A city's population exceeds 1000?Don't plant it again!5 move away from 7 major data traps!

Original | Chen Ying

"× City's permanent population exceeds 1,000 people" (should be 10,000 people) "2021 net profit of 696.959 billion yuan" (should be 10,000 yuan) "year -on -year growth exceeds 70" (should be 70%) ...

Do you feel ridiculous when you see such a expression? These "harmful and insulting and strong" data errors are actually from the hands of the media. Some of them are on the newspaper, and some are widely spread on the network platform.

News report data errors have greatly hurt the credibility of the media. How to accurately control the data? What are the traps in all aspects of data acquisition and use?

In mid -June, Wang Tao, deputy editor -in -chief of Xinhua Finance and Economics of the Media Tea Society, Zhang Zehong, Director of the Surging News Data News Department, Chen Liangxian, reporter from the Data News Department, Editor Ren Jiao, editor -in -chief of the Creative Newspaper Department of the Beijing News, Li Zhen Essence

Data trap order the media frequently planting

In early June, the nostalgic chickens had not paid social security for nearly three years, and the social security chairman came out to apologize and clarify the social security payment situation, saying that the media had the problem of repeated calculation data.

News like a fellow chicken is not alone due to errors due to data.

In April of this year, a city in Henan issued a notice of economic operation in the first quarter, and the city's GDP growth rate ranked second in the province. A media directly wrote "GDP's second place in the province" into the title, allowing readers to mistake the growth rate second to the second. Soon after, the manuscript was deleted. In fact, anyone who is slightly familiar with the regional economy of Henan knows that the city's GDP has long been in the province for a long time, and there is still a small gap with the second place.

For another example, in March 2021, the first financial report released after the Kuaishou Technology was listed, and the report of "Kuaishou's net loss of 116.6 billion yuan in 2020" was brushed.

In fact, in the international financial accounting standards, preferred shares will be included in liabilities, and changes in the fair value generated are recorded as losses. The equity holder in the fast -handed financial report should account for 116.6 billion yuan during the year, including 106.845 billion changes in the fair value of convertible redeemable limited stocks. In other words, the operating loss of Kuaishou 2020 was 10.3 billion yuan, and the loss after adjustment was 7.948 billion yuan.

Li Zhen, the editor -in -chief of Dahe Cai Cube, told the media tea that the data was wrong in the news report. A large part of the reason is that some editors and reporters have a serious dependence psychology. Give the press publishing person. This is easy to lead to the "three trials, three schools" review process, and it is easy to cause factual errors, even political errors.

"Calculating errors, information folding, exaggerated details to cover up the facts, uncertainty in the hidden data, and the establishment of errors and causality through establishing false relationships, etc., make the data full of deception." Zhang Zehong, director of the surging news data news department, revealed that he was based on it, in order to use it to use it to use it. Data news reports at the core also have data deviations.

From the perspective of the editor -in -chief of the Beijing News Shell Financial Intelligent Media Creative Newspaper, data visualization can make financial data and news more simple and intuitive, and the situation of error use of data in data news may appear in any process.

As a senior financial media person, Wang Tao, deputy editor -in -chief of Xinhua Finance from the front -line reporters, has a deep observation of data in the news report.

"The cause of data errors is more diverse, the source of data is not relatively true, the awareness of authority is insufficient, the lack of professional knowledge, the randomness of data reference, and the blind obedience to experts and scholars may cause the data to be true." Media tea words will analyze.

What are the common data traps?

Why did there be data errors in the news report? What traps are there in data acquisition, identification, calculation, interpretation and other links?

In Wang Tao's view, common data errors are mainly common sense errors, data source errors, induction errors, professional errors, and other categories. There are traps in each link.

1. Common sense errors

For example, a certain article mentioned many years ago that "more than 500 million people in my country have their own family doctors", which has been separated from the daily feelings of the people.

2. Data source trap

"Our requirements for editing are‘ there is no word and no place ’. Of course, this sentence does not just refer to data.” In Wang Tao ’s view, this concept of traceability is very necessary.

"The most irreversible in news report is the data source error." Zhang Zehong said, for example, the interviewer may provide error data, especially the data of oral interviews may be wrong. "In addition, there may be too low data sampling, insufficient data coverage, etc."

Ren Jiao reminded that the channel for data acquisition is whether the official release, authoritative release channels, and data release time have been invalidated for too long. "Be alert to the error of the accuracy of data. If the data is not the data obtained from the real measurement, the statistics and scope are large."

"The quality of data is high and low, not all data is worth citing." In Wang Tao's view, data sources are important criteria for data quality. "Now there is a tendency to be abused, and some data released by institutions without credibility. There is no authority at all. As long as there is a gimmick, there will be media citations."

3. Data selection, cleaning trap

The differences between the media and statistical caliber selection of data expressions may also cause errors. For example, there are often total amount and new expression differences in economic data annual reports; the difference between data units should be careful, whether "10,000" or "billion". In addition, these three situations must be noticed. One is to summarize the data of different levels together, the other is to over -calculate the comparison rate or proportion, and the third is the mixed proportion and percentage.

4. Data analysis trap

Ren Jiao believes that when the media conducts data news reports, in the process of data collection, storage, association, transformation, and cleaning, the data and class levels may not match or mix error data due to incorrect forms and incorrect processing. In addition, there may be inconsistent or compatible fields such as measured units or dates. "The missing or duplicate brought by different data gathers may also change the original data distribution."

5. Data thinking inertia trap

"First to form a hypothesis to find data verification or overthrow assumptions, it is often easy to rush to believe that the assumption is correct, and give up the analysis or ignore the data that does not meet the hypothesis." Ren Jiao reminds that people are subjectively prefer to prove their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations, such as their original expectations. For example When the famous "Simpson Paradox", when the two sets of data were discussed, the main body of a set of data accounted for an advantage. Once the merger was merged, it caused the opposite conclusion.

In addition, "data noise" and "data strain" are also prone to occur during big data analysis and processing. If a certain data is too small or too large, it will affect the overall analysis results. Whether such a result can represent the whole. Actually make judgments.

6. Data visualization trap

In data news reports, the phenomenon of a large number of chart stacking is common to pursue the "cool" and aesthetics of the chart, which leads to the truly important data hidden, and a large amount of decorative elements are drowned for the real important information.

During the data visualization, the low -level errors of different coordinate axes are alert to the chart, and the use of colors and elements should be carefully used. For example, financial data often uses red represents the stock price to rise, green represents decline, while green in US stocks rises, and red is a decline. For example, the Winter Olympics logo is different from the Winter Paralympic LOGO and cannot be misused.

7. Data induction trap

Wang Tao reminded that the publisher of the data may have their own interests and demands, especially some institutions may create data for business purposes. A verification of data is a bitter issue, and editors are often misleading.

For another example, in an interview with some enterprises, we must prevent enterprises from being comprehensive and introduce to reporters to the most conducive to their statistical caliber data. Many industry associations will publish statistical data in accordance with the monthly, quarterly, semi -annual, and annual cycles, but different statistical dimensions and statistical calibers are often different. Press lack of experience will be selected by the company's selective reference data. Bringing the direction, leading to the not objective and comprehensive content of the report.

"Compared with text, the chart has more advantages in spreading. If you want to achieve the purpose of inducing readers, you will conclude that various parameters will not be consistent with the data itself." Chen Liangxian, a reporter from the surging news data news department, shared the sharing A case.

The following two pictures are maps of Phoenix News (left) and BBC (right) reporting the number of Chinese crowns. The data update time is on the afternoon of February 14, 2020. The BBC provinces with more than 500 cases of diagnosis have given the heaviest dark red, causing 529 cases of Chongqing and 51,986 cases of 5,1986 cases at that time. And Phoenix.com is divided by the 10th side, and Chongqing and Hubei are distinguished. Therefore, in addition to the incorrect map of the map, the map of the BBC also has the problem of inducing readers to misjudge the real situation of the epidemic.

Picture source: Surging News

5 Zhaoyuan data trap

How can we effectively avoid the common data traps and errors? The interview object gives 5 suggestions.

1. Correct attitude, awe data

"The news facts reflected in the data must be awesome, do not impose contact, do not interrupt conclusions, and keep their thinking clear." Ren Jiao said that recognizing the data must be a certain period of time and scope. Reflexes and data existence must have corresponding conditions.

2. Tour the initial source to ensure the authority authority

"The accuracy and objectivity of the data are always the first place in news reports." Li Zhen reminded whether the data comes from the official release channels; for the data published in quarterly and monthly data in economic data, verify whether it is the recently released data The information obtained from non -authoritative channels should be verified with the authoritative department before reporting.

He also mentioned that for the interviewees, we must make a judgment on their identity and position. "Even if there are some interviewees who have a certain identity, they cannot be blindly superstitious.

Zhang Zehong further added that sometimes "the data source obtained is relatively one -sided, not enough, but it cannot be supplemented or avoided." Sex. In the case of insufficient data disclosure, this is a compromise method that is not worthy of advocating but repeatedly. "

3. Strict editing process and establish a cross review mechanism

For data news reports, the regular audit method may be difficult to find the data trap, and the cross -review mechanism can be established. The key data in the assumption of the article is errors, pushing down the re -calculation, analysis, and then compare. Essence "In surging news, the availability of data is the first criterion for judging whether the data selection report can be established. When having data sources, repeated multiple times of data source verification is the most effective avoiding data source errors and data. The means of deviation. "Zhang Zehong said.

4. Improve professional ability and obtain corresponding qualifications

Li Zhen suggested that the reporter edit self -charging, and obtains the qualifications of securities, funds, banks, accounting and other industries to promote learning to avoid problems in various aspects such as data collection, analysis, prediction, and release.

It is understood that Dahe Cai Cube requires reporters to edit the qualifications of securities, funds, accounting, banking industry and other industries, encourage the qualifications of securities analysts, futures analysts, and even higher -level CPAs and CFA certificates.

In addition, when encountering more professional data analysis and reports in some industry fields, such as the requirements for accounting standards in the breeding industry, technology companies and other fields, they can seek the help of investment banks and professional analysts who are proficient in segments.

5. Into the tools cleverly

Good at using data analysis tools can also be added. For example, data analysis software such as Wind, Flush, Choice Data, Sky Eye Investigation, Corporate Early Warning and other data analysis software.

For another example, Xinhua Finance National Financial Information Platform also provides a lot of practical functions, including corporate equity penetration and some economic data models. In addition to the database itself, the platform also provides a large number of tools to help reporters carry out data statistics, collation, analysis, and mining of data.

Part of the picture in this article is provided by petals as beautiful

- END -