Documentation
Construction of the Soft-Power Shock (SPS) series, version 1.0. Same content as CONSTRUCTION.md; column definitions are in the codebook.
1. What the shock is
A giant-panda birth in a foreign zoo is a dated event whose timing is set by biology. Implantation of the embryo is delayed by a variable amount and pseudo-pregnancies are common, so the month of a birth is hard to predict, both for the public and often for the keepers. The first announcement comes from the zoo, and no government decides anything on that date. The birth raises the host public's attention to the panda, the best-known symbol of China. The soft-power shock (SPS) is an exogenous rise in a foreign public's exposure to a country's soft-power asset that carries no change in policy. It is identified from the births of giant pandas at foreign zoos and measured as the dose of attention each birth draws; the doses are added up to a host × month SPS series.
Hosts (24): AT, AU, BE, CA, DE, DK, ES, FI, FR, GB, HK, ID, JP, KR, MO, MX, MY, NL, QA, RU, SG, TH, TW, US; monthly, January 2004 to December 2025.
2. Event database
events.csv records 491 dated events in the lives of the pandas on loan to the 24 hosts,
1994–2025: loan announcements and agreements, arrivals, debuts, pregnancies, births, failed
pregnancies, namings, cub debuts and departures, extensions, returns, recalls and deaths. Each row has
- the physical date of the event and the date of its first public announcement: a zoo release, a ministry statement or the first wire story. The announcement date is day 0 for attention. It is taken from dated sources, never from the peak in the search data;
- a bundling code: \(+1\) if a state visit, summit or major agreement between the host and China took place within ±7 days of the announcement, \(-1\) if a diplomatic conflict, sanction or punitive recall did, and \(0\) otherwise;
- at least one source URL (zoo release, wire service, national broadcaster; 55 rows rest on a Wikipedia page only, which the notes say).
Bundling is common for government decisions and rare for biology: 18 of the 25 loan announcements are bundled, against one of the 67 births.
3. Daily attention
Source. Google Trends, topic entity Giant panda (/m/03bj1), which
pools searches in every language and script; the topic China (/m/0d05w3) is the
comparison series. The monthly indices \(A_{i,t}\) cover 2004-01 to 2025-12. For each event announced on
day \(d_e \ge\) 2004-02-01 with day precision, a daily window runs from \(d_e-60\) to \(d_e+120\). Windows
are pulled for the national panda topic, the national China topic and, where Google reports it, the panda
topic in the region of the zoo.
Rescaling. Google scales each request to a maximum of 100. Let \(v_{w,d}\) be the raw value on day \(d\) of window \(w\) and \(\bar v_{w,m}\) its mean over month \(m\), using months with at least 20 days in the window. The scale factor is the slope through the origin of the monthly index on the window's monthly means:
\[ s_w = \frac{\sum_m A_{i,m}\,\bar v_{w,m}}{\sum_m \bar v_{w,m}^2}, \qquad \tilde v_{w,d} = s_w\, v_{w,d}. \]It requires at least three overlapping months and a window that is not all zeros.
Standardisation. The host standard deviation \(\sigma_i\) is computed from \(\tilde v_{w,d}\) pooled over the usable windows of the host and topic, on non-event days only, that is, outside \([-10,+40]\) of every dated event of the host. A window is usable if fewer than half of its days are zero and it passed the scale check. Attention in host-SD units is
\[ a_{i,d} = \tilde v_{w,d} / \sigma_i . \]One unit means the same thing in every host: one standard deviation of that host's ordinary daily fluctuation in interest in the panda. Regional series get their own \(\sigma\). Regional Taipei City windows are not used.
4. Event intensity
For event \(e\) in host \(i\) announced on day \(d_e\),
\[ n_e = \frac{1}{3}\sum_{d=0}^{2} a_{i,d_e+d} \;-\; \frac{1}{5}\sum_{d=-7}^{-3} a_{i,d_e+d}. \]The post window covers late-evening announcements, which reach Google's UTC data the next day. The
pre window ends on day −3 to keep leaks out of the baseline. Each event has three doses: national
(dose_national), best-of (dose_best: the zoo's regional series if usable,
otherwise national) and China topic (dose_china).
5. The monthly SPS series
With \(\mathcal B_{i,t}\) the births announced in host \(i\) in month \(t\) that pass the inclusion rules,
\[ z_{i,t} = \sum_{e \in \mathcal B_{i,t}} n_e^{\text{national}}, \]and \(z_{i,t}=0\) in months without such a birth (z_birth_nat, the SPS (national dose), the headline). The variants
are
z_birth_reg, the SPS (regional dose): the same sum with the best-of dose;z_birth_count, the SPS (count): the number of births;z_birth_cap90, the SPS (capped dose): the headline series capped at the 90th percentile of its nonzero 2011–2025 values (15.70);z_death: minus the summed doses of deaths;z_hard: the signed intensity of hard political events (Section 6).
Inclusion rules
- Not bundled. Biological events with a bundling code \(\neq 0\) are excluded: one birth (Berlin 2019, four days before Chancellor Merkel's visit to China), three deaths and three namings.
- Not a later media day. Rows recording a later media date of the same cub are excluded: the first veterinary examination, the sex reveal, the first photograph or video.
- Measurable dose. The announcement has day precision and falls on or after 2004-02-01, and the national window is usable (fewer than half of its days zero, scale check passed).
- Not a measurement zero. Biological events at Adventure World (Wakayama) and Kobe Oji Zoo announced before 2017 are treated as missing doses: Google's topic barely captured Japanese-script searches before 2017. Births at Ueno Zoo in Tokyo stay in.
Sample. The paper's instrumental-variable sample starts in 2011-01, because most earlier
birth windows are too sparse to measure a dose. The monthly file starts in 2004-02 so that users can
choose; every z_* is missing in 2004-01. Over 2011–2025 the headline series has
33 nonzero host-months in 14 hosts, with a median dose of 3.08.
6. Hard political events
hard_events.csv lists 98 dated events between China and the hosts, plus a few other
countries. They are coded \(-1\) for conflicts (54: meetings of heads of state or government with the
Dalai Lama, sanctions and boycotts, Taiwan-related and territorial disputes, arrests, human-rights
conflicts) and \(+1\) for rapprochement (44: summits, state visits, major agreements). Their intensity
\(n_k^{\text{China}}\) is measured like a birth's, on the China topic. It is missing if half or more of the
window's days are zero.
The counts use every in-panel hard event, with or without a measurable intensity. Governments choose these events, so they are not exogenous; in the paper they serve as a positive control.
7. Companion files
attempts.csv: 144 announced breeding attempts with outcomes and the paper's success / fail / drop assignment.placebo_animals.csv: 17 non-China animal events (Knut, Harambe, Marius, Moo Deng, ...) with the intensity on their own Google topic.policy_regimes.csv: 139 dated travel and trade regimes (Chinese group-tour bans, visa waivers in both directions, flight suspensions, FTAs) with sources.attention_daily_windows.csv,attention_spillover_windows.csv: the daily data behind every dose.
8. Known measurement issues
- Japan before 2017. Attention to the panda (monthly Google index) averages about 6 before 2017 and about 44 after, with a permanent jump at the birth of Xiang Xiang (June 2017). This reflects topic coverage of Japanese-script searches, not interest, and is the reason for inclusion rule 4. Taiwan and Macau show milder versions.
- Spain and Mexico, 2004–2006. The panda topic is very high and then decays,
most likely because searches for the antivirus company Panda Security were mapped into it. This affects
the early level of
A_pandabut no event in the 2011+ SPS series. - Google collection changes in 2016-01 and 2022-01 appear as small steps in several monthly series.
- Single pulls. Google Trends values come from a sample of searches and differ slightly between pulls; v1.0 uses one pull per series.
- Sparse regional data. Regional windows are usable for 157 of 351 events.
- Low-confidence dates. Japanese and Mexican announcements before 2011 are often copies of the birth date. None is in the 2011+ SPS series.
9. Reproducing a dose
import pandas as pd
w = pd.read_csv("data/attention_daily_windows.csv")
w = w[(w.event_set == "panda_event") & (w.topic == "panda")]
post = w[w.rel_day.between(0, 2)].groupby("event_id").value_std.mean()
pre = w[w.rel_day.between(-7, -3)].groupby("event_id").value_std.mean()
dose = post - pre # equals events.csv dose_national
10. Example: panel local projection
\[ y_{i,t+h} - y_{i,t-1} = \alpha^h_i + \lambda^h_t + \beta^h z_{i,t} + e_{i,t+h}, \qquad h = 0,\dots,24, \]with standard errors clustered by month: lp_example.py (Python, statsmodels) and lp_example.do (Stata, reghdfe). Outcome data are not distributed; download_outcomes.md lists the sources.