Tax Authorities Activity on Facebook
TARC BBS
Duccio Gamannossi degl’Innocenti
12-10-2017
Overview
- Introduction
- Data
- Pages and Users Activity
- Time Evolution of Activity
- Text Analysis (plain and Alchemy)
- Sentiment Analysis (Alchemy)
- Open Questions
Communication in the digital era
- Taxpayers perception of tax authorities is crucial for compliance
- Rising importance of social media in information retrieval
- Huge amount of good quality data available
- Various novel techniques to be applied
Research Goals
- Characterize tax authorities activity on FB
- Characterize users activity on tax authorities FB pages
- Analyse taxpayers’ engagement
- Analyse taxpayers’ sentiment (forthcoming)
- Topics extraction and text analysis
Data
- Sample restricted to English speaking authorities
- 30+ Pages considered
- 43,186 Posts - 127,714 Comments- 1,131,173 Likes
- Huge variability in population
The Dataset
| Name | Tot Post | Tot Likes | Tot Comments | Tot Shares | Base Volume | |
|---|---|---|---|---|---|---|
| 1 | UK | 4826 | 12360 | 12202 | 19853 | 44415 |
| 2 | UK-Graduates | 3290 | 6440 | 2577 | 632 | 9649 |
| 3 | Ghana | 259 | 8517 | 1252 | 1766 | 11535 |
| 4 | Nigeria | 84 | 71 | 37 | 4 | 112 |
| 5 | Namibia | 24 | 219 | 8 | 36 | 263 |
| 6 | South Africa | 871 | 16517 | 11081 | 111819 | 139417 |
| 7 | Botswana | 209 | 6399 | 915 | 2003 | 9317 |
| 8 | Zambia | 908 | 27924 | 1854 | 1536 | 31314 |
| 9 | Rwanda | 1038 | 1948 | 213 | 390 | 2551 |
| 10 | Uganda | 2343 | 38497 | 11061 | 8126 | 57684 |
| 11 | Malawi | 373 | 2721 | 1072 | 683 | 4476 |
| 12 | Africa | 1092 | 2761 | 152 | 676 | 3589 |
| 13 | Tanzania | 587 | 310058 | 7549 | 6929 | 324536 |
| 14 | Mozambique | 258 | 8136 | 682 | 4070 | 12888 |
| 15 | Kenya-KRACare | 1302 | 383328 | 43231 | 50225 | 476784 |
| 16 | Kenya-StoriZaUshuru | 263 | 53569 | 2230 | 4949 | 60748 |
| 17 | Pakistan | 492 | 4238 | 719 | 1303 | 6260 |
| 18 | India-CBEC | 442 | 4912 | 174 | 653 | 5739 |
| 19 | India-RevenueService | 655 | 40621 | 1493 | 8034 | 50148 |
| 20 | India-IncomeTax | 115 | 3584 | 356 | 449 | 4389 |
| 21 | India-itaxirs | 755 | 56577 | 1944 | 24825 | 83346 |
| 22 | Sri Lanka | 11 | 7 | 0 | 2 | 9 |
| 23 | Bangladesh | 2476 | 162494 | 2153 | 29926 | 194573 |
| 24 | Malaysia | 5705 | 700931 | 52575 | 173060 | 926566 |
| 25 | Singapore | 964 | 11239 | 351 | 5866 | 17456 |
| 26 | Singapore-Customs | 842 | 14524 | 1216 | 4601 | 20341 |
| 27 | Australia | 10246 | 216011 | 57292 | 1393168 | 1666471 |
| 28 | Papua New Guinea | 17 | 167 | 14 | 2 | 183 |
| 29 | New Zealand-Business | 258 | 2027 | 274 | 551 | 2852 |
| 30 | New Zealand-Student | 335 | 552 | 275 | 60 | 887 |
| 31 | US | 20 | 4145 | 25024 | 1464 | 30633 |
| 32 | Jamaica | 2034 | 1806 | 314 | 880 | 3000 |
| 33 | Sierra Leone | 92 | 1859 | 311 | 233 | 2403 |
| 34 | Average | 1309 | 63793 | 7291 | 56326 | 127410 |
Pages and Users Activity on Pages
Users Activity on Pages
Total Posts
Unique Active Users
Engagement
Time Evolution of Activity
Posts Time Evolution
Time Evolution by Post Type
Authority Day-time Activity (HMRC)
Users Day-time Activity (HMRC)
Coverage Day-time Activity (HMRC)
Lifetime of Posts and Users (HMRC)
Users attention patterns (HMRC)
Text Analysis
Wordcloud of HMRC Posts
Wordcloud of URA Posts
Comments WC, URA page
Word Freq. of HMRC Posts
Word Freq. of URA Posts
Word Freq. of HMRC Comments
Word Freq. of URA Comments
Digrams, HMRC Post
Digrams, URA Post
Digrams, HMRC Comments
Digrams, URA Comments
HMRC most associated terms
URA most associated terms
Sentiment Analysis - Alchemy
Comments Sentiment in Time
Concepts (Alchemy)
Keywords (Alchemy)
Open Questions
Work in Progress
Identify possible patterns in communication linked to economic development/complexity:
- Tax authority and Users activity (Evolution in time, Type of posts, etc.)
- Most frequent words and topics treated (cluster of words/digrams based on correlation?)
- Evolution in time of words more associated with tax authority
Investigate the effects of interaction between Users and Tax Authority
- How the reaction of users to interaction has evolved over time (any pattern?)
- Is it possible to identify if a particular style of interaction has better outcomes?
- Are there concepts more prone to backfire?
Thank You!
Questions?
Tax Authorities Activity on Facebook
TARC BBS
Duccio Gamannossi degl’Innocenti
12-10-2017
Comments WC, HMRC page