effective but at a lower cost, in this noting the CCTV network developed throughout London.
Nurses Pay & Death
In healthcare payment for nurses is based on equity, in that everyone does the same job therefore they should have the same pay. However, this may have unintended consequences on the service that nurses provide, and the extent to which agency nurses are employed.
If we consider where there is a greater use of agency nurses and death rate from heart attack we see a concentration around London, the Home Counties, and a few Northern cities. In this one may suggest that the quality of agency nurses is lower hence the death rate. However, we can note that there may be reverse causation. Hospitals with high death rates struggle to employ and maintain staff; thereby making agency nurses a requirement. It is also an assumption that agency nurses are more likely to give poor service.
We can breakdown this relationship through considering outside wages as a third variable because it affects the use of agency nurses but not necessarily the quality of the hospital. Carol Propper and John Van Reenan outline the employment of agency nurses and the quality of healthcare in greater detail in the paper “Can Pay Regulation Kill? Panel Data Evidence on the Effect of Labour Markets on Hospital Performance”
School and Wages
How much is education worth?
Education has an impact on fiscal policy, welfare, labour, and crime. Increasing education could lead to a decrease in crime, and a variety of other relationships. Highlighting the importance we place on education in economics.
Considering a distribution of hourly wages and years of schooling we may note a positive relationship whereby years of schooling increases alongside hourly wage. Some problems we run into when considering education:
- Education is not randomly distributed amongst the population.
- People choose how long to remain in education.
- There are unobserved and unmeasureable factors that lead to higher wages.
- We hope that we can pick two random people and by years of education be able to assign them a wage
The natural experiment used here looks at the school leaving ages from 1949 to 1967:
The people who were forced to stay in school just stayed the one more year after 15. This made them move from the 15 age group to the 16 age group leaving, at that point around 60% had left school but this had a positive effect on wages.
RoSLA – Raising of the School Leaving Age
Pre-RoSLA and Post-RoSLA comparison, the average age of leaving education increased, and meant that 1/3 people stayed in school longer. The policy change made this an effective natural experiment. We can consider the indifference difference:
A whole extra year of schooling meant a 6% increase in hourly wage. Why did it have this effect?
Breaking up periods, when you could leave school. This resulted in a far greater proportion with academic qualification for those that could not leave before the summer exams. It does it for a greater extent to those that are post-RoSLA.
Dickson and Smith in 2011 have looked in detail at the labour force survey and the effect of raising the school leaving age if we were to exploit another increase in the minimum leaving age. This would be exploited again, by moving the age bracket again achieving higher qualification, resulting in a greater wage in theory. The question brought here is to what extent will this relationship hold?
Education is always a difficult variable due to our inability to measure to what extent it has been effective, the issue that arises is that even if everyone got the same education individuals would have gained different benefit. Let alone the fact that there are a variety of programs and courses, which pupils may take.
Copyright Almog Adir © 2014 · All Rights Reserved · My Website