# Repost: Writing Advice




> Archiving material I think is good, including advice on literary aesthetics and economic research.
>
> Under the advance of economic imperialism, no one knows where economics is heading. Some say economics is a skill for slaying dragons, but this world has no dragons to slay; some say economics is over-mathematized and has long drifted from the reality of the world; some say economists only use assumptions to solve problems... Positivism, post-positivism, constructivism — research paradigms crisscross without a single certain answer. Today, economics increasingly resembles the fable genre of constructivism, using data and examples to tell well a fable we need. From this angle, economics and literature have quite a few similarities, so here I collect and arrange articles I like, as reference.

## Literature

### Maugham's "Writing Advice"

1. You must become bored before your audience does.

2. The writer's motto must be: cut what can be cut, cut again, straight to the peak of the audience's attention.

3. What we writers must study is still the ordinary person. Ordinary people are the writer's more fertile soil.

4. He preferred strong, forceful words to pleasant-sounding ones.

5. Only years later did I understand that writing is a delicate art, acquired only through hard work.

6. It is useless to strive at something for which one has no talent.

7. Surprised at the poverty of my vocabulary, I took paper and pen to the British Museum, wrote down the names of curios, the Byzantine colors of old enamel, the feel of fabrics against the body, and then carefully devised sentences to put all of it into. Fortunately, I found no occasion to use this material; it still lies in that old notebook, ready for anyone who wants to write nonsense.

8. I wanted to cast aside ornate phrasing and write as plainly and unaffectedly as I could.

9. I discovered my own limits, and the only sensible thing for me was to set my goal at the best I could achieve within those limits. I knew I had no gift for lyricism.

10. I thought to myself that I should aim at clarity, simplicity, and euphony.

11. In a writer's work one can find two kinds of obscurity, one arising from carelessness, the other deliberate. People write obscurely often because they have not taken the trouble to learn how to write clearly. Another cause of obscurity is that the author is not quite certain of his own meaning.

12. Many writers do not think before writing but only as they write — the pen itself gives birth to the thought. This is indeed a danger the writer must constantly guard against.

13. It is easy to persuade oneself that a word one does not quite understand contains far more meaning than one realizes.

14. To achieve clarity requires strict training.

15. Simplicity and naturalness are the truest marks of excellence.

16. Euphony is the last of the three qualities I mentioned, and a good many readers, as well as many respected writers, lack it.

17. Words have weight, sound, and appearance; only by taking these into account can you write sentences that are graceful and pleasing.

18. I do not write as I wish; I write as I am able.

19. A far-fetched, worn-out, even affected word is not unsuitable in my eyes as long as it sounds better than a plain, clear one, or gives a sentence better balance. Though you may yield to pleasing sound without hesitation, you should never give way to a word that blurs the meaning.

20. I think the author had best have a stronger faculty of boredom than the reader, so that he senses boredom before the reader does.

21. Once people labor to form a style, they can rarely afterwards write with complete freedom.

22. If you can write clearly, simply, euphoniously, and vividly, you will write perfectly; you will write like Voltaire.

23. Good style shows no trace of effort; what you write should look as though it came to you by a happy accident.

24. If in the end I achieved an effect of ease, it was only through hard effort.

25. People should write in the manner of their own time. Language is living and changing; striving to write like authors of the distant past can only lead to unnaturalness.

26. I would rather a writer be worldly than affected; for life is worldly, and what the writer pursues is life.

27. In my view, many writers do not observe life at all but merely create their characters, as they already are, out of the many figures of their imagination. They are like draftsmen who draw patterns from memory of antiques and never try to draw a living model. At best they can only give their inner fantasies a plausible shape.

28. The stupid prejudice of the intellectual lies in thinking that only their knowledge matters. Truth, goodness, and beauty are not the privilege of those who attend expensive schools, haunt libraries, or frequent museums. The artist has no excuse for thinking himself superior. If an artist thinks his knowledge is more important than others', he is a fool; if he cannot meet others happily on an equal footing, he is a blockhead.

29. The writer may rest assured that the works he himself wishes to forget will also be forgotten by his readers.

30. I put books aside only because I realized that time was passing and life was my proper business.

### Camus's 1957 Nobel Speech, "The Artist and His Time"

Your Majesties, Your Royal Highnesses, ladies and gentlemen:

Your Academy, in the spirit of freedom, has generously bestowed this honor on me, and since I feel my achievements fall far short of its weight, I am all the more sincerely grateful. Everyone longs for recognition, and the artist more than anyone. I am no different. Only after comparing the effect of your decision with my real self did I truly understand why you made such a decision. A still-young man who has nothing but doubts, whose work is not yet formed, who is used to living alone in his work and shrinking from every friendly advance — for such a recluse, suddenly seized and thrown into this dazzling spotlight, how could he not feel a kind of panic? When the other writers of Europe, even some of the greatest among them, are forced into silence, when his homeland is suffering endless affliction, with what feelings is he to accept this honor?

I experienced just such inner dread and unease. To regain calm, I could only accept this generous gift of fate. Since my achievements cannot match this award, I can only lean on the conviction that has sustained my life and that I have never abandoned even in the hardest circumstances: my view of my art and of the writer's role. Allow me, with gratitude and friendship, to express this thought to you as briefly as I can.

For me, without art I could not survive. But I have never placed this art above everything else; on the contrary, it is indispensable to me precisely because it is closely bound to everyone and allows someone like me to live as everyone does. Art, in my view, is not a solitary celebration. Art is a means that, in its own way, presents the common suffering and joy of humanity and thereby moves the majority of people. So it forces the artist not to isolate himself and makes him submit to the simplest and most universal truth. And usually, the person who chooses an artistic career thinking himself different soon finds that only by acknowledging what he shares with all beings can his art and his distinctive talent draw nourishment. It is in this constant coming and going between himself and others, in his dealings with the beauty he cannot set aside and the community from which he cannot be detached, that the artist tempers himself. That is why true artists despise nothing; they demand of themselves that they understand everything, not judge everything. If they must choose a side in this world, they can perhaps belong only to the kind of society built in Nietzsche's great words: a society ruled by creators rather than judges, whether those creators are laborers or intellectuals.

Likewise, the writer's role is burdened with a difficult duty. As a writer, in times like ours, he should not serve those who make history; he should serve those who endure it. Otherwise he will be isolated and will also lose his art. A writer who keeps company with a tyrant cannot escape that loneliness even if the tyrant marches with a host of thousands. But at the other end of the world, a nameless prisoner abandoned to humiliation — his silence is enough, again and again, to rescue the writer from this lonely exile, as long as, while enjoying the rights of freedom, he never forgets that silence and, through the means of art, makes it resound.

None of us is great enough to take up this mission. But in all the circumstances of his life, whether neglected or celebrated for a time, whether crushed under the yoke of tyranny or enjoying a momentary freedom of speech, only by serving truth and freedom loyally and to the best of his ability can the writer's profession become great and can he win from the people a vindication that comes from the heart. The writer's mission is to unite as many people as possible; this mission must not yield to lies and servitude, for in lands ruled by lies and servitude, lonely souls are imprisoned everywhere. Whatever weaknesses we have as individuals, the nobility of our profession is forever rooted in two promises that are not easy to keep: never to lie about what one knows, and to resist all oppression to the end.

Through more than twenty years of an absurd journey, isolated and unaided, I and my contemporaries were lost in the upheavals of the age, sustained only by a faint inner feeling: that in today's world writing is an honor, because this act bears a mission and compels you not merely to write. Above all, it compels me, in my own way and with my own strength, to share with all who lived through that history as I did the suffering and hope we held in common. These people were born at the start of the First World War; they were in their twenties when Hitler's regime was established and the first revolutionary waves rose. Then, as if to complete their experience, they lived through the Spanish Civil War and the Second World War; they lived through that Europe scarred everywhere by concentration camps and prisons; and now these very people must, under the threat of destructive nuclear weapons, raise their next generation and fulfill their mission. I think no one has the right to demand optimism of them. I would even argue that, while continuing to struggle against them, we should understand what they do: only because of their growing despair did they commit shameful acts and fall into the nihilism prevalent in this age. Yet, whether in our country or across Europe, most of us still refuse nihilism and are still searching for a kind of justice. We need to forge an art of living through troubled times, so that we may be reborn from the ashes and then calmly fight the death instinct in the course of history.

Perhaps every generation is confident that it bears the mission of reshaping the world. Yet our generation knows that we are powerless to do so. But our generation's mission is perhaps greater, because our mission is to keep this world from falling apart. What we have inherited is a broken history, mixed with failed revolutions, technology gone mad, dead gods, and exhausted ideologies. Even in such an age, any mediocre force can destroy the world, but that mediocre force has only the power of negation, and when reason degrades itself into a slave of hatred and oppression, that power of negation cannot teach our generation to rebuild, in our hearts and in the outside world, a little of what can give dignity to life and death. Faced with a world that might collapse at any moment, the great judges among us will perhaps forever build kingdoms of death; but our generation knows that we should, while racing frantically against time, build among different peoples a peace that bows to no servitude, restore the relationship between work and culture, and join hands with everyone in the world to form an alliance. No one can be sure whether our generation can complete this vast task, but we are sure that they are already spread across every corner of the world, fighting for truth and freedom and ready at any moment to die for it, without complaint or regret. It is these people who deserve our respect and encouragement, whenever and wherever — and especially where they have fallen. In sum, I want to transfer the honor you have just conferred on me to them, and I believe you will feel the same.

At the same time, having spoken of the nobility of the writer's profession, I want to restore the writer's true appearance: apart from the identity he shares with his comrades, he has no other. He is at once fragile and stubborn; he cannot remain just forever, yet he eagerly pursues justice; before everyone's eyes he silently builds his work, neither ashamed of it nor proud of it, endlessly torn between pain and beauty, ultimately in order to distill, from this double existence, what he stubbornly wants to create out of the ruins of history. That said, who can still expect him to offer ready-made answers and perfect moral creeds? Truth is mysterious, elusive, always still to be conquered. Freedom is certainly exhilarating, but practicing it is likewise dangerous and difficult. We must move toward these two goals, through hardship and a long journey, yet unwaveringly and with undying resolve. That being so, what writer with self-knowledge would still dare to set himself up as a preacher of virtue? As for me, I must say once more that this is not at all my identity. I have never been able to give up the light and happiness of life, nor the free life, things that have accompanied me since childhood. Although this nostalgia has led me into many mistakes, it has undoubtedly also helped me better understand my profession and helped me stand without hesitation beside those silent people — those who, apart from retrieving from memory that little bit of brief, free happiness, have nothing else to go on in this world.

Now that I have shown you my true self, you know my shallowness and limits, know how much I owe to others, and know my hard-won conviction. In closing, I can at last express more freely the breadth and generosity of the honor you have conferred on me, and can more freely say to you that I accept this honor and will regard it as a tribute to all those who, like me, fought without receiving any distinction and only endured suffering and persecution. Finally, from the bottom of my heart I thank you, and openly, with gratitude, make to you an ancient promise — the ancient promise that every true artist makes to himself in silence each day — and that is: fidelity.

### Chuck Palahniuk's (author of Fight Club) Writing Guide

In six seconds, you'll hate me.
But in six months, you'll be a better writer.

From now on — for at least the next six months — you may not use "thought" verbs. These include: think, know, understand, realize, believe, want, remember, imagine, desire, and the hundred others you love to use.

The list also includes: love and hate.
It also includes: be and have, but let's get to those two in a moment.

Until around Christmas, you can't write: Kenny thought Monica probably didn't like him going out at night...
Instead, you have to unpack that sentence into: "Some mornings Kenny stayed out late, missed the last bus, until he had to hitchhike or pay for a cab, and came home to find Monica pretending to be asleep, because she never slept that quietly; those mornings she would only put her own cup of coffee in the microwave. Never his."

Instead of writing what your character knows, you must now provide details that let the reader know those things.
Instead of writing that a character wants something, you must now describe that thing so the reader wants it too.
Instead of saying: "Adam knew Gwen liked him." you have to say: "Between classes, when he went to open his locker, Gwen was always leaning against it. She'd roll her eyes, push herself off with one foot, leaving a black heel print on the painted metal, but she also left her perfume. On the combination lock there was still the warmth of her hip. Next period, Gwen would be leaning there again."

Simply put, there are no shortcuts anymore. Only concrete sensory details: action, smell, taste, sound, and feeling.
Typically, authors use these "thought" verbs at the start of a paragraph (you can call them "thesis sentences" here; I'll complain about that in a bit). In a sense, they point out the whole paragraph's intent. What follows explains those intents in detail.

For example: "Brenda knew she couldn't make the deadline. She was coming from the bridge, past eight or nine exits. Her phone was dead. At home, the dog needed to be walked or it would make a mess to clean up. Plus she'd promised to water the neighbor's flowers..."

Do you see how that opening thesis sentence steals the thunder from the paragraph that follows? Don't do that.
At the very least, cut the opening line and put it after all the other sentences. Better yet, move it somewhere else and change it to: Brenda couldn't make the deadline.

Thinking is abstract. Knowing and believing are invisible. If you show only physical actions and the details of the character, and let your reader do the thinking and knowing, as well as the loving and hating, your story will always be stronger.

Don't tell your reader "Lisa hated Tom." Instead, argue your case like a courtroom lawyer, speak through details, and present every piece of evidence.

For example: "At roll call, in the breath after the teacher called Tom's name, in the instant before he answered, right then, Lisa would whisper 'jerk.' Just as Tom said 'here.'"

One of the most common mistakes of beginning authors is leaving their characters alone. When you write, you can be alone. When they read, your readers can be alone. But your characters should very, very rarely be alone. Because a lonely character will start thinking, worrying, or wondering.
For example: "While waiting for the bus, Mark started to worry about how long the trip would take..."

A better way to show it: "The schedule said the bus would arrive at noon, but Mark's watch showed it was already 11:57. You could see to the end of the road, all the way to the mall — not a single bus. No doubt the driver had stopped at the transfer station, at the far end of the route, and was dozing. The driver was relaxed and sound asleep, and Mark was about to be late. Or earlier, the driver was drinking, and he'd pull over and charge Mark seventy-five cents to give him a gruesome death in a serious traffic accident..."

A character who is alone must enter fantasy or memory, but even then you can't use "thought" verbs, or any of their abstract-verb relatives.
Oh, and you can forget the words "forget" and "remember."
No more transition sentences like: "Wanda remembered how Nelson used to comb her hair."
Again, unpack it. Don't take shortcuts.

Better yet, have your character meet another character. The sooner the better.
Keep them together, and get them acting. Let their actions and talk reveal what they think. Don't — you — get inside their heads.

While you're avoiding "thought" verbs, pay special attention to those two ordinary verbs, "be" and "have."
For example: "Ann's eyes are blue." "Ann has blue eyes." Replace them with:
"Ann coughed, raised a hand and waved it in front of her face, shooing the cigarette smoke away from her eyes, blue eyes, and smiled..."

Instead of using flat "be" and "have" statements, try burying the details of what a character is and has into action or gesture. At bottom, this is showing rather than telling your story.
Much later, once you've learned to unpack your characters, you'll hate lazy authors who settle for: "Jim sat by the phone, wondering why Amanda hadn't called."

Please, at this moment, hate me however you like, but don't use "thought" verbs. After Christmas, write however you like — but I bet you won't.

...This month's assignment is: read through your work and circle every "thought" verb. Then find ways to kill them. Kill them by unpacking them.
Then read some published novels and do the same thing. Show no mercy.

"Marty imagined the fish, leaping in the moonlight..." "Nancy remembered the taste of the wine..." "Larry knew he was a dead man..."

Find them, and then find ways to rewrite them so they become stronger.

## Economics

### Advice on Choosing a Topic

[How to choose a topic for empirical economics research?](https://core.ac.uk/download/41460312.pdf)

— He Xiaobo, Shanghai Finance University

[How to get started on research in economics](https://econ.lse.ac.uk/staff/spischke/phds/How%20to%20start.pdf)?

— Steve Pischke (2009)

You can also look here: [Resources for Economics PhD Students](https://econ.lse.ac.uk/staff/spischke/phds/)

### Samuelson | Why Am I the One Who Wrote Economics?

Source: [Tales of the Economics World](https://mp.weixin.qq.com/s/yVPlnaUUu4QI23t4Nl44Kw)

> This article was written by Samuelson for the fiftieth anniversary of the publication of Economics, from Scientific Papers of Paul Samuelson No. 587. **Translator: Lü Ji'er**, forwarded by teacher Lai Jiancheng. Wang Laohu revised parts of the text, hoping not to have run counter to the translator's original meaning. It is shared here because I like the article; the original-protection mark is added to protect teacher Lü's rights as translator. Teacher Lü invites friends to offer corrections so the translation can be improved. For your corrections, the translator expresses heartfelt thanks! **Translator's email: zjjier@163.com**

#### Economics Turns Fifty

Economics has reached its fiftieth birthday, and over this half century it has witnessed the considerable development of economics. Economics has come a long way, yet to this day it has no right to boast of having reached the state of an exact science — as is probably true of any social science. Economics is forever a discipline that merges science and art, and there is a world of difference between a vigorous economics and a pedantic, outdated one.

Today, the things economists can do are increasing day by day, whether in government, the private sector, or academia. Compared with the late 1940s, the volume of data in statistical databases has grown explosively. Now computers can obtain and process statistical information and all kinds of knowledge at the speed of light, and we can obtain information quickly. Whether theoretical or applied economics, research in political economy has benefited enormously from advances in computer hardware and software — more, perhaps, than any other field.

Even so, the foundational areas of economics have barely moved forward. The world has learned some economic principles from history (at least from economic history) and has studied and tested them diligently. Economists treat fickle economic slumps and the inflation that produces false prosperity as laboratory mice, conducting controlled experiments like chemists in a lab. (The translator omits one sentence here.) At the same time, the perfectly competitive free market has failed to bring humanity equality of opportunity and equality of income.

The fact that "poverty is always with us" forces democracies around the world to evolve into a "mixed economy" — neither a laissez-faire market nor a Robin Hood utopia. Only after studying the basic principles of economics can the public understand and be convinced of how to strike the right middle course between self-interested individual initiative on the one hand and government regulation, the maintenance of stability, and redistribution on the other. Willingly or not, a mixed economy must be a "limited mixed economy."

#### Looking Back

A fiftieth birthday is an excellent occasion for looking back. What story can we see in the popularity of this best-selling textbook, which has led the field? How was this textbook born? We have to pull our gaze back to 1945. The Second World War ended with the defeat of Germany and Japan, and the people returning home filled American universities to overflowing. At that time, economics happened to be entering a golden age. The Great Depression of 1929 to 1935 turned economics around, casting aside the old ideas that had once held orthodoxy and replacing them with ambitious plans, with strong monetary and fiscal policies. Britain and the United States adopted new ways of mobilizing their economies for war, something Hitler, Mussolini, and Hirohito could never have dreamed of. At the time, we could not yet know of the later Marshall Plan and the American occupation of Japan, but they were about to set the stage for decades of postwar economic growth.

Students should understand all of this. Yet we teachers of that generation all knew that the best-selling economics textbooks of the time were seriously outdated, which was deeply distressing. No wonder beginners generally felt bored with economics. When I taught at Harvard and MIT, I often saw the glazed-over expressions on my students' faces.

When Economics finally appeared in 1948, "macroeconomics" — the study of the factors determining social employment, the price level and inflation rate, real GNP growth, and such questions — was a word not yet in the dictionary. By then, even if I had not written guns and butter into microeconomics, someone would certainly have done so. It was time for the introductory economics course to be reborn. Yet the real question is: why was that person me?

At that time I had just returned to the economics department from the MIT Radiation Laboratory. At the Radiation Lab I had done mathematical research on designing automatic servo mechanisms to evade enemy bombers, and I "longed" to return to economics research and teaching. I was exactly 30 that year, in the prime of life, a good time for writing a textbook and innovative papers. It was also around then that my earlier work, Foundations of Economic Analysis, had been published — the book that won me the Nobel Prize in Economics twenty-five years later.

In those days, promising scholars basically did not write textbooks, certainly not the basic texts used by first- and second-year students; only ghostwriters did that. But considering that I had already published so many research papers, and that my reputation and my prospects for tenure gave me room to maneuver, I enthusiastically responded to the MIT department chair's request for me to write a new textbook. Out of overconfidence, even some conceit, all I needed to do was satisfy myself. I thought it was time to call a group of leading figures in economic research back to the trenches of general education.

#### Refined and Polished

Starting a project is an easy thing; completing it well requires painstaking effort. Each time a chapter was written, the mimeograph machine would creak and groan as it printed it out for our MIT students to try. This approach placed high demands on both me and the material, but I took pleasure in it. I naively predicted I could finish the work in a year; who knew it would take three. I devoted all my weekends and summer vacations to it, working hard to simplify complex basic principles into clear, easy-to-understand teaching material, and I had to give up my hobby of tennis. In the course of writing I found that, to make "economics, the dismal science" lively and exciting, even the most traditional economics charts had to be redesigned.

#### The Crucial Moment

In the autumn of 1948, the first edition of Economics finally rolled off the presses. However hard the earlier work had been, or however optimistic the dream, no one could be sure what the future result would be. Fortunately, this new way of handling economics seemed to resonate with readers from the very start. Many universities, large and small, adopted the new book, and each time a reprinting sold out, Economics would set the presses running again.

A Guggenheim Fellowship brought me to Europe, and in several major cities I visited every large bookshop to see whether I could buy copies translated into French, German, Italian, Spanish, and Swedish. Apart from wanting to satisfy an author's innate vanity, as an educator I was also pleased to see German citizens being put to the test of the latest mainstream economics.

Reviews of the book rapidly heated it up. The first to strike was John Kenneth Galbraith, then the editors of Fortune, a conservative business magazine. He predicted that the next generation would all study Samuelson's Economics. Although the praise was intoxicating, I have to admit that the book's enduring popularity did genuinely surprise me. In fact Galbraith was more prescient than I was; Economics did establish a new, lasting paradigm.

#### A Bumpy Road

Writing the book did not bring me only pleasure. In the days when Senator Joseph McCarthy was in full opposition, when accusations of radicalism erupted violently from the podium and the classroom, my book was also attacked. A conservative male student at MIT warned the university president, Karl Compton, that if Paul Samuelson were allowed to publish an apology for a "mixed economy," his academic reputation would be harmed. Dr. Compton replied that the day his faculty were subject to censorship was the day he would resign the presidency. Forty years later this seems almost comical, but in an era when a textbook could be branded subversive, being a professor at a public university was no joke. (There was an excellent textbook, published a year before mine, that was strangled in the cradle from the start by being falsely and maliciously accused of Marxism.) In fact, when you are slapped by the right, the pain can perhaps be partly offset by a slap from the left. In the 1960s, student activism was in full swing on campuses at home and abroad, and the two-volume Anti-Samuelson was duly published, accusing the market of being a dog-eat-dog, mutually destructive arena, and me of being a staunch champion of the market, an apologist for the laissez-faire world, and a loyal running dog of capitalism.

#### Forever Young

Just as a child takes on a distinct personality different from its parents', so it was with Economics. At first the initiative was in my hands; but later, by its own strength, it turned the tables and led me by the nose. As the years passed, my hair turned from blond to brown and then to gray. But like the portrait of Dorian Gray, which never grows old, Economics is forever 21. Its cover changed from green to blue, from brown to black, and now to gold. With the help of hundreds of letters and suggestions from teachers and students, Economics evolved and matured within its covers. Historians of mainstream economic doctrine, like paleontologists studying the bones and fossils in different strata, can identify the eras when intellectual tides rose and fell by analyzing how the first edition was revised into the second, and so on until the sixteenth.

So it was. The work was full of hardship, but well worth it. The call of tennis was always there for me, and the day finally came when I could no longer resist it. I said to McGraw-Hill: "I have paid what I had to pay. While I enjoy the good life of a professor emeritus, let someone else continue tilling the field of research that most interests me." McGraw-Hill answered offhandedly: "Let us find you a coauthor. We will draw up a list of economists whose ability and views you admire and whose interests match yours." And so the search for the perfect William Nordhaus began.

Yale is only 150 miles from MIT, and it was there that the real Nordhaus would be found. The one helpful thing was that Bill had already earned his PhD at MIT. Since then he had served on the President's Council of Economic Advisers and been posted successively to Cambridge (England), Delhi, and Vienna, gradually rising to prominence. It was formally proven that, like Gilbert and Sullivan or Rogers and Hart, we were a very well-matched pair.

As in a classic fairy tale, we lived happily ever after. What matters is that the book is still young, pointing in the direction the mainstream of economics is about to flow.

#### Science? Art?

Why has economics become the most popular elective at most universities? The reasons are many. Economics concerns the real life around us. When inflation comes, none of us can escape it. When recession hits, the tide reaches every boat. And the bird that rises early to find a worm needs to understand supply and demand. So do the elderly living only on a limited pension and the young idealists determined to improve and reform society.

The economics of good sense is not entirely obvious. The common sense you bring with you from home to college will not help you understand why both rich and poor countries can gain enormous benefits from free international trade. (Without a good course on comparative advantage, your senator wouldn't understand this either.) On the other hand, once you have mastered the teachings of so-called microeconomics and macroeconomics, there is no mystery left. If it does not foster good sense, it is not good economics.

A first course in economics will not make you a master adept at understanding and interpreting intricate and abstruse topics. But based on the experience of students everywhere, I can tell you one thing: your best economics course will be your introductory course. Once you step into this strange and wonderful garden of ideas, the world will never again be the same. And when you look back on the experience years later, even the things you didn't quite understand at the time will have clearly matured.

Enjoy it to the fullest!

**Paul A. Samuelson**

**MIT, Cambridge, Massachusetts**

**1998**

### Lu Ming: Experience and Advice for Scholars Doing Empirical Research

[Source: Econometrics Circle]()

"Carrying Empirical Research Through to the End" is an article I wrote back in 2004, published in The Economist's Teahouse. It drew a certain response, beyond my expectations, which also shows that empirical research is receiving more and more attention in China. At such a time, we do need some articles like this to help everyone clarify some understandings about econometrics.

Today I especially want to help everyone clear up several misunderstandings — misunderstandings that happen to be ones I heard from students myself in the course of working with them, so they are not something I invented.

The first misunderstanding is that empirical research is applied, and applied work is not academic enough and has little value, as if it only answers a question in reality and the conclusions found seem to be things we already knew.

The second is a phrase everyone often says: empirical research is easy to do, easier than theoretical research, and when you can't continue with theory you switch to empirical research. Speaking of which, this winter vacation I was just reunited with one of my former students, now a PhD student at KU Leuven in Belgium. Her recent work is related to social economics, and she does theory. When we met over the winter break and talked, I asked her, why don't you do empirical work on this topic? She said empirical work is too hard. So I think that on some questions, it is not as everyone imagines — that when you can't do theory you do empirics; it may well be the reverse: when you can't do empirics you do theory. I'll talk about the relationship between empirics and theory later, and I'll tell everyone that both theory and empirics are very important and complementary.

Third, students often think empirical research is very simple: just put the data into the computer, the results come out, and you can write a paper. Once, a colleague said, with some self-deprecation: he felt very uneasy, because the data wasn't collected by himself — much of the large-sample data we use was collected by others — and the idea wasn't very new, not his own; then the programs were ready-made, Stata already coded. After that, we put data we didn't collect and an idea that wasn't too new into Stata, the results came out, and we felt very uneasy. And our students think that a master of economic research should be full of mathematical symbols plus programs they wrote themselves, and then draw very fancy figures — that's what counts as high level.

The fourth misunderstanding is that many students think that when we now study China's problems, we should do theory, because theory is in line with international standards. Especially for beginners: when taking the master's and doctoral entrance exams in economics, you are tested on micro and macro, and in your first year you take a great deal of micro and macro theory, so you feel that is the mainstream of economics and only that is prestigious; even when doing research on China, you should write China's problems as mathematics, or simply not do research related to China at all. Is that really so? Should Chinese economics research really do theory first?

The questions I'm going to talk about next are an attempt to clear away the misunderstandings in everyone's minds. Today I will cover these questions: first, why do we do empirical research? Second, what is the relationship between empirics and theory, including ideas? Third, what is the relationship between empirical research and the development of Chinese economics? Fourth, how do you complete a piece of empirical research?

#### 1. Why Do Empirical Research?

I think the functions of empirical research in the process of modern economic research and development are as follows:

First, empirical research is used to test theory. Because for economists, we have too many theories, but in reality which theory is correct and which is more or less important — in fact, without practice we have no way of knowing. This is the first important aspect of empirical research. And from the path of making economics scientific, as everyone knows, the first Nobel Prize in economics was awarded to econometricians, and in recent years it has been awarded to econometricians in a row. As everyone knows, for science, an important criterion is that it can be falsified, so how do we know whether a theory is right or wrong? Economists in fact rely heavily on empirical research to make economics a science that can be falsified or confirmed.

Second, to challenge the theory. Once a theory is produced — especially a mathematically modeled theory, as everyone knows — once its mathematical logic is established, it has self-consistency and logical scientificity, because it relies on mathematics, and mathematical logic is rigorous, so in a sense it is correct. But sometimes when we look at a theory, we find that when its conclusion is that x and y are positively correlated, we feel that in real life this doesn't seem to be the case. The possible problem is that the existing theory has not correctly captured this relationship in reality, or has not captured the more important aspect of the x–y relationship. So how do we know? Economics relies on empirical methods to see whether x and y are positively or negatively correlated. If it really is so, then the existing theory is very likely correct to a large extent. At this level, the relationship among econometrics, empirical research, and theoretical research is like that between experimental physics and theoretical physics.

Third is to discover some net effects. We know that many theories, especially modern economic theories, are already extremely complex, and there are different theories about the relationships among variables. Some theories hold that x and y are positively correlated; others hold that they are negatively correlated. As everyone knows, when making policy you need to know whether x and y are positively or negatively correlated. Each mechanism, viewed separately in theory, may be correct, but for the policymaker, if different mechanisms predict opposite directions for the x–y relationship, then you need to know: once the policy is implemented, is the effect positive or negative? Theory cannot help us answer this question. Different theories may well reach opposite directions, and current economic research can come up with different mechanisms on many questions, so that under this mechanism the two variables are positively correlated and under that one negatively correlated,

And many people who do theory often like to be unconventional: everyone thinks x and y are positively correlated, and they write a paper arguing the two are negatively correlated — this is a shortcut to publishing abroad. In modern economic theory, a great many theoretical economists are already doing this kind of work to reveal the different mechanisms of the x–y relationship. So when we look at the real world, we may be confused by these theories: which one is actually more important? Without econometrics, we have no way of knowing whether the net effect is positive or negative.

Fourth, going one step further: the relationship between x and y, and between x and z, may be clear in theory, but empirical research can tell you which effect is stronger. As everyone knows, when making policy — for example, with a budget of one million — we need to consider which thing to spend the money on. For instance, there is a famous debate abroad: for the development of education, at least two approaches can be proposed. The first is training teachers; the second is reducing class size, because people think that if a class is large, each student gets less attention from the teacher, and then students' performance declines. So for the government, it needs to consider: which thing should this one million go to? Setting aside the correlation between teachers and students, and the correlation between class size and students — whether they are positive or negative, whether significant, is itself uncertain. Even if it were certain, the government still needs to know whether to spend the money on reducing class size or on improving teachers: which is more worthwhile?

So we must know: under large-sample observation, how much does doing this improve students' performance? For econometricians, a very important goal is to make the coefficients of the econometric results accurately match the size of the true effects, so that the effects of different policy variables in an econometric analysis can be compared. This helps raise the effectiveness of policies and the efficiency of resource allocation, and helps improve human welfare. This is work that theory cannot tell us about; theory cannot tell us how large this effect actually is.

Fifth, in frontier research in modern economics, what is more important is to identify the mechanism, especially in much macro research. I myself also do some macro research, for example on inequality and growth. There are so many theories saying the two are negatively correlated, and also some theories saying they are positively correlated. Where has our work reached? In fact we look at the net effect: under different theoretical mechanisms, I can tell you that on the whole the two are negatively correlated. But there is another hard task: what causes this negative correlation? In our research we actually don't know, but one of the hard jobs of econometrics is to tell everyone which mechanism is at work. In theory there may be several mechanisms telling you the two are negatively correlated, but through which mechanism does it operate? The importance of identifying this mechanism lies in knowing, when making policy, which mechanism to make policy on. And this is very difficult, because in the process of collecting data it is hard to distinguish such mechanisms. Especially when you are not yet very experienced in econometrics, you often just want to look at the relationship between x and y. When doing econometrics and collecting data, you only consider how to measure x and how to measure y, then put the data into the machine and that's it. But we may rarely think about what the intermediate variables between x and y are. We may overlook these things, and in the end we have no way of seeing what these intermediate mechanisms are, which keeps the research from reaching a higher level. These tasks require, before doing empirical research, a comprehensive and clear understanding of the various mechanisms from x to y, and when designing the questionnaire you already need to know how to use certain methods to separate the various mechanisms. You must make it very clear and let everyone be convinced that the mechanism you see is very "clean." The word "clean" is a colloquial expression: it means that in the x-to-y relationship I see, I can convince you that it is because of the mechanism I described, and does not include other things. To achieve this there are many methods.

The second question I want to talk about is: at what point in the history of economics are we? The first point I want to make is that the whole theoretical framework of economics is already basically complete, which means it is already very hard to make major breakthroughs in the progress of economic theory. This is an important reason empirical research has become the worldwide trend. Here I want to mention Chen Zhijun, who works on industrial organization theory and came from a physics background. I think what I just said may not be very credible coming from me, because much of my current work is empirical. He came to Fudan last semester, and after a seminar we ate together and he said that economics has basically developed enough, so people should just use it now. He said this: "Sometimes I feel that being too good at mathematics is harmful." This sentence is especially worth pondering for students who are good at math.

Second, in some research fields we have too many theories but insufficient empirical evidence, and in these fields the frontier is now often pushed forward by empirical research. What are some of these fields? I will return to several of them later. The first is, for example, IO — empirical IO has now become the frontier of IO research. Another very representative field is personnel economics, where empirical research is also very hot right now.

The third aspect is applied micro and policy evaluation. Right now, across the whole world, the public sectors of governments are very large, and if you look at the share of government public spending in GDP, it is rising in developed countries. So where has all this public spending gone? To labor market, health, and education. In these fields we run into the question I just raised: does a policy actually work, and which policy works better? The government also wants to know this. For example, the small-class example I just mentioned is a very good one, which I also like to use: in the United States, enormous sums are invested each year in reducing class size. Because, as everyone knows, reducing class size requires more teachers, classrooms, and hardware, so the federal and state governments put a great deal of money into it. But does it really improve students' performance the way we imagine? That is not certain, because in a society with choice, if one class is small and another is large, you immediately think: who would choose the small class? Usually the rich, because they have money; and families where the parents are highly educated also think a small class is better, so they send their children to the small class, and so we see that the small class performs well. But does the small class perform well because the parents are rich? Because the parents' own academic performance is better? Or because the class size was reduced? So the good performance of small-class students observed with our naked eye proves nothing. If this policy has no effect, you should know that the money saved for the government could have built who knows how many aircraft carriers. So in such research, the improvement of human welfare and the creation of social value brought by policy evaluation far exceed a purely theoretical study.

So why are applied research and policy evaluation so important? Many people feel that the humanities seem useless, that the humanities are just people tapping their heads and thinking up ideas. But current empirical research can already play a very important role in improving social welfare and raising the efficiency of resources. Looking back at China, there is a sentence I always say: there are policies but no evaluation. We have made a great many policies — have we evaluated them? Do we know whether these policies work? We don't.

Over roughly the last ten years in economics, a new wave has emerged in three fields, which I have also been talking about: comparative economics, political economics — actually the new political economics — and social economics. Here I want to especially emphasize comparative economics and the new political economics. A new research direction of the new political economics is to look at how the political structure and political institutions of various countries affect each country's economic growth and performance, so the rise of the new political economics itself carries a comparative perspective. This immediately raises a question: do different countries' social and political structures have different effects on economic performance and growth? This question is first of all an empirical question, because before theory develops, the first thing we must confirm is the fact I just described. If it cannot be confirmed, then don't do theory — there would be no comparative economics or new political economics to speak of.

When you now study economic theory, when you study micro and macro theory, you will see the names of many Nobel laureates in your textbooks, and so you have a misunderstanding: you feel these represent the frontier direction of economics and the research direction of mainstream economics. What I want to remind everyone is that what you learn in micro and macro — especially in the public-course micro and macro, not the specialized micro and macro seminars or workshops — is all at least twenty years old, because the Nobel Prize is usually awarded for achievements from twenty or thirty years earlier. You should pay more attention to the Clark Medal, and look at what recent Clark Medal winners have been doing, and what the people at the very frontier of world economics are doing. Let me name a few for you. I cannot say these people do empirical research, but I must say that most of them do both theory and empirics, and their recent work is largely concentrated on empirics. What kind of people? Murphy, now at Chicago, a student of Becker. Levitt, a Clark Medal winner from a few years back — he's interesting; there's a joke from when he was a PhD student: in a microeconomics class he asked his classmate what a total derivative is, and his classmate looked at him and said, "You're done for." Yet a person like that ended up winning the Clark Medal. His work is entirely empirical, and later I'll give examples of his work, like how he looks at the relationship between abortion and crime, and whether political business cycles can be confirmed empirically. The next two people are now very famous, and I estimate they may be future Nobel laureates: Shleifer and Acemoglu. One is at Harvard and the other at MIT, and both came from theory. Shleifer used to do corporate finance; Acemoglu mainly does labor economics, especially human capital. Recently these people have done a great deal of empirical work, and in the last couple of years their most famous empirical study — also a study that has sparked a debate between the two — is institution and growth, the relationship between institutions and growth. There is another person who has not yet won the Clark Medal, but I think he fully deserves it; he is also a student of Becker, Edward Glaeser. He is also at Harvard, and he also does theory, but recently a great deal of his work has been empirical too, for example in social economics, and he has also worked on institution and growth. This person does everything, and in urban economics he is also a leading authority. Look at the research of these people and you will know where the international frontier is.



#### 2. What Is the Relationship Between Empirics and Theory?

Standing here today teaching empirics, I will promote how important empirics is, but that does not mean theory is unimportant. Once in class, a student heard me say this and asked: Professor Lu, you say empirics is important, so is theory more important or empirics more important? I said: people still ask this kind of question? If someone tells you theory is more important than empirics, you can just treat him as a lunatic. Theory and empirics are both important. When I say today that empirics is important, it does not mean theory is unimportant; on the contrary, theory is very important. So what is the relationship between theory and empirics? Let me first tell you two things, neither of which is my own; someone else said them, and I'll relay them.

The first is my classmate, now at Queen's University in Canada, Ding Weili. She once said something: why did I make empirical research my career choice? Because once at an international conference, two people doing theory were arguing endlessly, and then an empirical economist said, "Stop arguing; let me show you the evidence," and the theoretical economists stopped arguing. She said that from then on she was firm in her confidence to do empirical economics.

The other is a student of mine I just mentioned, now at KU Leuven, called Ji Yuemei. This winter break I also talked to her about the confusion many students face — that students don't value empirics, think empirics is simple and easy, and only do empirics when they can't do theory. She now does theory, but when she heard me say this her eyes were full of bewilderment; she thought, how could Fudan students be like this? Then she said this: "I think empirical economists should read more theory papers, and theoretical economists should read more empirical papers." Why? Think about it: what is economics? Borrowing a phrase from teacher Wang Yongqin, economic theory lies not in creation but in discovery. Do economists have the ability to say, let's create a theory? Economists have no such ability; their ability lies in discovering what regularities exist in real life. And what is a regularity? A regularity is nothing but explaining why x leads to y. Before that, you first have to know that x leads to y, and then you explain why x leads to y. So empirical economics, for theory, can help you distill the facts that exist in this world. Just this Tuesday, Li Ting, who is doing a PhD in Toulouse, came to our school for a seminar. She does theory, and she said: "Coming back this time, I feel there is so much that can be done in China. Every day I read newspapers and the news, and I tell my mom, so many things could become a paper." What does she mean? In fact, the inspiration of theoretical economists comes from observing reality. A newspaper is one kind of observation, television another. What empirical economists can do is provide a more scientific observation that everyone can be convinced by — that's all. The conclusions econometricians reach based on large-sample data and scientific methods are more credible — that's all. So they can provide the factual foundation for theoretical economists.

I just explained why empirics is important for theory; next I want to say that empirics must also be based on theory. Here I have to rebut a view of yours: when you read a lot of econometric papers, what you see is econometricians lining up a dozen or so variables on the right-hand side of an equation and the results come out — Stata reports them automatically anyway — and when writing the paper they say x and y are positively correlated, x and y are negatively correlated, and that's it. It is not like that.

Empirical research must be based on theory; the second aspect is to avoid data mining; third, — perhaps this goes a bit too far, — for theoretical economics; the next question to discuss is the relationship between mathematical models and econometrics; first, the theory already exists and can be tested directly, at this point your; second, we already have enough competing theories, so we go and test; third, if the mechanism is already very clear, there is no need to write everything; the second example, as everyone knows in the real world there are so-called residential areas — empirical research must be based on theory. Of course, by "based on theory" I do not necessarily mean based on those already-published mathematical-model theories; that is not what I mean. When doing empirical research, from the very start — from the selection of variables, to the measurement of variables, to the setting up of the model — you must base everything on theory, or more colloquially, you must have a theoretical foundation and be able to say why. Otherwise it is very easy to raise the R² of an econometric equation: if a linear term doesn't work, add a quadratic term; if the quadratic doesn't work, add a cubic; if the cubic doesn't work, add a quartic... but what's the point? Doing econometrics that way is meaningless. When you add each higher-order term, we have no idea why the relationship between variables is like that, because if current economic theory tells us there is a nonlinear relationship between variables, it usually stops at the quadratic term; very few people tell you there is also a cubic relationship.

The second aspect is to avoid data mining. When you do empirical work in the future, you will find that you often encounter this situation: a result comes out that doesn't match your prior expectation — sometimes it is insignificant, sometimes you expected positive but the result comes out negative. There is a practice called data mining, which is trying: trying over and over, constantly adding and removing variables, constantly adding quadratic and cubic terms or removing them, adding a log term, adding an interaction term, and then changing the measurement of the data from FDI to trade... In this sense, everyone who does econometrics knows that econometrics can help you get whatever result you want — that is data mining. If econometrics is done this way, that is too frightening; then we might as well not do econometrics, since the conclusion is already known in advance. So how do we avoid data mining? Before concretely running the regression, you should already have a theoretical judgment, and then run the regression. When you find that the result differs from your theoretical judgment, you have to be careful: first think about why, and where the problem is. Then find the problem, and adjust your model and data based on the reason you found for why the problem arose — that is not data mining. In actual practice, the boundary between data mining and the theory-guided adjustment of data and models I described is very blurry, but only results obtained strictly following the guidance of theory can stand up to testing. Under the current academic system, if you only patch together results and make some very obvious mistakes, it is impossible to get published.

Third, — perhaps this goes a bit too far, — for theoretical economists, you may be one-sided yet profound, but for empirical economists, you must be comprehensive and carefully observe reality. As everyone knows, in doing theory, there are many possible relationships or correlations between x and y. For people doing theory, their work is really just to describe one possible mechanism. If you want your model to be more complex and richer, you can describe several mechanisms, but you don't need to spell out every mechanism between x and y in one model. So in a sense, theoretical economists can be one-sidedly profound, but econometricians cannot do this, because when you omit an important mechanism or an important variable, the first thing that appears is bias in the coefficient estimates. So for econometricians, when you think of a question to work on and then go to choose data and build a model, you must be very familiar with the literature in advance: what has theoretical economics roughly done, and what theories are there? How did predecessors set up models and choose data when doing similar work? Only then do your own work. At this point, omitting any important variable or literature is very dangerous.

The next question to discuss is the relationship between mathematical models and econometrics. There is now a view that before doing econometrics you should first write a mathematical model, so that it looks more high-end — it is not like that. If your econometric model can be built on a mathematical model, I can only say that is better, at least better than or equal to having no mathematical model. But sometimes it is not a necessary thing, and what I especially oppose is everyone, in pursuit of having a mathematical model before the econometric model, forcing a mathematical model in. I have come across many papers where the earlier mathematical model and the later econometric model don't match up at all. So when is the mathematical work not necessary before the econometrics? Let me list a few cases:

First, the theory already exists and can be tested directly; at this point your innovation is to provide evidence. For example, in theory, in the American economics-of-education community, everyone believes that competition among schools helps raise school efficiency and thus has a positive effect on student performance. In theory everyone thinks so, but is it actually so? You don't need a mathematical model; you just need to test it directly. And the work now being done in this area all has no mathematical model, which has also become a frontier question in economics-of-education research.

Second, we already have enough competing theories, and we test which theory is more important. For example, as I mentioned, the relationship between inequality and growth, which we have worked on ourselves. Some say it is positive, some say negative, so we put them together and see whether it is positive or negative. Some say it is positive in the short run and negative in the long run, so we also put them together and see whether the short run and long run have such a difference.

Third, if the mechanism is already very clear, there is no need to write everything as a mathematical model. Mathematical models help us see mechanisms that the human brain may not see clearly. If the human brain's line of thought is already clear enough, there is no need for a mathematical model. For example, in social economics or the economics of education there are some very hot topics. The first is peer effects, which means your performance is influenced by your classmates. For instance, among students you'll see that those who like academia are often in the same dormitory; everyone influences each other and all come to like academia, and those who like playing computer games also live together. That is peer effects. So do we need to write a mathematical model for why people who like games influence the classmates around them? For people doing empirical economics, everyone feels this thing may exist; I am only identifying whether it is because people are influenced by peer effects only after living together, or whether people of the same type choose to live together in advance. For empirical research, the latter question is more important.

The second example: as everyone knows, in the real world there is the so-called segregation of residential areas, where the poor live with the poor and the rich live with the rich. Then someone says that when social interaction exists, because the poor are together they all exert negative influences on each other, and the rich together also exert negative influences on each other. So the accumulation of the poor's human capital slows down, and their probability of unemployment rises. But here we run into a big empirical problem: is the phenomenon you observe really caused by social interaction, or did poor people who don't value human capital choose to live together in advance? We don't know this either. Does this need a mathematical model? For empirical economists it can also be left out, because what is more important is to identify which situation caused it. That people who live together influence each other does not need a mathematical model written out by empirical economists.

The third example I want to give is the social multiplier. As everyone knows, economics loves multipliers — for example, the Keynesian multiplier, where a small amount of monetary or fiscal policy can be amplified through the Keynesian multiplier. Now the development of social economics has proposed a new concept called the social multiplier. That is, if mutual influence between people exists, then the effect of a little policy can be continuously amplified through mutual influence between people. For example, Shanghai has a project called "A Million People Learn Etiquette," where the government invests money to let everyone improve their manners. Think about it: after you learn these manners and return to your community or colleagues, you may influence them; everyone will think, why does this person dress better or have better table manners, and they will learn from you. So this effect exists without the implementation of government policy. When you evaluate the overall effect of a policy, what you see is the final effect; dividing this final effect by the effect on the person the policy directly acts on gives the size of the social multiplier. Empirically, does this social multiplier exist or not? How large is it? This has also become a very frontier topic in empirical research, and it doesn't need mathematics either, because we all know this thing either exists or doesn't, and it arises through interaction between people.

In some research topics, economists have actually already developed a division of labor between theoretical and empirical economists, just as physicists have divided into theoretical and experimental physicists. At this point, if your comparative advantage is in empirics, you don't need to do the work that theoretical economists should do. Because in some research fields, the entry cost between the tools used by theoretical economists and the tools used by empirical economists is very high. Let me give a very simple example: social networks. The econometrics of social networks has a set of methods, but the mathematical tool used for mathematical modeling of the formation of social networks is graph theory. I don't know what graph theory is; I only know the word. In my own work, I also study the formation and influence of social networks. But for me to do the theoretical work of how social networks form would be suicide. So in fields like this, when you enter you should be clear about what kind of economist you want to be. This does not mean that those who can use graph theory are more clever or smarter than those who cannot. Because for those doing theory, they also don't know whether these things really exist in reality. At this point, if you position yourself as an empirical economist, just go ahead boldly.

#### 3. What Is the Relationship Between Empirical Research and the Development of Chinese Economics?

First, I have already given many examples to tell everyone that empirical research is the frontier in many fields. So in which fields does empirical research constitute the frontier? First, in labor economics, empirical research has become more and more important since the 1990s, because labor economics is a typical discipline whose theory has basically become complete. So after the 1990s, if you say you are a labor economist, everyone basically understands you as an applied microeconometrician rather than a theoretical economist. This discipline has developed to a stage dominated by empirics.

Second, the two examples I mentioned: in industrial organization and personnel economics, these disciplines developed from theory, but during the great development of their theory, empirical research was very scarce because of problems with data availability and data cost. In recent years, empirical research in these fields has constituted the frontier.

Third, in many research questions, we simply don't know what the facts are. We can all feel that the development of China's economy, China's institutions, culture, human behavior, and social and political structures are somewhat different from some things we learn in current economics. We vaguely know they are different, but from the standpoint of economic research, we actually lack a difference that can be confirmed by economic science: does the difference exist or not? If it exists, how large is it? We don't know. So currently, much of our understanding of China's problems is still based on theories developed on the foundation of Western facts, used to study and understand China's facts. But doing this, in many cases we cannot see China's problems through clearly. If you want to propose an understanding of China and construct a new theory, the first thing to do is — indeed, we need a new theory, because I can tell you that empirically there is a difference.

I especially want to emphasize several reasons why empirical research is particularly important in China. The first is what I just mentioned: China is a bit different, but in what way? The facts are still not very clear. The second is the need to make policy. At China's current stage of economic development, we don't yet have enough resources to support a group of economists who can ignore real-world problems and just read and write papers in their studies, doing so-called pure theory. The demand for applied research at China's current stage must be very great, and for this, what is very important is the policy evaluation I just mentioned, based on empirical research. As everyone knows, the Chinese government now talks about scientific decision-making. What is scientific decision-making? In English there is no phrase "scientific policy making," but there is a phrase "research-based policy making." I think this is the English translation of scientific decision-making. The scientific outlook on development we now speak of also means doing more research. In addition, China's empirical research is also a focus and trend of international attention. When you get a chance to attend conferences in the future, especially international conferences related to China, you will find that the share of empirical research is the overwhelming majority.

This reflects economists' awareness that empirical research is the first work we need to do on these questions. For example, at the end of last semester, an article by my student Zhang Shuang and me was accepted by a very high-profile labor economics conference at Xiamen University, organized by Professor Hong Yongmiao, who invited many big names, including Heckman, a Nobel-level figure. Someone at the conference said: we came to China hoping to know what is happening in China, but we didn't hear it. So, doing research in China, the first thing is still to honestly tell everyone what is actually happening in China. On this foundation, if we find differences, then we need new theory.

Finally, because of the preceding three points, this relates to your funding. When you go after projects, including some international projects, they all involve the funding question.

#### 4. How Do You Complete a Piece of Empirical Research?

Empirical research, like theoretical research, is something where our students often tend to focus on the middle part. It's like eating a fish: you see that the fish's belly is fat, but you don't know that there is a lot of work to be done before and after doing the research, and this before-and-after work often takes up 70% of the time. But you only see the middle part, the paper itself. So for empirical research, what needs to be done beforehand?

First, before you do empirical research, you need a great issue as your research background. This is the question of "keeping the big picture in mind," but you must also have a small point, which is what I often call "starting from a small point." Having a great issue makes your research important, and having a small point makes your research operable. After that you need a good question. For example, as I mentioned: do institutions matter for economic growth? Institutional economists will tell you they matter a lot, but for empirical research, we don't know. Because when you see that countries with good institutions also have high economic development, we actually don't know whether it is because their development is high that they have good institutions. So for empirical economists, the task is to look at whether this causality from institution to growth exists — that is a great issue. So what is the small point? In Acemoglu's research, he thought: how do we pin down this causality? He found a very important instrumental variable. He looked at the natural conditions of Africa, Latin America, and other colonies during the colonial period, one very important condition being the mortality rate. His idea was: if the mortality rate in these colonies was high at that time, white people would be unwilling to settle, and so they would adopt extractive institutions. If they were willing to settle, they would transplant Europe's good institutions. The state of institutions at that time would affect today's differences in the quality of institutions in various places, and this would very likely bring about differences in growth. Through the action of this chain, we can confirm that institutions are the cause of growth. This approach is a small point that made his research operable.

The next example I'll give is tea and sex. As everyone knows, in developing countries there is a very important problem of sex imbalance — too few girls and too many boys. One explanation economists give is that for a family, having a boy is better than having a girl, because boys have higher productivity and higher wages, so the return on investing in a boy is higher. Many people dislike this explanation, saying that economists explaining fertility behavior from an economic angle is wrong; they think it is a cultural influence, that everyone dislikes girls and discriminates against them. So are the economists right? In the tea-and-sex research I mentioned, there is a Chinese-American named Nancy Qian, a star on the American job market the year before last, now at Brown, who has visited Fudan two or three times. In her research she uses Chinese data to test whether family fertility behavior is an economic decision. She realized that before reform and opening up, the prices of various commodities were controlled, and after reform and opening up prices were gradually liberalized, so the price of cash crops, especially tea, rose. As everyone knows, picking tea is a comparative advantage of girls, because girls are more careful. So, according to the inference of economic theory, a rise in tea prices would raise the return to girls, which would make families more willing to have girls, and after the child is born they would give girls more care to reduce their mortality. So she used a natural experiment — the price reform raised the return to girls, thereby raising the birth rate and survival rate of girls. Her research found exactly that.

Everyone may immediately think: these are all too complicated, I can't do them. If you can't, you can do something simpler. For example, let me talk about a mine-accident example — though of course this isn't simple either, because studying mine accidents would be very troublesome. But I want to give an example. At the economics annual meeting the year before last, a teacher said he observed that after the central government required local governments to raise compensation for workers who die in mine accidents, mine accidents actually increased. Why? According to economic theory, this could cause more severe moral hazard. If safety measures are determined by economic decisions, and life has value, then the rise in the mine-accident level means that if you reduce safety measures, once an accident occurs the compensation level is higher, which increases people's moral hazard and reduces safety precautions, which could lead to a rise in mortality. The problem is: if you have studied information economics, it is very easy to construct a theory of the story I just told. But there is a problem here: is the story he just told actually true? Did the mine-accident rate actually rise after compensation was raised? Or is there a causal relationship between these two things? We don't know. If you can make this paper, it would certainly be an excellent one, because it is a big question that can prove whether human life can be priced.

After you have proposed a good question, you must do a thorough literature survey. This can serve several functions. First, through the literature survey you can demonstrate why your research is important. For example, in the mine-accident problem I just mentioned: if your literature survey is done this way — in such-and-such a year and month someone said the two are positively correlated, so my paper will prove this — then nobody will be interested in this paper. If your literature survey is like this: in economic theory someone proposed that people's investment in safety facilities is an economic calculation, and many social scientists dispute this; for example, some sociologists think these views of economists are wrong, and this paper uses the mine-accident case to tell you whether economic theory is correct — then people will find it very important. The second thing you must do is explain what's new. The easiest thing to overlook in research is what's new. You make something — why should your readers read it? In the process you must try your hardest to dig out something new in your work to attract your readers, because reading a paper has a cost for everyone. Many of our students think: I did great work, but I won't say so, I'm cool. I tell you, if you hold this attitude in research now, you're dead. When you do research, you must do your utmost to compare your work with existing research on idea, method, and data, to tell everyone: I used a different idea, I used a different method, and my method can produce more accurate or unbiased estimates. My data is new; others' samples are small, mine is large; others don't have this measurement, I have it, so I'm new.

Now let's talk about the middle of the research. At this point you first need a theory. As I already said, the first question this theory and analytical framework involves is: do I need mathematics? I already answered this, so I won't repeat it.

Second is the data. Where does your data come from? Of course we have a lot of ready-made data — for example, those of us who do empirical research all have some data on hand. But when you do empirical research in the future, whether in a company or in academia, you often need to collect data yourself, and at that point your data is first-hand. Especially at the initial stage, your sample cannot be too large; a large sample needs money, and you don't have that much money, so your sample isn't very large. So to do this research, how do you make it a good piece of research? You have to think about it. First, you need to be new; you need to pay attention to whether, in your not-so-large sample, there are some variables that other large-sample data don't have; third, you need to have a very smart idea — of course this is hard, so...; finally there is econometric analysis (econometric analy...; for example, in a study we did ourselves we ran into such a situation; after the research we get many results, and at this point you still have some work to...; if you did as I described just now, then I will congratulate you, because...; five, some comments; next I will make some concluding comments; besides theory helping us think about these mechanistic things, ca...; the second situation is... — or you study a problem that no one has ever studied before, and you are the first to study it by collecting data.

Second, it must be well defined, that is, your data must be very clearly defined. For example, according to the existing literature, social capital is defined as these several aspects, all measured in this way, using what kind of questions to ask, so in my own questions I ask the same way, and in this way you can get a good measure. Finally, your data must be well structured, that is, you should have the variables that ought to be there, and you should also have the intermediate variables of the mechanisms you want to identify, so that even if your sample is a bit small, you can still do very good research.

Third, you need a very smart idea. Of course this is hard, so the requirements get higher and higher. In econometrics, usually the relationship you see between variables can only be called a correlation, but current empirical research is increasingly unsatisfied with correlation and wants to look at causality. A commonly used method for looking at causality is finding an instrumental variable. So what method do you use to look at causality? Use a natural experiment? Use exogenous variables that are convincing enough? Or find an instrumental variable? This requires you to have very smart ideas to convince everyone that what you see is really causality, not a correlation whose direction is unclear. Of course this requirement is relatively high, and in many research fields it is a frontier topic. For us, sometimes don't casually say you found causality — finding a correlation is already not easy.

Finally there is econometric analysis, which is to analyze, regress, explain the results, and so on. After we get the results we may run into this problem: first we find that the result seems wrong, that the reality we sense seems not like this, specifically shown as the sign or significance not being what we expected beforehand. At this point you must be careful; the problem may be in the following aspects. You must think about which aspects went wrong, and then adjust them, rather than doing data mining. First, the data may be bad. The first reason the data may be bad is that you used Chinese data. Many of China's macroeconomic statistics are very rough; for example, China's unemployment rate itself is not the real unemployment rate, and using it causes problems. Also, your sample may not be a random sample, and your sample size may be too small. These problems can all lead to low data quality, so that what you want to see cannot be seen. The second possibility is that the theory went wrong. This has two possibilities: one is that the theory you based on before is wrong, or it doesn't reflect reality at all. The other is that there may be a new theory or mechanism that offsets the mechanism you originally thought of.

For example, in a study we did ourselves we ran into such a situation. When we were doing research on the correlation between economic openness and domestic market segmentation, we originally thought that openness intensifies competition, and intensified competition should integrate the domestic market. But later we found that when we put it in, it was a linear relationship, and positive — that is, openness promotes market segmentation. So we thought: why? It may be that in the early stage of economic development, openness promotes market segmentation, and when economic development reaches a certain stage, openness promotes market integration. So there may be a quadratic relationship. When we put the quadratic term in, sure enough a very beautiful U-shaped curve came out. When you do econometrics under correct theoretical guidance, you will know how to add things — whether to change the form of the model, change the data, or change your measurement. Afterwards, the results you get under theoretical guidance can help you discover new things, because the U-shaped curve I just mentioned tells us that in theory the relationship between economic openness and market segmentation depends on the degree of openness — something not present in existing theory. In fact, if any of you here are interested, you can absolutely go do such a theory to describe the relationship between economic openness and market segmentation as a U-shape like the one we saw. At this point you'll know what the relationship between theory and empirics is. Where does theory come from? It is not conjured up in the head; it is to formalize the facts and regularities already discovered.

After the research we get many results, and at this point you still have some work to do. The first is to explain them. Many of our students, after finishing the econometrics, say: my conclusion is that x and y are positively correlated, and x and z are negatively correlated. This is meaningless — just like making a mathematical model: after you make it, what is its economic meaning? You have to explain why it is positive, why it is negative, why there is a nonlinear relationship. Also, everyone should distinguish two significances. One is econometric significance, which refers to whether the relationship between x and y is statistically significant. The other is economic significance, which refers to how large the coefficient is — the economic significance, which can tell you how important a factor is. Then on this foundation you need to get the policy implication.

If you did as I described just now, I want to congratulate you, because you are already advancing along the right road toward becoming a good economist.



#### 5. Some Comments

Next I will make a few concluding comments. First, although I said earlier how important econometrics is, the first comment I want to make to everyone now is: large sample plus econometrics is not everything. That is, don't think that the only path to doing econometrics and empirical work is to do large samples plus regression. First, much of what econometrics does is often still not enough to identify the mechanism of action between two variables. At this level, econometrics and theory are complementary. Theory can help you explain why x and y are positively correlated and what the mechanism is; what empirics helps you see is only that x and y are positively correlated — that's all.

Besides theory helping us think about these mechanistic things, cases are sometimes also very important. We economists now look down on cases a great deal, but I think cases are important, because sometimes the mechanism is unclear in the econometrics, and at such times you need to do some case studies and observe some realities to tell you why, in the econometrics, you may see that x and y are correlated.

The second situation is that sometimes many things cannot be measured. A premise econometrics relies on is that everything can be measured, but can everything be measured? For example, what is an institution? What is law? What is property rights? What is democracy? In fact, sometimes some measurable things may not accurately capture the mechanisms in real life.

Third, sometimes data is almost unobtainable, but the question is very important. For example, last semester I gave a report at the School of Economics discussing privatization and corruption in China, using cases to tell everyone that the reason privatization leads to corruption in China is that we set up a very wrong institutional structure before. This question is very important, but if you want to use an econometrician to tell everyone that privatization caused corruption, you need a large-scale sample. For example, you'd need 500 firms, and you'd need to find a measure of corruption — can you get the size of corruption through a questionnaire? So in this situation cases become very important. Having said this, I especially want to emphasize: econometrics has its advantages — it is large-sample and relatively credible. So on questions that can be measured and where data can be collected, try as much as possible to use econometric methods. Cases should be used on questions where data is almost unobtainable and where the question and the mechanism are very important. You must never do something everyone agrees can be done with econometrics and say you did it with cases, because then no one will believe you, since a case may have only a few observation points.

In 2004 I published an article, "Carrying Empirical Research Through to the End," and there are a few sentences I want to show everyone. First, econometrics has traps. As I said just now, different people have different moral bottom lines, and what each person reports depends on himself. Under current academic norms, this is very hard to achieve. Abroad, if you publish an empirical article, your data must be made public so that your reviewers can replicate it; at that point you won't dare report results casually.

Second, you will find that in the development of empirical research, there is often an empirical article saying that the previous results are untrustworthy and that the results I got are trustworthy, yet the results are the opposite. Then everyone panics. Why? Because in theory, you only need to prove that x and y are positively correlated, and it is indeed correct on paper. But in empirical work, after it has been found that x and y are positively correlated, you clearly find they are unrelated, and then you panic, fearing that three years later your article will have no citations. But I think a question needs to be resolved here: what exactly is econometrics? The theoretical tools of economics, especially the econometric tools, are like the telescope we use to study astronomical phenomena — how far a telescope can see corresponds to the degree of scientificity of the econometric method.

Econometric methods are constantly advancing. As methods advance, results obtained before may well become untrustworthy. It's like looking at the world through a telescope: up to now you haven't seen God, so you think God doesn't exist? Not necessarily. It's just that the current telescope can't see it. In this sense, science in the end actually approaches theology infinitely. So we must not be afraid, feeling that our methods aren't advanced enough and that we haven't seen the right thing. In this sense, I want to say that, viewed with a developmental eye, all research is wrong. Because without the original geocentric theory, we would still think the world rests on the back of a giant turtle. It was precisely because of the geocentric theory that the heliocentric theory gradually came later.

Next, look at a cartoon. It tells the meaning of SARS: many blind people feel an elephant; one says it's a terrorist attack, another says it's a virus, another says it's bacteria — just as different econometric approaches may get different results. Who should we believe? Feldstein once said: "The significance of this story is not that each blind man goes away with a one-sided, 'incorrect' impression, but that the wise prince (the economist), after studying the findings of the five blind men, can piece the parts together into a correct judgment of the elephant — especially if he had seen other four-legged animals before."

What does this mean? It means the relationship between econometric research and our accurate understanding of the world is this: it is not that you should believe any one article. If you want to form an overall view of the world, you put many empirical articles together. Each article may be like a blind man feeling an elephant, but for you yourself — for a good economist — you should put these articles together and form a relatively complete view of the world.

For example, the relationship between inequality and growth I just mentioned: most people's research holds that they are negatively correlated, and at that point you should believe the relationship really is negative. So this is why the same question is done repeatedly by different people — some improve the method, some improve the data — because in the process of doing it continuously and repeating it continuously, we gain a clearer understanding of the relationship between variables.

Having said all this, it is about time to end. Let our economic research move toward reality, and let us carry empirical research through to the end.


---

> Author: [滑翔闪](https://github.com/hzp2333/hzp2333.github.io)  
> URL: https://blog.huaxiangshan.com/en/posts/xz/