Understanding Factbase's Author-Level Research Metrics

Author-level research metrics assess individual researcher contributions to specific topics. This guide explains how publications are attributed to authors and how metrics are calculated at the individual level.

Author metrics require different methodological considerations than country or institutional analysis due to name disambiguation challenges, career mobility, and appropriate use constraints.

How Papers Are Attributed to Authors

Author Identification

Author attribution is based on byline authorship declared on publications.

Example:
Paper: "Quantum Error Correction Advances" Authors: J. Smith, A. Chen, M. Patel, L. Zhang, R. Kumar

Each author receives credit for this publication

---

### Name Disambiguation Challenge

**The fundamental problem**: Multiple people share the same name, and individuals publish under name variations.

**Example ambiguities**:

"J. Smith" could be:

• Jane Smith (MIT, quantum computing)
• John Smith (Stanford, quantum computing)
• James Smith (Oxford, different field entirely)

Same person, different names:

• "Jane A. Smith" (early career)
• "J. Smith" (mid-career)
• "Jane Smith" (later career)
• "Jane Anderson-Smith" (after marriage)

---

### Factbase's Disambiguation Approach

**Primary method**: ORCID matching

ORCID (Open Researcher and Contributor ID):

• Persistent digital identifier (e.g., 0000-0002-1825-0097)
• Author-controlled
• ~20% of global authors have ORCID

**When ORCID available**:

Author: Jane Smith ORCID: 0000-0002-1825-0097 → All papers with this ORCID attributed to same person ✓

**When ORCID unavailable** (majority of authors):

**Algorithmic disambiguation using**:
- Name string matching
- Institutional affiliation patterns
- Co-author network analysis
- Publication topic consistency
- Publication timeline coherence

**Example**:

"J. Smith" papers at MIT (2015-2024) in quantum computing

• Co-authors: frequently with A. Chen, M. Patel
• Topic: consistent quantum error correction focus → Likely same person

"J. Smith" paper at Harvard (2023) in marine biology

• Co-authors: different network
• Topic: unrelated to quantum computing → Likely different person

**Confidence levels**:
- High confidence: ORCID match or strong algorithmic signals
- Medium confidence: Partial signals, some ambiguity
- Low confidence: Common name, sparse data, flagged for caution

---

## Fractional Credit for Authors

### Multi-Author Papers

**All authors on a paper receive equal fractional credit** regardless of author order.

**Formula**:

Author_fraction = 1 ÷ (Number of authors)

**Example**:

Paper with 5 authors: Smith, Chen, Patel, Zhang, Kumar

Each author receives: 1 ÷ 5 = 0.2 fractional papers

**Note**: Factbase does **not** weight by author position (first, last, corresponding). All authors treated equally.

**Why equal weighting**:
- Author contribution conventions vary by field
- First author ≠ always primary contributor (varies by discipline)
- Standard practice in bibliometric analysis (OECD, CWTS Leiden)

---

### Author Paper Count

**Definition**: Sum of fractional credits across all papers.

**Example – Dr Jane Smith in Quantum Computing (5Y)**:

| Paper | Total authors | Smith fraction |
|-------|---------------|----------------|
| A | 5 | 0.20 |
| B | 3 | 0.33 |
| C | 2 | 0.50 |
| D | 8 | 0.125 |
| E | 1 | 1.00 |

Smith fractional papers = 0.20 + 0.33 + 0.50 + 0.125 + 1.00 = 2.16 papers

**Interpretation**: Dr Smith contributed the equivalent of **2.16 papers** to quantum computing (from these 5 papers).

---

## Author-Level TMCM

**Formula**:

Author_TMCM = Σ(Paper_TMCM × Author_fraction) ÷ Σ(Author_fraction)

**Example – Dr Jane Smith in Quantum Computing (2023)**:

| Paper | TMCM | Authors | Smith fraction | Weighted contribution |
|-------|------|---------|----------------|---------------------|
| A | 4.0× | 5 | 0.20 | 0.80 |
| B | 3.0× | 3 | 0.33 | 1.00 |
| C | 2.0× | 2 | 0.50 | 1.00 |
| D | 1.0× | 8 | 0.125 | 0.125 |

Smith TMCM = (0.80 + 1.00 + 1.00 + 0.125) ÷ (0.20 + 0.33 + 0.50 + 0.125) = 2.925 ÷ 1.155 = 2.53×

**Interpretation**: Dr Smith's quantum computing papers from 2023 received, on average, **2.53× the median citations** for the topic.

---

## Special Author-Level Considerations

### Issue 1: Career Mobility

**Challenge**: Authors move between institutions over time.

**Example**:

Dr Smith publications:

• 2018-2020: MIT (USA)
• 2021-2023: ETH Zurich (Switzerland)
• 2024+: University of Tokyo (Japan)

**Factbase approach**:
- Author profile aggregates **all publications** regardless of institutional affiliation
- Papers attributed to **institution at time of publication** for country/institutional metrics
- Author TMCM reflects **total body of work** across all affiliations

**Result**: Author metrics show career-long contribution, institutional metrics show contribution while affiliated.

---

### Issue 2: Early Career vs Established Researchers

**Challenge**: Metric interpretation differs by career stage.

**Early career researcher** (5 years post-PhD):

Papers: 8 fractional papers TMCM: 3.5× Top 1%: 25% of papers

Assessment: Strong start, high quality output

**Established researcher** (25 years experience):

Papers: 45 fractional papers (5Y) TMCM: 2.8× Top 1%: 8% of papers

Assessment: High productivity, sustained quality

**Different profiles, both valuable** – metrics must be interpreted with career context.

---

### Issue 3: Collaboration Patterns

**Solo vs collaborative authors**:

**Solo-focused author**:

10 papers as sole author = 10.0 fractional papers Average TMCM: 2.5×

**Collaboration-focused author**:

30 papers, average 5 authors = 6.0 fractional papers Average TMCM: 3.2×

**Comparison**: 
- Solo author: Higher fractional count per paper
- Collaborative author: Potentially higher quality (TMCM)
- Different research styles, both legitimate

---

## Author Metrics Summary

### Core Author Metrics

**Volume**:
- Fractional paper count
- Papers per year (productivity rate)

**Quality**:
- Mean TMCM
- Median TMCM
- Periodic TMCM (3Y, 5Y, 10Y)

**Excellence**:
- Top 1% paper count
- Top 10% paper count
- Excellence share (% in top tiers)

**Collaboration**:
- Average co-authors per paper
- International collaboration rate
- Unique co-author count

---

## Appropriate Use of Author Metrics

### ✅ Appropriate Uses

- **Research profile assessment** – Understanding author's contribution to a topic
- **Collaboration network analysis** – Identifying research partnerships
- **Topic expertise identification** – Finding leading researchers in specific domains
- **Trend analysis** – Career trajectory and evolving research focus
- **Aggregate analysis** – Studying patterns across many researchers

### ❌ Inappropriate Uses

- **Hiring decisions** – Author metrics insufficient for employment evaluation
- **Promotion/tenure decisions** – Narrow focus on publication metrics ignores teaching, service, mentoring
- **Individual performance rankings** – Gaming potential, field differences, career stage variations
- **Funding allocation** – Past metrics don't guarantee future productivity
- **Quality of person judgments** – Metrics measure output, not character or potential

**Critical caveat**: Author-level metrics are **one input among many** for understanding research contribution. They should never be the sole basis for decisions affecting individuals' careers or livelihoods.

---

## Confidence and Limitations

### Low Confidence Flags

Author metrics flagged when:
- **Name disambiguation uncertain** (<80% confidence in author identity)
- **Low sample size** (<5 papers in analysis period)
- **ORCID mismatch** (conflicting ORCID claims)
- **Career interruption** (gaps in publication history)

### Known Limitations

**Name disambiguation errors**:
- ~5-10% error rate even with best algorithms
- Common names more problematic
- Non-Western names may have higher error rates

**Incomplete coverage**:
- ORCID only ~20% of authors
- Historical papers (pre-2000) less reliable attribution
- Non-English publications less complete

**Collaboration bias**:
- Equal credit may undervalue/overvalue contributions
- Author order conventions vary by field
- Corresponding author often most responsible but not always

**Career stage bias**:
- Recent PhDs disadvantaged (limited output)
- Established researchers accumulate advantages
- Career interruptions (parental leave, illness) penalised

---

## Summary

**Author-level metrics** require special methodological care:

**1. Name Disambiguation**
- ORCID matching preferred (20% coverage)
- Algorithmic disambiguation for remainder
- Confidence levels flagged

**2. Fractional Credit**
- Equal credit to all authors (1 ÷ number of authors)
- No author position weighting
- Aggregates across career

**3. Context-Dependent Interpretation**
- Career stage matters
- Collaboration style varies
- Field norms differ

**4. Responsible Use**
- Appropriate for research profiling and network analysis
- Inappropriate as sole basis for individual decisions
- Combine with qualitative assessment

**Author metrics are one component of Factbase's research intelligence**, primarily useful for understanding research networks, topic expertise, and aggregate career patterns rather than individual evaluation.

---

*For related methodology guides, see:*
- *[Understanding TMCM: Topic Median Citation Multiple](#)*
- *[Understanding Fractional Credit in Research Output Attribution](#)*
- *[Understanding Country-Level Research Metrics](#)*
- *[Understanding Institution-Level Research Metrics](#)*

*For questions about author-level analysis or disambiguation methodology, contact the Factbase team.*