Debunking the Wage Penalty: Why Finding a Job via Friends is a Methodological Minefield

TL;DR

Is finding a job through a friend actually bad for your salary? A prominent study on Singapore’s meritocracy claimed so, finding a negative correlation between "personal contacts" and earnings. However, Gerhard Krug’s critique reveals that this "wage penalty" is likely a methodological artifact. By confusing how we search with how we find, and failing to account for why people use networks in the first place (self-selection), the original study may have misread the data entirely.

Context: The Meritocracy Paradox

In a meritocracy, success should be based on "what you know," not "who you know." Vincent Chua (2011) argued that in Singapore, social networks might even be damaging to one's career because institutions favor formal credentials over informal influence. His data showed that those who found jobs via contacts earned significantly less. But is the network to blame, or is there a deeper statistical ghost in the machine?

The Pain Point: The Search vs. Find Ambiguity

Krug’s primary critique centers on the Montgomery Caveat (1992). Most surveys ask: "How did you find your current job?" This is a Job-Finding Method.

The problem? If a worker uses multiple search methods (LinkedIn + Referrals + Job Boards), the one that ultimately "finds" the job is usually just the fastest one, not necessarily the one that yields the best wage.

The Sequential Search Theory Logic:

1. High-Offer Frequency (q): Networks might give you more job offers quickly.
2. Wage Distribution (F): Formal methods might offer higher average wages but fewer offers.

If you receive two offers—one from a friend and one from a recruiter—you pick the highest. But if you only received one offer from a friend because you are a "lower-tier" candidate, you take it. In the latter case, the network didn't lower your wage; your low market value simply made the network your only viable path.

Methodology: Where the Analysis Went Wrong

Krug identifies three critical areas where the original analysis fails to separate signal from noise:

1. Misinterpretation of "Multiple Methods"

Chua attempted to dismiss the multiple-methods problem by showing that few people found their job through a combination of methods. Krug argues this is irrelevant. The issue isn't whether people found a job through two doors at once, but whether they knocked on multiple doors during their search.

2. Information vs. Influence

Does a friend give you a "tip" (Information) or "pull strings" (Influence)?

• Information is valuable even in a meritocracy (knowing a vacancy exists earlier).
• Influence is where meritocracy creates barriers. By using a general "contact" variable, the original study conflates the two, making it impossible to see if the network was helpful for info but blocked for influence.

3. The Self-Selection Bias

This is the most damning critique. Authors often assume "everything else is equal" (ceteris paribus). However, if graduates with poor grades (unobserved in the data) are the ones who turn to personal contacts because they can't compete in the formal formal market, the negative coefficient for "contacts" is actually just measuring "poor grades" in disguise.

Note: In the original paper, the comparison between Model 1 (raw effects) and Model 2 (with controls) shows that while human capital controls reduce the negative effect of contacts, they don't eliminate it, leading Chua to claim "downward mobility." Krug argues this remaining effect is simply the unobserved human capital bias.

Experimental Insight: Alternatives to the Trap

Krug suggests three ways to fix this research:

1. Observe the Unobservable: Use longitudinal data to track wage changes for the same person across different search methods.
2. Structural Assessment: Follow Mouw’s (2003) approach—look at the structure of the network (size, education level of friends) rather than just the "finding method."
3. Search Strategy Indicators: Ask what people used to search, not just what worked. If a person used their network but found a job through an ad, the network still influenced their "reservation wage" (the minimum salary they would accept).

Note: A re-evaluation of the 2005 Project Network Survey data suggests that using more direct questions like "Was there someone in the company who helped you?" provides a much cleaner measure of network influence than the generic "how did you find this job?" question.

Critical Analysis & Conclusion

Krug’s critique is a masterclass in econometric skepticism. It reminds us that in social science, a correlation between a behavior (using a network) and an outcome (lower wages) is often a reflection of the reason for the behavior rather than its consequence.

Key Takeaways:

• Context Matters: In meritocracies, networks provide info, not just favors.
• Selection Bias is King: People with fewer formal options use informal ones; researchers must account for this "lack of resources" motivation.
• Policy/Product Implications: Job platforms (like LinkedIn) shouldn't just track "hires via referral" as a success metric without adjusting for the candidate's initial market position.

Limitations: Krug’s critique is theoretical; he does not re-run the Singapore data with a full structural model (due to data limitations), but he provides the blueprint for how it should be done.