Predictive Maintenance and Principal-Agent Problem: Why the new LTA SMRT $1b deal might not be so good news for passengers

Predictive Maintenance simply means using analytics to predict when a component is likely to fail so as to allow maintenance to take place before the break down. The SMRT and LTA have been in the headlines in the last couple of years, and predictive maintenance, I’d have thought should be something high on their agenda.

The principal agent problem is said to occur when there is a disconnect between the entity taking an action (the agent) and the entity on whose behalf or under whose authority the action is taken (the principal), and the agent sometimes goes counter to the principal.

Straits Times published an article over the weekend on the new deal between the regulator LTA and the publicly listed transport operator SMRT. http://www.straitstimes.com/singapore/transport/government-and-smrt-reach-agreement-on-new-rail-financing-framework

Prior to this deal, SMRT made its own decisions on the maintenance, upgrading and replacement of rails (including light rails). The rails are a key asset and it makes sense for SMRT to maintain them if you believe in privatisation. Theoretically, they would ensure these assets are in good enough working order so as not to damage their profits. Hence they would balance the risk of breakdowns for example, with the potential costs, loss of business, fines from the regulators... 

The current deal was for 30-40 years, putting the onus firmly on SMRT to take care of the rail asset. The new deal is only for 15 years. SMRT pays a fee into a sinking fund that will be used by the LTA for replacements, upgrades. The difference is the time horizon (shorter) and the exposure, unknown cost today compared to known fixed cost tomorrow.

Under the new deal, SMRT is now only concerned about the maintenance, but decisions on replacement, upgrades are taken by the LTA. What's also interesting is that the LTA will share the revenue risks, so if the rails give problems and there is loss of revenue, the LTA is hit too. Tax payers might have a think about that, but this is not what I am after.

To me what this deal means is that the SMRT has been found to have failed in its role in taking care of the rail asset. That is why the regulator has to step in. It's not that the SMRT totally failed in taking care of the rails; after all, the services didn't go totally kaput. They just did what was necessary, what was profitable and no more, and this was found to be inadequate by LTA.

The SMRT has a virtual monopoly of transport in the areas it operates. When the trains breakdown today, most people have few cost-efficient alternatives. At most SMRT losses amount to the rail revenue foregone while the network is down, plus the costs of providing bus services to ferry stranded passengers, plus the fine from the regulator, LTA. To me that's where the problem lay. The fine was too small in the eyes of the SMRT, a publicly traded company.

A quick look at the SMRT's accounts will show why. On page 29 of PDF of the annual report, https://www.smrt.com.sg/Portals/0/InvestorRelations/Annual%20Report/2016/SMRT%20Annual%20Report%202016_LR.pdf , the largest chunk of the SMRT's earnings (EBIT) comes from rental (almost 20X). Not rail. Rental. The R in SMRT might as well stand for Rent.

Source: SMRT Annual Report 2016

Given that the SMRT has to make profits, it makes more sense to make sure that the rental income is unaffected, rather than the rails are close to perfect. The regulator might have wished the SMRT spent more on rails, but it has little incentive to do so. The costs of rail breakdown is so low. http://www.straitstimes.com/singapore/transport/smrt-fined-record-54-million-for-july-7-breakdown Not even the maximum 10% of rail revenue for a major breakdown. The breakdowns do not hurt SMRT. Hence the SMRT does not have the incentive to keep rails up 100% of the time; the price to pay for failure is low, and that price is determined by the regulations.

Do not blame the SMRT, it was just playing within the rules it was given. The agent played within the rules set up by the principal. And now the principal is changing the rules, not by simply making the fines higher but by changing the game altogether.

So what about the new arrangement, is it much better?

Under the previous regime, the SMRT had to make sure there wasn't a 'catastrophic' breakdown in rail services that would affect the rent income and the rail income for the medium run, these costs were hard to quantify. Furthermore they were growing with the arrival of competitive ride-sharing and car-pooling apps putting transport by car as a more viable alternative to train transport for some. Now the amount the SMRT would have to pay to repair/upgrade the rail network is fixed, to the amount they put in the sinking fund.

Hence, to me, the principal agent problem has not been fixed, the agent (SMRT) is even freed up from the impact of variable costs of replacements; in fact the problem can arguably be said to have been made worse.

Let me put it this way. Imagine you own a bridge between 2 villages, and the next crossing point is miles away. You also own the vehicles allowed on the bridge. You also own the malls at the end of the bridges where people spend their time. And the malls make more $ for you than the bridge crossing business. Every spare dollar you would like to reinvest would naturally go to the malls, because the impact on your revenue is greater, better air conditioning, face-lift... However, since the bridge is also owned by you, you will ensure that it doesn't get structural faults. You will watch the load, and speed on vehicles, and manage it so as not to impact the structure. Of course you will cover potholes and maintain the bridge too.

Now someone buys and leases the bridge back to you at a fixed cost, and you can use it as you wish. He will ensure the bridge in reinforced and new struts placed as wear and tear takes place. If it's more profitable to ferry people in 2-tonners rather than individual cars, or even pile up as many people you can in the buses, you would do it, since you don't care about the structural risks to the bridge that are now someone else's problem.

But what I am most concerned about is that the fact that predictive maintenance is very unlikely to yield an optimal actions.

SMRT knows the usage of the rails, the weight, the speeds of each train on the network at every point. This data is critical in predicting failure of components of the trains - the more weight you carry as higher speed, the more likely your suspension is to be affected for example. But that same data is also critical in predicting the failure of the rails. And by separating these costs, making part of the rail costs fixed, the cost-benefit equation is seriously impacted. The SMRT might not have that much of an incentive to manage the loads/speed of trains to take care of the rails anymore. It's simply not their problem anymore. 

Yes the LTA, will have access to data on rail fatigue and they are unlikely, unless provision is specifically made, to have access to train speeds, loads that are a major component in predicting when the rails need to be replaced. Furthermore, since these factors are controlled by SMRT, LTA would not be in a position to manage these factors to manage the rails. Even if predictive maintenance models were fed enough data, the ability to make these models actionable is likely to be limited.

And that’s a real shame because analytics is easily able to help solve some of the major problems that SMRT has been facing. However the new structure of their relationship with the LTA is unlikely to allow the application of such models so as to avoid costly and irritating rail network issues.

(As a side note, since the SMRT now can be said to have limited risk on rail operations, it can use its funds more efficiently and is probably more valuable with this new arrangement)