The Economic and the Social: Anatomy of a Contract Change in India

By Rajshri Jayaraman and Debraj Ray

Productivity rose among tea pluckers when their base wages increased, despite a fall in monetary incentives, but this response vanishes within months.

In a variety of interactions, the economic and the social rub shoulders, often in uneasy co-existence. Think of blood donation, for example. People probably do it out of a genuine sense of social altruism, or because helping their fellow human beings makes people feel good about themselves. Placing a monetary reward on this activity by paying people to donate may well crowd out this intrinsic motivation (Costa-i-Font et al. 2011) and (if the monetary price is low) drive down donations. Similarly, we might respect the idea of standing in line, or being punctual to pick up our kids at daycare, or volunteering to work in a poor neighbourhood, but once monetary rewards are placed on such activities – imagine, for instance, being able to pay to skip an orderly, egalitarian queue – the social incentives may be eroded. ‘Extrinsic incentives’ linking performance to money may crowd out ‘intrinsic incentives’ to exert effort.

Extrinsic and intrinsic incentives in for-profit organisations

Is the same true of paid employees in organisations governed by the profit motive? In this case, the thought experiment runs in the opposite direction. The baseline relationship is based on a monetary contract, unlike in the examples above. Can employer generosity, good gestures, or a happy work environment translate into greater employee effort and higher output, out of a sense of gratitude or reciprocity? This is an important question, but a difficult one to tease apart. You could easily imagine that firms would be interested in knowing the answer, if for no reason other than the possible cost savings to be had. After all, a show of generosity or bonhomie or the handing out of employee-of-the-week awards is cheaper than performance pay.

We don’t pretend to have comprehensive answers to these questions, or measures of the quantitative importance of ‘firm culture’. The truth is that nobody has a conclusive answer, claims on TED talks and airport bookshelves notwithstanding. We all hear about companies such as Google, where a relaxed, hands-off atmosphere is supposed to foster productivity. But do we really know this? And if it is true, then how does this work? For instance, does the deliberate adoption of such a culture change the productivity of existing employees, or does it change the composition of personality types who choose to work in such environments? It is important to appreciate this distinction, because it says something about whether a change in firm culture actually works for a pre-existing set of employees, or whether it only matters for new start-ups or for firms undergoing major workforce expansion. In any event, the bottom line is that numerous experimental studies emphasise that individuals respond positively to (unconditional) acts of generosity. In contrast, the traditional consensus in economics is that money – and financial incentives – do the real talking. How do these two viewpoints interact in for-profit organisations?

Our study: A new contract for tea plantation workers in India

In a recent paper, we study these perspectives empirically in the context of a developing country (Jayaraman et al. 2016).1 Our setting is a large plantation in a region of India where tea is a dominant source of low-skilled employment. The tea plantation employs around 2,000 workers to pluck tea from bushes planted in several fields. There are permanent workers and some temporary workers hired to fill labour gaps in the peak plucking season, but – by law – they must be given identical individual wage contracts. The contract stipulates a daily fixed wage, plus additional piece rates per kilogramme of tea-leaf plucked. Regular tabs are kept on the weight of plucked leaf, and wages are calculated daily and paid at month-end.

Wage contracts in the tea industry are renegotiated every three years and apply to all plantations in the locality. Our plantation fell under a consortium of 20 unions and plantation owners (covering about 10,000 workers) that received a new contract in mid-2008. The state government happened to be deeply involved in this one: it mandated an increase in the baseline daily wage by over 30%. Plantation owners contested the increase, but their petition was dismissed by the state high court. In response, plantation owners felt impelled to cut back on incentive pay over and above the new baseline. Worker incomes still jumped up under the new contract, though the high-powered incentives were not so high any more. In summary, the new contract consisted of a shift up, and an accompanying ‘flattening’ of the wage structure, and we have the perfect stage to examine the effects of a change in monetary incentives.

An unexpected outcome: Productivity rises as monetary incentives fall

Observe that traditional contract theory makes a clear prediction about what follows such a change: worker productivity should fall, as the monetary rewards to effort swivel downwards. What happened was unexpected and quite dramatic. Output in the following month increased anywhere between 20-80%, the wide range of possibilities coming from the particular controls we include and the counterfactuals we use. Figure 1 plots the raw data, and shows the striking increase in daily output under the new contract (after ‘Day 0’), relative to two counterfactuals with no contract: a plantation in a nearby location which follows a different three-year contract renegotiation schedule, and our plantation in the previous year (2007).

There are a number of obvious explanations that we will need to initially address. First, the contract change occurs near the start of the plucking season, so the subsequent increase could represent a seasonal effect. And indeed, there is certainly an increase in output under both counterfactuals: in our plantation in 2007, and in the nearby plantation in 2008. But it is nowhere near the 75% we see in Figure 1 in the September following the contract change in 2008. Second, the increase could have reflected a change in participation rates, but in fact we see no such change in the data. Third, the increase could reflect a shift in technology, with more workers using shears in the post-contract change period. However, the jump in productivity is present for each of the two technologies: hand and shears. Finally, there could have been local changes in weather patterns, specific to the plantation at the time of the contract change.

Figure 1 Raw output increase in the month following the new contract

[[{"fid":"7525","view_mode":"left","type":"media","link_text":null,"attributes":{"height":227,"width":315,"class":"media-element file-left"}}]]
Technically, we can deal with all such potential confounds in a multivariate regression, by accounting for field fixed effects, daily rainfall, time-varying technologies (shears versus hand), and the intensity with which the field has been harvested in the recent past. We still find a net increase in output of at least 20% over and above the control plantations and generally more, depending on the precise composition of controls and counterfactuals. By any standard, this is a remarkable increase. More to the point, it is not consistent with the traditional theory which – as we’ve observed – unambiguously predicts that productivity should weakly decline under the new contract.

But even apart from the obvious suspects aside, five other classical explanations could account for the increase we see.

  •  First, ‘dynamic incentives’ may be at work, as opposed to the ‘static incentives’ provided by piece rates. After all, given the better terms of the new contract, workers would presumably have more to lose from contract termination. This explanation doesn’t hold water because the temporary workers on the plantation are if anything less responsive to the contract change than their permanent counterparts, who cannot be fired.
  • Second, supervisors may have exerted more pressure on workers. This channel is a possibility: we show that it could account for up to a quarter of the increase we see.
  • Third, output was unusually low in the first few weeks of observation for the treatment plantation, so one argument is that the increase we observe just reflects ‘saving up’ on leaves and then chopping them off in larger quantities in the period after the contract change. But this argument is not valid: tea leaves become useless if they are not plucked in regular (shorter) intervals.
  • Fourth, one might hazard the guess that this is a learning effect of some kind. But that would be near impossible: most of the workers in this plantation have been plucking tea for decades.
  • Finally, the increased productivity could be the result of improved nutrition. But wages are paid only at the end of the month – long after the observed increase in productivity – and credit markets are highly imperfect.

This process of systematic elimination pushes us to conclude that what we see in the immediate aftermath of the contract change is indeed a ‘behavioural’ response. The data do not allow us to unearth the precise behavioural mechanism, but given the long tenure of workers at this plantation and the close relationship they share with management, it seems likely that gratitude and reciprocity played their parts here. After all, it is very likely that once the contract was sealed, the plantation owners made no bones about their apparent magnanimity. In any case, the bottom line is that the output increase was strong, and inexplicable from the perspective standard economic arguments.

A different picture emerges over the long run

But this is only the first part of our study. Indeed, many ‘behavioural’ papers on incentives stop at this point. An experiment is run, typically in a lab, sometimes in a somewhat contrived work setting (e.g. students are ‘hired’ to stuff envelopes), effort responses to various incentives are measured, and researchers conclude that the resulting patterns of behaviour contradict the predictions of a classical model, but are consistent with the predictions of a behavioural model. Workers care about fairness. They respond to generosity. They engage in reciprocal gift giving. At this point, the vindicated behavioural economist typically packs his bags and goes home. We would have probably come to a similar conclusion had we terminated our analysis a month into the new contract.

However, we were in the fortunate position of having data four months into the contract change, up until the end of the plucking season. The picture that emerges through these later months is rather different. Starting around the second month, a decline sets in until, four months after the contract change, output returns to pre-change levels. This is immediately clear from visual examination of the data, as well as parametric and semi-parametric estimates, which account for the control variables mentioned earlier.

Figure 2 Reversal to standard economic incentives four months out

[[{"fid":"7529","view_mode":"left_no_caption","type":"media","link_text":null,"attributes":{"height":215,"width":315,"class":"media-element file-left-no-caption"}}]]
Figure 2 extends the analysis to four months after the contract change. It plots output residuals relative to the pre-contract period, after accounting for rainfall and other controls. The horizontal line denotes the average residual right at the start of the contract in 2008. This panel then plots the daily residuals to Month 4. It shows that the Month 1 productivity response persists through the second half of Month 2, but then tapers off until, by the end of Month 4, output retreats to the levels in the week before the contract change.

This reversal is equally dramatic, as the entire process appears to complete an eventful but ultimately ephemeral round-trip. What, then, accounts for the long-run productivity response to the contract change? The answer, as it turns out, is a very simple classical model with static incentives. Structural estimates using this simple model show that in the short run, it is way off the mark. But by the end of month four, it does a remarkably good job of predicting the productivity response to the new contract.

Concluding remarks

Where does this leave us? Well, behavioural economics has captured public attention in recent years, and for good reason. We all know that people are tempered by altruism, morality or notions of justice. It would be foolish to ignore this possibility, and pretend that narrow-minded economism is some implacable law. Indeed, our own results show that behavioural economists make an important point even in this most obviously economic of transactions: manual wage labour. The question is not whether social considerations are present or whether they matter at all. It is whether behavioural models that capture those sentiments do a better job of predicting behaviour than workhorse neoclassical models.

Our results show that behavioural models may well have the upper hand in terms of predicting workers’ responses to incentives, but only in the short run. In the long run, however, the simple neoclassical models appear to do a better job. At the end of the day, which model is ‘right’ for prediction will depend on what you are interested in – the short run or the long run. But before we rubbish workhorse classical models and hurriedly classify important economic phenomena as fundamentally ‘behavioural’, we should at least study these phenomena over a longer period of time.

Photo credit: will- on-board @ flickr.

References

Costa-i-Font, J, M Jofre-Bonet and S Yen (2011), “Non-monetary incentives can overcome motivation crowding out”, VoxEU.org, 4 August.

Jayaraman, R, D Ray and F de Véricourt (2016), “Anatomy of a Contract Change”, American Economic Review 106(2): 316-358.

Endnotes

[1] We have also summarised this paper in an article for Ideas for India, and as such there is unavoidable overlap in the exposition here, and in that article.
 

Corporate Responsibility and Ethics, Leadership and Change

Add new comment