"What your post shows is no correlation, whatsoever, between union coverage and labour productivity. Unless you choose to highlight a couple of examples, and ignore the rest of your data. Which isn't a pattern at all."
This made me think, well a simple scatter graph could solve this. So I whacked the two data sets into excel, and this was the result .....
In addition, I calculated the Average Annual Labour productivity growth for those countries with a collective bargaining coverage figure of below 40% and above 60%. The respective averages were 1.7% and 2%. So there's certainly a significant pattern, if not a drastically pronounced one.
I will say however that calculating any statistical significance is nigh on impossible because of the numerous other factors that would need to be included and accounted for. As such these findings should be taken as indicative evidence rather than concrete proof of any correlation between collective bargaining coverage and labour productivity growth.
Posts
Two things:
ReplyDeleteFirstly, there doesn't seem to be any significant pattern in the data.
Secondly, what is the theory behind unions increasing labour productivity. If the marginal revenue from investment in labour was going to give the firm enough money to cover the marginal cost, then they would invest.
The cross-country comparison is not satisfying enough to even say that there is a link that connects labour unions and labour productivity, let alone what direction the causation runs in (countries where the labour is more productive may have more ability to set up unions and coerce government).
Ultimately, you need a model, nice mathmatical models allow you to distill these sorts of things, and make it easier to compare your assumptions with other people.
Hi Matt. Once again I would like to thank you for your constructive criticisms. You are a rare calm voice in the maelstrom that is NZ's political blogsphere :-)
ReplyDelete"The cross-country comparison is not satisfying enough to even say that there is a link that connects labour unions and labour productivity"
I agree - the data I provided is far from conclusive. I do however look at it as indicative evidence.
"Ultimately, you need a model, nice mathmatical models allow you to distill these sorts of things, and make it easier to compare your assumptions with other people."
Granted. I did a little hunting for such a model and found this econometric study which concludes
the following:
www.rdg.ac.uk/Econ/Econ/workingpapers/emdp429.pdf
"The MG and PMG estimation techniques that we
used together with more traditional methods are at the forefront of panel data
econometrics. Our time-series and cross-country analyses revealed that this
relationship has been different among countries and over time. Controlling for
possible time and country-specific effects, the panel data analyses allowed the
estimation of a common across countries long-run coefficient. The good performance
of our regressions and the stability of our results, we interpret as evidence in support
of the appropriateness of the econometric method we employed.
Our basic results provide robust evidence of a positive impact of unionism on
productivity. Both the long- and short-run effects are positive and statistically
significant, although we also offer some evidence suggesting that country-specific
factors, like the strategies employed by the national trade unions and the degree of coordination
among them and between them and the employers, might play an important
role at the short-run."
Hey Philip John, did you understand the models in that paper you have quoted?
ReplyDelete"Hey Philip John, did you understand the models in that paper you have quoted?"
ReplyDeletehehe - hell no. Did you?
Roger Nome said...
ReplyDeleteHehe no. Did you?
Yep, I have developed a number of different variants of the Auto-regressive algorithms mentioned in that paper for a commercial financial analytics application that would be available locally in the next few months. The variant described in the paper auto-regressive distributed lags (ARDL) is a sub-class of the general ARMA (Auto-Regressive Moving Average) model. The analysis in that paper focused on time-series analysis, meaning that the series is regressed on its past history, where there were no external inputs at all to the system. The general model for economic dynamical systems analysis of today use the : is Auto-Regressive Moving Average Exegenous ,ie, ARMAX (linear case) for short or NARMAX (non-linear ARMAX). ARMAX or NARMAX can model many external inputs into a system (be it economic, climate or population system, etc) which then drive the system response. So, ARMAX is more general than ARMA, which is more general than AR (auto-regression) used in that paper, you've have quoted. I can point you out to some economic papers on dynamical system analysis using general ARMA.
Phillip John, the last comment is mine.
ReplyDeleteCheers,
Falafulu Fisi
Interesting, I just had a look over the article. I haven't got much experience with panel data, as I mainly use simple time series data at work, but there are a couple of things I'm skeptical about.
ReplyDeleteThe hypothesis we are testing is that unions cause higher labour productivity right. This study shows that unionisation rates are correlated with labour productivity, not that they cause productivity. As I said earlier, in a society where the labour is more productive, they have more power as an interest group. This implies that they are able to capture government, and improve there bargaining position.
Furthermore, the study does not look at alternate variables that may explain this link. For example, if unionised countries have a lower level of labour force participation, then the productivity of a labour unit will be higher (all other things equal). Although they run a fixed effects model, I think it would be appropriate to see if the relationship exists between other variables, and then choose the model that makes the clearest economic sense.
Thirdly, they state that the majority of literature finds that the union effect is negative. As a result, it would be nice to see the results of this study replicated.
Fourthly, they use a cobb-douglas production function, which is not an accurate representation of aggregrate production. A CES production function would have been more appropriate.
And lastly, they do not create a microeconomic model that sufficiently explains how unions influence labour productivity. An empirical model gives us correlation. It needs a logical model to explain causation.
h984"The hypothesis we are testing is that unions cause higher labour productivity right. This study shows that unionisation rates are correlated with labour productivity, not that they cause productivity"
ReplyDeletereally?
"Our basic results provide robust evidence of a positive impact of unionism on productivity."
Sounds like a little more than just a correlation to me.
"As I said earlier, in a society where the labour is more productive, they have more power as an interest group."
The study measured PRODUCTIVITY INCREASES, not "productivity".
"For example, if unionised countries have a lower level of labour force participation, then the productivity of a labour unit will be higher (all other things equal)."
But that just isn't true.
http://en.wikipedia.org/wiki/List_of_countries_by_employment_rate
(I know employment rate is different to LFPR, but if you do some research you'll see that in highly regulated labour markets LFPR is much higher relative to the employment rate than it is in liberalised labour markets).
"Thirdly, they state that the majority of literature finds that the union effect is negative."
Yes but they also say that other studies generally don't use a decent number of country to country comparisons.
"Fourthly, they use a cobb-douglas production function, which is not an accurate representation of aggregrate production. A CES production function would have been more appropriate."
Can't answer that one.
"An empirical model gives us correlation. It needs a logical model to explain causation."
I agree, though there are plenty of logical reasons for the positive effect of unionism on productivity - i.e. lower worker turnover, higher moral, more work-place training = higher workforce skill levels.
Hi,
ReplyDeleteI'll just go around answering the point by numbers:
1) They paint it out as measuring a relationship, however there empirical model can do is show a correlation, it doesn't matter how they say it. You need a logical model before you can move to causation
2) They are looking at growth labour productivity. A possible model to explain this would be that if there is a labour union, firms will invest more in capital but hire less labour (all things equal). Both these factors will lead to growth in labour productivity, but this isn't necessarily a good thing, I'll come back to this.
3) Firstly, there was nothing on that wikipedia page, but I know what you were trying to say. The fact is employment and labour force participation are completely different. Labour force participation is the proportion of the population that is willing to work, employment is the proportion of the people willing to work who have jobs. If there are unions, it is more difficult for people to get jobs and people will drop out of the labour market, implying that labour force participation is lower.
4) You need more than one paper to make a point. I am sure that other people have used pooling techniques to study unions, its a popular topic and pooling techniques have now been around for a while. Just because the paper says it is the best doesn't mean it is.
5) I don't agree with the 'logical reasons' for labour unions being a good thing, although these reasons could be used to explain why labour force productivity growth could be higher with unions. Firms could improve conditions and reduce labour turnover without unions, if it was optimal for them to do so. If a firm is willing to provide shitty conditions, then it must be more efficient for the firm to do so.
Now there are two questions:
First, do unions increase labour productivity growth?
Second, is this a good thing?
Now the first question is tricky, I agree, and I can actually imagine cases where unions can increase labour productivity growth. However, they don't do it by making workers happier etc, they do it by making labour so expensive that firms buy machines instead of workers, thereby making workers more productive.
The second question is clear to me, unions that are too strong are a bad thing. Why? Well, they constrain the firms choice, they make labour artificially expensive forcing the firm to invest in capital instead. As the price of an input is higher, marginal costs are higher and output growth is lower. This implies that production growth is lower for the economy, which is not good.
Ultimately, it would be possible to make a model where unions increase labour force productivity growth. However, labour force productivity growth is not the goal, maximising output with our limited set of resources is the goal. Unions can only be defended on an equity basis, and I think current employment policy is sufficient to achieve this.
I hope this clarifies my position. I do not disagree that unionization could cause higher labour productivity growth, however the case behind this is tenuous (which is why a analytical model would be useful). However, I do disagree that the increase in labour productivity that could come about from unions is a positive thing, as it would fundamentally lead to lower output (as it would increase firms marginal cost).
Thanks for the very interesting discussion matt. Just a few ideas in response to the points you raised in your last post.
ReplyDelete"1) They paint it out as measuring a relationship, however there empirical model can do is show a correlation, it doesn't matter how they say it. You need a logical model before you can move to causation"
Ok - I get you. I don't really have the time or the expertise to devise a model, though I think I can provide some good reasoning to support their empirical findings - and will do so below.
"If there are unions, it is more difficult for people to get jobs and people will drop out of the labour market, implying that labour force participation is lower."
I know where you're coming from matt. I just don't think that reality supports this theory. Here's that link that I posted before - It was cut short for some reason.
http://en.wikipedia.org/wiki/List_of_countries_by_employment_rate
You may also know that the Scandinavian countries have the highest union density levels in the world.
"Firms could improve conditions and reduce labour turnover without unions, if it was optimal for them to do so. If a firm is willing to provide shitty conditions, then it must be more efficient for the firm to do so."
You would certainly think so Matt. The problem, and I beleive there is very good evidence that shows this, is that many employers are reticent to give up control, and it's to their own detriment.
Many studies show that in free-contracting environments no bargaining/negotiation takes place in contract formation in the vast majority of instances. The employer simply hands the prospective employee a standardised individual employment contract and asks them to sign. This dynamic continues throughout the employment relationship, and creates a situation where the only means by which an employee can change their situation is resignation. Obviously in a unionised workplace this isn't going to be the case. As a result non-unionised workplace usually have lower staff-moral, a higher labour turnover rate (one study found that th annual job turnover was 53% the USA and ranged from 7-23% in Europe), and consequently lower productivity growth.
www-ilo-mirror.cornell.edu
(page 6)
"4) You need more than one paper to make a point. I am sure that other people have used pooling techniques to study unions, it’s a popular topic and pooling techniques have now been around for a while. Just because the paper says it is the best doesn't mean it is."
Admitedly this is where I'm out of my depth. Though I do know enough about statistics to know that, all other things being equal, a study which involves a large sample is of greater indicative value than a much smaller one.
"The second question is clear to me, unions that are too strong are a bad thing. Why? Well, they constrain the firms choice, they make labour artificially expensive forcing the firm to invest in capital instead. As the price of an input is higher, marginal costs are higher and output growth is lower. This implies that production growth is lower for the economy, which is not good."
It's not just a question of how strong the unions are, it's also a question of how moderate they are in the economic philosophy, and how coordinated their bargaining they are. i.e. moderate, highly centralised and powerful unions often exercise real wage restraint during recessions, so as to minimise job-losses. The union movements within the Scandinavian countries, the Netherlands and Ireland exemplify this.
"However, I do disagree that the increase in labour productivity that could come about from unions is a positive thing, as it would fundamentally lead to lower output (as it would increase firms marginal cost)."
This is a values based argument. I believe that there is a balance to be found between output and work-life balance. i.e. countries with highly centralised collective bargaining generally have a much lower average hours worked per worker than countries with liberalised labour markets (employers simply force employees to work longer hours). And there is good evidence to support this
http://en.wikipedia.org/wiki/Image:Yearly_working_time.jpg
As has been shown, there is little evidence to support the idea that a strong union movement causes a lower employment rate. In fact overly generous unemployment benefit levels have been shown to be responsible for most of the difference in unemployment levels between the Germ/France/Spain group and the US/NZ/AUS/GB group.
Sorry Matt - just realised that link for OECD labour turnover doesn't work.
ReplyDeleteHere's another (it's table 2 under section 2.2)
http://www-ilo-mirror.cornell.edu/public/english/employment/strat/publ/etp48.htm
Hi,
ReplyDeleteI'll reply about the points a bit later on (I'm just a little tied up this week as building data is coming out). However, I did see this interesting post from an economist that explains how labour market rigidities (like unions) are tested.
http://rodrik.typepad.com/dani_rodriks_weblog/2007/09/are-labor-marke.html
It is interesting as he provides some alternative ways of defining the labour market issue. It doesn't face the questions we are discussing here directly, but its still interesting.
Rogernome
ReplyDeleteI've got a proposal to put by you in private.
I would appreciate it if you contact me at BluecollarGreenie@Gmail.com so that we can discuss it.
interesting article matt. Have you read any econometric studies regarding unemployment and labour market institutions? I've read quite a few, and most conclude that unemployment benefit generosity is highly correlated with unemployment, while worker protection levels are either irrelevant, or all but irrelevant.
ReplyDeleteAlso, using unemployment as a way of assessing the success of a country's labour market policy is a actually quite misleading. i.e. if jobs are more desirable then more people will be dragged into the labor market, and consequently the unemployment rate will be higher. So are we to say it's a bad thing to have more high quality, well paid jobs? Of course not. In fact, it makes more sense to look at "employment rate", It measures the proportion of your working age population in paid employment, whereas unemployment merely measures the proportion of the labour force that is looking for work - rather meaningless IMHO.
Did you get my reply rogernome?
ReplyDelete