Minimum Wage: Preliminary Impact Assessment (Warning: another long and wonkish post)

It’s been nearly a year since Malaysia’s minimum wage law came into effect. Although enforcement has been held off until next year from the loads of companies applying for a postponement, most have already complied.

So here’s my crack at trying to figure out what the minimum wage’s impact on Malaysian employment, unemployment and incomes. Most of this material is taken from a presentation I gave at UTAR earlier this week.

There are two basic arguments on the impact of a minimum wage on employment:

1. Higher incomes draws greater labour supply which leads to greater consumption, which in turn leads to higher employment; but...
2. Higher wage costs means slower job creation/hiring, or higher unemployment

What we really want is a long term assessment, but since we’re less than a year into this, that’s obviously out of the question. Moreover, public data on the data we’re interested in is short-spanned and at an aggregated level only. Monthly wage data has only been regularly collected for manufacturing, and only since 2009; the monthly employment and unemployment from DOS is of the same vintage. There are no Malaysian time series for overall wages.

We also have some identification problems as many factors influence wages, employment and growth. Any attempt at teasing out the impact of the minimum wage alone thus has to be treated with caution.

The results

To save the less econometric minded from boredom, here’s my results:

1. Both overall employment and unemployment appear to have been affected, though unemployment more strongly than employment;
2. Overall employment was raised about 5% (4.8% without seasonal dummies, 5.3% with), or about 640k jobs;
3. Overall unemployment was raised by around 8.1%-8.8%, or about 36k-39k jobs;
4. Manufacturing shows a different response – employment did not increase, but average incomes did;
5. I suspect this may be due to high propensity to export (demand is externally derived), which means employment will not vary with domestic economic activity (mis-specified model and/or omitted variable bias)
6. Average incomes in the sector were raised by 7.3% (w/o seasonal dummies) or 5.6% (with seasonal dummies)

In short, the minimum wage appears to have had the desired effect, at least in raising incomes over the short run.

As a side effect, unemployment was also raised but offset by greater employment. How do you get an increase in both at the same time? Basically, higher wages at the low end have drawn people who are out of work, or in the informal economy, into the formal economy. Technically, we got a higher labour force participation rate.

The boring bits – A look at the aggregated data

Basically, I applied two very simple models to test the effects of the minimum wage:

EMP = f(Y) + D and
UNEMP = f(Y) + D

Where:

EMP = employment
UNEMP = unemployment
Y = economic activity (proxied by IPI)
D = dummy variable for imposition of minimum wage

Why IPI and not GDP? Because IPI is available as a monthly series, and GDP is only quarterly. Plus, because IPI is a pretty good predictor of GDP.

We also need to account for structural changes in labour force i.e. registration exercise of illegal foreign workers circa Dec 2010-Jan 2011, which adds one more dummy variable to the model(s) specs.

Note: By rights, I should include seasonal dummies as well, but as the results for these were virtually identical with the unadjusted models, I’m not going to bother reporting them except where necessary (despite some interesting seasonal effects).

I’m also going to test the changes in the labour force participation rate, but this will be a simple test of a change in the intercept.

Employment model:

This one’s pretty unambiguous – there was a statistically significant increase in employment at the same time the minimum wage came into effect. This doesn’t change if we add seasonal dummies to the model.

Unemployment model:

This one’s a little harder, because the model is not a good fit for the data. Nevertheless, there is a statistically significant positive change in unemployment occurring at the same time as minimum wage coming into force.

Labour Force model:

I really ought to specify this model to test the change in slope, not intercept, but for my purposes this does the trick just as well. In essence, the LFPR has been generally increasing over the span of the data, but is trend stationary. That changed in about March 2013, when it started leaping up quite remarkably. While the causality might be iffy, I can’t think of anything else that could possibly be a factor – certainly GE13 shouldn’t qualify – and it fits the minimum wage narrative pretty well.

In depth with the manufacturing sector

I’ll be using the same approach here, although we can’t test for manufacturing “unemployment” because there’s no such thing. So instead, I tested the impact on incomes (average wages per worker).

Again two simple models:

EMP = f(Y) + D and
INC = f(Y) + D

Where INC = average income per worker

Manufacturing employment:

I had to use an AR(1) term here because of serially correlated errors, but while the model is a good fit, IPI is not a significant explanatory variable for manufacturing employment. I can think of a number of reasons for this, mainly stemming from high foreign ownership and an export-oriented approach – domestic production wouldn’t be a good indicator for foreign demand, or take into account the globalised supply chain of which Malaysia is a part.

However, the dummy for the minimum wage is statistically significant and positive, indicating that it did raise employment in the manufacturing sector. As it turns out though, this wasn’t really the case as explained below.

Manufacturing employment with seasonal effects:

[Note: the residual chart is virtually identical to the model without seasonal dummies]

Adding seasonal dummies to the model above, the minimum wage dummy turned out not to be significant at all. In the result above, the dummy really was just capturing a seasonal effect – hiring is generally much higher in January, which coincided with the implementation of the minimum wage. In fact, hiring in the manufacturing sector is also significantly higher in November and December as well.

But I can only confirm this for the manufacturing sector alone, as the models for the aggregate employment and unemployment data do not show the same seasonal effects i.e. my results for aggregate employment and unemployment remain valid.

Manufacturing wages:

There is no such ambiguity with the income data – the average wage did rise with the minimum wage.

Final Thots

There are a whole bunch of caveats to go with this sort of analysis:

1. Short sample span, and only in expansion phase of current business cycle; dynamics are not tested either (i.e. there should be a lag structure to employer/employee responses to MW)
2. MW should affect only labour market for workers with wages around the MW level (short term), and not total employment. Aggregate numbers may hide the truth
3. Presence of large illegal foreign labour force substantially masks impact of MW (variability of official employment/unemployment stats may be muted by unseen workforce)
4. Low wage workers might be concentrated in certain sub-sectors of the economy, but lack of data to establish for certain
5. New research paper suggests employment impact is mainly on the margin i.e. new job creation for smaller companies, rather than employment in the economy as a whole
6. The literature suggests a nonlinear threshold response of employment to minimum wage levels – Malaysian MW might be too low to show significant unemployment effects

Those are the ones I can think of offhand, and there are bound to be more.

So take this for what it’s worth, a highly preliminary look at the economic impact of the minimum wage. A more concrete assessment will need to wait a few years, and hopefully with much better data than I have at hand.

Technical Notes:

Data on employment, unemployment, labour force participation rate, and the Industrial Production Index (IPI) from the Department of Statistics