Path-goal theory and leader-member exchange in technical teams
Two classical leadership theories, path-goal theory and leader-member exchange, examined from the perspective of an engineer who often leads without formal authority, with two small models: motivation as a product of beliefs, and the unintended formation of in-groups.
Engineers tend to prefer rules that apply to everyone equally: the same process, review and standards. This appears fair, and in software it usually is.
People differ from code. One colleague needs clear direction on a new task, another needs autonomy, a third needs to see where the work leads. Treating all three identically is likely to be fair to only one of them. Parents of several children meet the same situation at home: one child needs help with homework, another needs to be left to work alone, and a third needs to understand why the work matters at all. The same rule for all three suits one of them.
The first module of my MBA asked students to test leadership theories against short cases and against their own work. Two of them were particularly useful because they explained patterns I had observed in engineering teams. This article examines why identical treatment fails and how unfair patterns arise in a team without anyone intending them. Each theory is illustrated with a small model.
The relevant literature is as follows.
Expectancy theory. Vroom (1964) proposed that the motivational force behind an effort is the product of three beliefs: expectancy (I can do it), instrumentality (doing it leads to an outcome) and valence (I value that outcome).
Path-goal theory. House (1971) built on Vroom: the leader's job is to clarify the path to the goal and remove obstacles, choosing a style (directive, supportive, participative or achievement-oriented) that fits the person and the task. Northouse's textbook (2016) is the standard summary.
Leader-member exchange (LMX). Graen and Uhl-Bien (1995) describe how each leader-follower relationship develops separately; some become high-quality partnerships, others stay transactional, creating in-groups and out-groups.
Cumulative advantage. Merton (1968) named the Matthew effect: early advantage attracts further advantage. Mathematically it is modelled by urn schemes, going back to Eggenberger and Pólya (1923), and by preferential attachment (Barabási and Albert, 1999).
Path-goal theory
When a capable engineer is not motivated, the usual response is more training or better tools. That strengthens expectancy only. As I wrote in one of the assignments: expertise improves what people can do, not whether they want to do it. If the work leads nowhere visible, or to something the person does not value, additional skill does not change motivation.
Engineers commonly value autonomy, interesting problems and a visible career progression. A leader who cannot offer a promotion can still offer the second and often the first.
Leader-member exchange
In technical teams, in-groups often form around whoever answers fastest. The person who replies within minutes gets the next urgent problem, then the next one, and soon all the visible work. Nobody decided this; it is the path of least resistance. Groups of friends show the same pattern: whoever organised the last gathering is asked to organise the next one, until organising has become that person's role without anyone choosing it. The simplest correction I know is to rotate high-visibility work deliberately.
There is also a resilience argument familiar from security: a team where escalation can only go to one person has a single point of failure. Distributing trust is fairer and also improves resilience.
The following models illustrate both points.
Model 1: motivation as a product
- Start with low valence (0.1) and press "Add training" a few times. Expectancy rises to 1, but motivation barely changes.
- Now press "Assign a valued problem". A single improvement in the weakest factor has more effect than all the training.
- Compare with the readout "if it were a sum instead". A sum compensates for a weak factor; a product does not.
Model 2: preferential assignment and rotation
In this model, six equally able engineers handle a quarter's worth of urgent tasks. Each task goes to someone with probability proportional to : the more someone has handled, the more likely they are to be asked again.
- Set α = 0. Work spreads evenly; inequality (Gini) stays close to zero.
- Set α = 1. This is a Pólya urn: the split is random, but whatever split emerges early tends to stick. Repeated runs give a different dominant engineer each time, with no difference in ability.
- Set α = 1.5. One engineer usually ends up with most of the visible work. The readout "people with ≥ 10% of the work" shows how few people could cover if that engineer were absent.
- Deliberate rotation. Keep α = 1.5 and rotate 40% of tasks. The distribution flattens: a modest deliberate policy counteracts a strong unintended dynamic.
Behaviour of the model. At , the classical Pólya urn result applies: the long-run shares converge, but to random limits that depend on early luck. For , results on generalised (non-linear) urns show that one option eventually takes almost everything. Neither outcome depends on any engineer being better than another.
Both theories lead to the same conclusion: identical treatment is not the same as fair treatment. Fairness lies in giving each person what they need to succeed, and being transparent about it. By the end of the module I had written down a working definition, which I expect to keep revising:
Leadership is building the conditions for capable people to make good decisions without me, one relationship at a time: clarity on the task, care for what each person values, and the judgement to know which one is needed.
The models add one practical point. Unfairness does not require bad intentions, because a feedback loop is sufficient to produce it. Breaking the loop requires deliberate rotation, which also makes the team more resilient.
References
- F. Eggenberger and G. Pólya, "Über die Statistik verketteter Vorgänge", Zeitschrift für Angewandte Mathematik und Mechanik, 3(4), 279-289, 1923.
- V. H. Vroom, Work and Motivation, Wiley, 1964.
- R. K. Merton, "The Matthew effect in science", Science, 159(3810), 56-63, 1968.
- R. J. House, "A path goal theory of leader effectiveness", Administrative Science Quarterly, 16(3), 321-339, 1971.
- G. B. Graen and M. Uhl-Bien, "Relationship-based approach to leadership: development of leader-member exchange (LMX) theory of leadership over 25 years", The Leadership Quarterly, 6(2), 219-247, 1995.
- A.-L. Barabási and R. Albert, "Emergence of scaling in random networks", Science, 286(5439), 509-512, 1999.
- P. G. Northouse, Leadership: Theory and Practice, 7th ed., SAGE Publications, 2016.