(ZT) 今天的失业率报告是扯淡的。统计学上站不住脚

nasd

Author: 大千的股弟   
      
统计学上有一个survivor bias。意思是说如果你只统计活下来的，而没有统计死掉的，那你的样本就是有问题的。

一个常常见到的扯淡的投资股市的理由，是说如果你按照某个规则选10个股票在十年前投资的话， 你现在就怎样怎样了。结论就是你应当投资股票，而且要按照他的规则选股票。

实际上这是扯淡的，因为十年后的今天选股票肯定不会选那些在这十年里死掉的那些股票，如果按照他的原则选股票，但是把那些死去的股票都包括在遴选过程中的话，有些死去的股票可能会被选中，那样的话你的投资肯定就不会这么好。

今天的这个失业率的报告让人大跌眼镜：预计有520，000但实际只有345，000新的事业。这是一大利好，表明就业市场改善了很多。

但是问题就在于他们的统计是根据payroll进行的，如何统计那些已经关门的公司的payroll？这是一个明显的survivor bias。这么大的改善，很可能来自于样本偏差。

xiaobailong

但是问题就在于他们的统计是根据payroll进行的，如何统计那些已经关门的公司的payroll？这是一个明显的survivor bias。这么大的改善，很可能来自于样本偏差。
nasd 发表于 2009-6-5 18:15

如果连这个都不考虑进去的话，这项统计数字是不是太儿戏了啊？

QuickHand

统计局就是拿执照的DATA COOK公司
ＧＳ就是拿执照的土匪
警察就是拿执照的...
医生就是拿执照的...

polopolo

There is something called "Birth/Death model"

Oh well, that one is not much better than fake, though.

小亥

true number=p_死了*（some number）+(1-p_死了）*current number

As long as p_死了是死猫就好了。

1# nasd

Persistance

1# nasd


In order to win, MMs and Fed can do anything under current situation.  Comparing to printing tons of money, change rules of  M2M, cooking data is nothing.  And they still have the chance to correct it later.
May is the beginning month of summer season, service industries should hire a lot of people.  NFP number should be positive.

Fig

统计局就是拿执照的DATA COOK公司
ＧＳ就是拿执照的土匪
警察就是拿执照的...
医生就是拿执照的...
QuickHand 发表于 2009-6-5 14:26

试一下填空：

统计局就是拿执照的DATA COOK公司
ＧＳ就是拿执照的土匪
警察就是拿执照的...黑社会
医生就是拿执照的...侩子手

超级潜水员

好像这个没有Survivorship Bias吧？失业率是用在找工作的人去除总的就业人口，这两个数据都是可以统计的，不存在Survivorship Bias。当然失业率有它的问题，最有名的就是因找不到工作而放弃的就不会在失业率中体现，但这和Survivorship Bias完全是两码事。另一个问题就是抽样偏差，因为调查失业率的数据其实不多。
最主要的问题是：什么是预期值？所谓的预期完全是几个经济学家的看法平均。如果你去分析历史数据的话，就会发生大部分的估计都是有问题的。像今天这样的偏差比比皆是，不要把它上纲上线。

QuickHand

8# Fig

老大, 想到有的老大是干医生和警察的, 没好意思写出来

小亥

谈的是 number of payroll reductions：

"Job losses -- while slower in May -- were still widespread. Construction companies cut 59,000 jobs, down from 108,000 in April. Factories cut 156,000, on top of 154,000 in the previous month. Retailers cut 17,500 positions, compared with 36,500, while financial activities cut 30,000, down from 45,000. Education, health care, leisure and hospitality were among the industries adding jobs in May. "

小亥

# of jobs has nothing to do with how much they earn.

Amy999

You are referring to the reduced payroll. That's why this number is out along with the unemployment which is 9.4%. Diff. definition.

coexist

"If job losses were smaller than expected, you might ask, then why did the unemployment rate go up? As it turns out, 350,000 people entered the labor force (those employed or looking for jobs) during the month. This could be a function of people who had previously given up their job searches putting their names out. As a result, the number of unemployed persons increased by 787,000 during the month to 14.5 million, the highest number on record going back to 1948."

Stein

不过，你如何考虑大众的心理因素呢？大家看到失业率下降，当然会对前景看好。

还有人为因素，大本一句话就会使股市下挫，某机构的点评会使某股停滞不前。

what is your comment about that?

reffo

?失业率明显在增加亚。这几天裁员还在进行，10%+几成定局

不过，你如何考虑大众的心理因素呢？大家看到失业率下降，当然会对前景看好。

还有人为因素，大本一句话就会使股市下挫，某机构的点评会使某股停滞不前。

what is your comment about that?
Stein 发表于 2009-6-6 02:04

finalfantasy

扯不扯并不重要，这反映了政府护盘的态度，就冲这个下周还是涨

happy_trader

还有信谢国忠的?

happy_trader

还有信谢国忠的?

happy_trader

还有信谢国忠的?