|
|
Post by danieldotwaweru on Jan 29, 2012 3:24:53 GMT 3
You get a "C" for trying to rebut what I said but that's an improvement from what you previously posted over what I said. You're far too kind. First, it is clear you are fond of using the word "nonsense," even clearly where it does not apply and my whole response was prove just that point and did so. Yes, your earlier post was nonsense. The words were clearly arranged in a grammatical order, and appeared to make sense. It looked as though they were an attempt to make some claim or other. In fact, they weren't, as should be obvious from your indifference to the facts. But see below for a fuller explanation. You now say that my saying no bachelor has been elected president in the US is "nonsense," because you googled and found out that Buchanan and Cleveland were elected as bachelors more than 127 years ago. Grover Cleveland's granddaughter was Philippa Foot, whose Natural Goodness I read and enjoyed years ago. I knew that her grandfather was a bachelor when he entered the White House. My other candidate was John Tyler, but I was unsure. So I went over to wikipedia to check. Luckily: because, although Tyler's wife died in the White House, he was married when he became President. I needed another example to be safe, and wikipedia supplied Buchanan. Did it ever occur to you I was referring to American modern history? Not really, no. Or rather, to be more precise, I have no idea what your next fallback position will be. You first stated your claim without qualification. Then you qualified it to democracies: The proposition is this simple: In politics, and I am talking about politics in functioning democracies as opposed to other forms of government, a leader generally first establishes their credentials locally and then progresses along the ladder to the top. In fact, a subset of this theory holds that one must demonstrate leadership at his or her very home, measured by whether or not they have family and if they do, how well they have raised that family and/or are still taking care of it. before confining it only to model democracies. (Apparently, this would include America.) You've now qualified yet again, to limit the claim to modern American politics. A claim so narrow is (i) incompatible with your earlier, broader formulation (ii) useless because uninformative. Many modern democracies have seen bachelors (or bachelorettes) in power: Edward Heath in the UK; Halonen in Finland; Kaczynski in Poland; Gillard in Australia (though she lives with her partner); Sigurðardóttir in Iceland (though she later married her partner). The list could be extended. Your latest most-qualified claim says nothing about them, so tell us nothing interesting about democracy. Suffice it to say that either large numbers of what we thought were democracies are not democracies, or your criterion is no use. Or that the point I made is nonetheless still supported by the very fact that no bachelor has been elected president in all those years and probably never will? As I've already indicated, you have a problem. If you insist on limiting it to modern American politics, then the generalisation is no use. Other democratic countries elect unmarried heads of state without an eyelid batted. So the modern American reluctance to elect unmarried Presidents shows nothing of interest---in roughly the same way that the modern American reluctance to elect atheists to office tells us nothing about how democracies work in other parts of the world. In any case, even if you were to assume, arguendo, that the point was contradicted by the missing "in modern history" in my piece or even by the fact these two presidents were elected as bachelors 127 years ago, would it therefore follow that what I said about that point is "nonsense" or merely incorrect? It is nonsense. That is, it was a claim without any serious attempt to communicate the facts. Perhaps you'll see what I'm getting if you turn to an example from Harry Frankfurt (though originally from Wittgenstein) when he's explaining what BS is: Frankfurt asks us to consider an anecdote told about Ludwig Wittgenstein wherein the great philosopher phones a friend named Fania Pascal who's just had her tonsils removed. How are you, Wittgenstein asks. Like a dog that's been run over, Pascal answers. Wittgenstein then replies testily, "You don't know what a dog that has been run over feels like."....
Is Pascal lying? No. She isn't trying to deceive Wittgenstein about how she really feels, and she isn't trying to deceive Wittgenstein about how a dog would feel if run over. Her error, Frankfurt concludes, isn't that she conducted a faulty inquiry into how a dog would feel if run over, but that she conducted no inquiry at all (in this case, because none is possible)."It is just this lack of connection to a concern with truth—this indifference to how things really are—that I regard as the essence of bullshit." The indifference to truth which Frankfurt points out here has elsewhere been called nonsense. I'm happy to award your efforts that title. Second, talking about differences, your saying you used Obasanjo as a "counterexample" as apposed to "analogizing" as I said you had is a distinction without a difference in the context I stated my proposition. Nope. Analogies tend to have a structure of the following kind: you argue that X. I produce an analogy for X by pointing to Y which shares some relevant property of X. The argument then proceeds from the claim of some shared feature to the claim that they share some other relevant feature. Counterexamples work the other way round. You argue for X---in this case, you argued that there is some feature shared by all successful politicians. A counterexample issues when I can find an example that does not have the relevant feature. In this case, Obasanjo did the trick, since he's a successful politician who was absolutely loathed in his homeland. Third, talking about context, Obama moved from Hawaii and made Illinois his adopted state and started his political career there, not his birthplace but that's precisely what the point is I was making, namely, Obama did not land at the White House without having first established his political launching pad locally elsewhere and in this context, locally is Illinois, not Hawaii. Was he born in Chicago? No. he transplanted himself to Chicago, and started his career there, before returning home when he had succeeded. This is a counterexample to your claim that no one can succeed without first establishing himself locally---as one can tell from the fact that you did not contest the Obasanjo example by saying that he had established himself 'locally' in a non-Yoruba area. Further, since you included family and home in one of the earlier formulations, it was obvious that you did not mean merely establishing oneself in some place or other, but rather meant establishing oneself in one's home, whether understood in some ethnic or familial sense. You've now redefined successful politicians must establish themselves locally to mean something like successful politicians must establish themselves in some place or other before they achieve high office. This is a sign that we have reached the end of the argument. Many---perhaps most---exciting claims turn out, on inspection, to be either trivial or false. Your original claim about politicians having to establish themselves locally turned out to be false on the obvious reading (too many examples of politicians who have succeeded despite establishing themselves away from home) so you have gone for the trivial reading (politicians have to start somewhere). The original argument, one might say, has died the death of a thousand cuts, since it has been qualified so often that there is nothing of the original left. |
|
|
|
Post by job on Jan 29, 2012 3:28:25 GMT 3
|
|
|
|
Post by job on Jan 29, 2012 4:19:04 GMT 3
Corridors of Power SATURDAY, 28 JANUARY 2012 00:08 BY POLITICAL DESK, The Star Supporters of a prominent politician who hope that Uhuru Kenyatta and William Ruto will endorse the VP if their quest for presidency hits a snag should think twice, if SMSes we have seen are anything to go by. An MP close to Uhuru sent Corridors a text reading: “We have tasked UK and William to help identify a neutralpresidential candidate to face off with ODM. The political opportunists and hyenas who have been hanging around for ICC spoils shall be in for a rudeshock.”
|
|
|
|
Post by Omwenga on Jan 29, 2012 6:31:41 GMT 3
You get a "C" for trying to rebut what I said but that's an improvement from what you previously posted over what I said. You're far too kind. Yes, your earlier post was nonsense. The words were clearly arranged in a grammatical order, and appeared to make sense. It looked as though they were an attempt to make some claim or other. In fact, they weren't, as should be obvious from your indifference to the facts. But see below for a fuller explanation. Grover Cleveland's granddaughter was Philippa Foot, whose Natural Goodness I read and enjoyed years ago. I knew that her grandfather was a bachelor when he entered the White House. My other candidate was John Tyler, but I was unsure. So I went over to wikipedia to check. Luckily: because, although Tyler's wife died in the White House, he was married when he became President. I needed another example to be safe, and wikipedia supplied Buchanan. Not really, no. Or rather, to be more precise, I have no idea what your next fallback position will be. You first stated your claim without qualification. Then you qualified it to democracies: before confining it only to model democracies. (Apparently, this would include America.) You've now qualified yet again, to limit the claim to modern American politics. A claim so narrow is (i) incompatible with your earlier, broader formulation (ii) useless because uninformative. Many modern democracies have seen bachelors (or bachelorettes) in power: Edward Heath in the UK; Halonen in Finland; Kaczynski in Poland; Gillard in Australia (though she lives with her partner); Sigurðardóttir in Iceland (though she later married her partner). The list could be extended. Your latest most-qualified claim says nothing about them, so tell us nothing interesting about democracy. Suffice it to say that either large numbers of what we thought were democracies are not democracies, or your criterion is no use. As I've already indicated, you have a problem. If you insist on limiting it to modern American politics, then the generalisation is no use. Other democratic countries elect unmarried heads of state without an eyelid batted. So the modern American reluctance to elect unmarried Presidents shows nothing of interest---in roughly the same way that the modern American reluctance to elect atheists to office tells us nothing about how democracies work in other parts of the world. It is nonsense. That is, it was a claim without any serious attempt to communicate the facts. Perhaps you'll see what I'm getting if you turn to an example from Harry Frankfurt (though originally from Wittgenstein) when he's explaining what BS is: Frankfurt asks us to consider an anecdote told about Ludwig Wittgenstein wherein the great philosopher phones a friend named Fania Pascal who's just had her tonsils removed. How are you, Wittgenstein asks. Like a dog that's been run over, Pascal answers. Wittgenstein then replies testily, "You don't know what a dog that has been run over feels like."....
Is Pascal lying? No. She isn't trying to deceive Wittgenstein about how she really feels, and she isn't trying to deceive Wittgenstein about how a dog would feel if run over. Her error, Frankfurt concludes, isn't that she conducted a faulty inquiry into how a dog would feel if run over, but that she conducted no inquiry at all (in this case, because none is possible)."It is just this lack of connection to a concern with truth—this indifference to how things really are—that I regard as the essence of bullshit." The indifference to truth which Frankfurt points out here has elsewhere been called nonsense. I'm happy to award your efforts that title. Nope. Analogies tend to have a structure of the following kind: you argue that X. I produce an analogy for X by pointing to Y which shares some relevant property of X. The argument then proceeds from the claim of some shared feature to the claim that they share some other relevant feature. Counterexamples work the other way round. You argue for X---in this case, you argued that there is some feature shared by all successful politicians. A counterexample issues when I can find an example that does not have the relevant feature. In this case, Obasanjo did the trick, since he's a successful politician who was absolutely loathed in his homeland. Third, talking about context, Obama moved from Hawaii and made Illinois his adopted state and started his political career there, not his birthplace but that's precisely what the point is I was making, namely, Obama did not land at the White House without having first established his political launching pad locally elsewhere and in this context, locally is Illinois, not Hawaii. Was he born in Chicago? No. he transplanted himself to Chicago, and started his career there, before returning home when he had succeeded. This is a counterexample to your claim that no one can succeed without first establishing himself locally---as one can tell from the fact that you did not contest the Obasanjo example by saying that he had established himself 'locally' in a non-Yoruba area. Further, since you included family and home in one of the earlier formulations, it was obvious that you did not mean merely establishing oneself in some place or other, but rather meant establishing oneself in one's home, whether understood in some ethnic or familial sense. You've now redefined successful politicians must establish themselves locally to mean something like successful politicians must establish themselves in some place or other before they achieve high office. This is a sign that we have reached the end of the argument. Many---perhaps most---exciting claims turn out, on inspection, to be either trivial or false. Your original claim about politicians having to establish themselves locally turned out to be false on the obvious reading (too many examples of politicians who have succeeded despite establishing themselves away from home) so you have gone for the trivial reading (politicians have to start somewhere). The original argument, one might say, has died the death of a thousand cuts, since it has been qualified so often that there is nothing of the original left. There is a point beyond which all arguments become nonsensical or without a purpose other than to advance the useless. We are at that point. P.S. Sitting in my bookshelf is an autobiography of Felix Frankfurter, one of the finest American jurists I admire his intellect not necessarily his judicial philosophy. I have contemplated lending several of these autobiographies to a friend sitting in our Kenya Supreme Court but only if they are interested. I am confident the new court and judicial reforms underway will in due course make a leader in African judicial philosophy and, to that end, I assume these justices are reading widely and deeply as they embark on this daunting but manageable task. |
|
|
|
Post by merlin on Jan 29, 2012 15:24:48 GMT 3
You are correct; I did not have any specific scientific evidence for my claim. I appreciate your effort to direct me to the study from the University of Connecticut regarding the elections in 2007. It confirms my proposition regarding mistrust between tribes and the need of control. (1) You presented no evidence whatever. The problem was the lack of evidence, not the lack of scientific evidence. (2) The study does no such thing. It confirms what has been found by a variety of researchers going back many years: that, of the major ethnic groups in Kenya, Luo are the least trusting of those outside their group. And this remained true after a period of several years--2005 to 2008---during which the politics of the country revolved around continuous Gikuyu hatred, and after a period of 15 years in which the government of the day carried out a campaign of rolling ethnic cleansing. Even after all that, it turns out Gikuyu aren't the least trusting of outsiders. These findings present a problem for your reasoning. ODMers tend to reason from the premiss that Gikuyu don't trust others to the conclusion that a variety of nasty things ought to be done to them. We can test their reasoning by checking whether the same things ought to be done to Luo. After all, Luo have even lower levels of inter-ethnic trust than Gikuyu. Naturally, ODMers have no interest whatever in suggesting that the nasty things should be done to Luo. That is how you can tell that the original anti-Gikuyu point had nothing to do with trust. Another way to reason to the same finding is to notice that ODM's ethnic baiting depends on pairs of directly contradictory arguments. For example: ODMers will argue that Gikuyu are incapable of living and cooperating with anyone else, while, on the hand, they will also argue that Gikuyu were mostly colonial collaborators. Likewise, ODMers will argue that Gikuyu are incapable of trusting anyone else, while at the same time arguing that Gikuyu owe their privilege to their cooperation with colonial power. Obviously not. You simply assumed that some bad thing must be true of Gikuyu, and proceeded to reason from that assumption. The problem is that Luo are significantly worse in that respect. In the case of Luo, your willingness to reason in the same way is significantly reduced. Interesting is also your reference/link further down the thread to “Wars, Guns, and Votes: Democracy in Dangerous Places” by Paul Collier. It gives only promotional write-up to buy the book and no information which could substantiate your claim. Follow the link. It will take you to Googlebooks. In particular, it will take you to page 70 of the Googlebooks version of Wars, Guns and Votes where Collier summarises his evidence of Raila's conduct regarding ethnic cleansing and ethnic mobilisation. You will be unsurprised, I take it, to hear that Raila was the leader in ethnic mobilisation, and that he campaigned on what amounted to a platform for ethnic cleansing. (I await your argument that Collier is a Gikuyu nationalist). I am not giving up yet. The essence of my entry is as follows, 1 Kikuyus don’t trust other tribes,
2 This forces them to be in control at any cost,
3 This isolates them from the rest of Kenya,
4 Hence a vicious circle,
5 The option ”at any cost” limits the choice for good leadership,
6 Leading to offering leaders of disputable qualities to lead the Nation,
7 The Kikuyu community is not homogeneous there are good leaders available within the community,
8 The feeling of mistrust offers a strong handle to bad leaders to manipulate them,
9 They should consider the cost of being in control,
10 The ICC shows the cost to be in control,
11 They should choose leaders who can break the vicious circle The entry should be read as a proposition for discussion as I have no scientific proof. However I do not understand your reaction about which tribe is most untrusting. We seem to be discussing about different subjects. |
|
|
|
Post by einstein on Jan 29, 2012 15:26:17 GMT 3
(2) The study does no such thing. It confirms what has been found by a variety of researchers going back many years: that, of the major ethnic groups in Kenya, Luo are the least trusting of those outside their group. And this remained true after a period of several years--2005 to 2008---during which the politics of the country revolved around continuous Gikuyu hatred, and after a period of 15 years in which the government of the day carried out a campaign of rolling ethnic cleansing. Even after all that, it turns out Gikuyu aren't the least trusting of outsiders. 20.8 (Not at all ); 42.0 (Just a little ); 28.8 (Somewhat); 7.5 ( A lot) Luo: 20.3 (Not at all ); 41.9 (Just a little ); 30.4 (Somewhat); 4.7 ( A lot) Folks,I need help. I think my math is letting me down here. Danielwaweru says, according to the statistics above, Luo are the least trusting of those outside their group. Is this really what the stats above indicate? Please also note that while the total % for Kikuyu in all categories of response sums up to 99.1%, for Luo it sums up to 97.3%. How does one explain this discrepancy in this survey? Source: www.econ.uconn.edu/working/2008-35.pdf on page 5 |
|
|
|
Post by merlin on Jan 29, 2012 17:16:35 GMT 3
Danielwaweru,
Onestone’s entry confirms my claim; you very successfully thrown your sabot into the machinery which also is a form of control though mostly destructive. You successfully polluted this thread as the subject has changed to; who is the most untrusting tribe in Kenya.
|
|
|
|
Post by danieldotwaweru on Jan 29, 2012 17:35:56 GMT 3
Danielwaweru says, according to the statistics above, Luo are the least trusting of those outside their group. Is this really what the stats above indicate? Yes. They're level pegging until the last column, when we find that far fewer Luo trust outsiders a lot. Another way to see the point is this: sum the numbers for the positive columns (trust a lot, and trust somewhat) and subtract the sum of the negative columns (trust a little or trust not at all). A larger deficit indicates lower trust. For Luo, (30.4+4.7)-(20.3+41.9) gives you -27.1; for Gikuyu (28.8 + 7.5) - (20.8 + 42.0) = -26.5. For yet another example of the same result, see p. 11 and Fig. 2 on p. 12 of Kasara 2011 Separate and Suspicious: Local Social and Political Context and Ethnic Tolerance in Kenya. Look at the double table labelled "Trust By Ethnic Group". On the left, it gives her results for how much members of an ethnic group trust their families. On the right side, it tabulates her results for trust of members of other ethic groups. In both tables, Luo are the outlier on the left-hand side. That is Luo trust their relatives less than any of the other groups; and they trust members of other ethnic groups less than any of the other groups. Please also note that while the total % for Kikuyu in all category of questions sums up to 99.1%, for Luo it sums up to 97.3%. How does one explain this discrepancy in this survey? As you'll often find in surveys of this kind, rounding means that not all columns add up to 100. |
|
|
|
Post by danieldotwaweru on Jan 29, 2012 19:26:04 GMT 3
Danielwaweru,Onestone’s entry confirms my claim; you very successfully thrown your sabot into the machinery which also is a form of control though mostly destructive. You successfully polluted this thread as the subject has changed to; who is the most untrusting tribe in Kenya. Standard causal test. Suppose you say that X causes Y. If X comes in degrees---if you can have more or less of X---then one obvious test is to check whether Y varies with the degree of X. If it does not, then the causal claim is in trouble. For a very crude example: if you say that chewing miraa causes sleeplessness, then chewing more miraa should cause more sleeplessness, and chewing less miraa should cause less. Your story is slightly more complicated than that, since the cause and effect vary inversely: that is, lower trust means a higher chance of all sorts of nasty things (Let chance of nasty things be P(N). Then your theory is roughly that P(N)=1/trust). But the same causal test applies. Yet another crude example: gravitational force is supposed to vary inversely with distance^2. That is, g=1/d^2. You can check that theory by decreasing the distance, which should increase the gravitational force. One can check the causal story you want to tell about trust in the same way. You suggest that low trust (in the Gikuyu case) causes all sorts of things, including the PEV. So if we reduce the trust even further, all the nasty things you identify should become more likely, since P(N)=1/trust. The lower trust of Luo respondents should make all the nasty consequences you identify more likely. Unfortunately, there's (i) not a lot of evidence in favour of that proposition, and (ii) virtually none of those who would employ that reasoning in the Gikuyu case are willing to employ it in the Luo case. So much for this particular causal story. |
|
|
|
Post by einstein on Jan 29, 2012 20:47:10 GMT 3
Danielwaweru says, according to the statistics above, Luo are the least trusting of those outside their group. Is this really what the stats above indicate? As I still take time to munch on the rest of your response above, please note that this assertion from yourself is NOT scientific at all in the realm of statistical surveys: "As you'll often find in surveys of this kind, rounding means that not all columns add up to 100."In statistics, a difference of 0.9% or 2.7% is a VERY HUGE difference!If you are going to talk about percentages in any scientific statistical survey, they MUST add up to 100%! I do not know what they teach you in the US of A or wherever you are based, but in Kenya as well as my current country of residence my professors taught me that any time I mention percentages the final analysis MUST add up to 100 %!! If it does not, it means I have forgotten or left out SOMETHING! This is the statistics rule 101 aka stats for beginners. If you asked me, that working paper you quoted is totally flawed. The final analysis was not done by somebody well versed in the realm of statistics. I'm just wondering aloud, is this how some folks get their PhDs?? Really?? I will get back to you on the rest of your mathematical/statistical analysis as soon as time allows. In the meantime, please respond scientifically to the issue about percentages that do not add up to 100% as stated above. OO, I would like you to follow this discussion with very keen interest. I hope you are not allowing a repeat pathological hate-monger to prowl the board unhindered! |
|
|
|
Post by merlin on Jan 30, 2012 0:11:06 GMT 3
Danielwaweru,Onestone’s entry confirms my claim; you very successfully thrown your sabot into the machinery which also is a form of control though mostly destructive. You successfully polluted this thread as the subject has changed to; who is the most untrusting tribe in Kenya. Standard causal test. Suppose you say that X causes Y. If X comes in degrees---if you can have more or less of X---then one obvious test is to check whether Y varies with the degree of X. If it does not, then the causal claim is in trouble. For a very crude example: if you say that chewing miraa causes sleeplessness, then chewing more miraa should cause more sleeplessness, and chewing less miraa should cause less. Your story is slightly more complicated than that, since the cause and effect vary inversely: that is, lower trust means a higher chance of all sorts of nasty things (Let chance of nasty things be P(N). Then your theory is roughly that P(N)=1/trust). But the same causal test applies. Yet another crude example: gravitational force is supposed to vary inversely with distance^2. That is, g=1/d^2. You can check that theory by decreasing the distance, which should increase the gravitational force. One can check the causal story you want to tell about trust in the same way. You suggest that low trust (in the Gikuyu case) causes all sorts of things, including the PEV. So if we reduce the trust even further, all the nasty things you identify should become more likely, since P(N)=1/trust. The lower trust of Luo respondents should make all the nasty consequences you identify more likely. Unfortunately, there's (i) not a lot of evidence in favour of that proposition, and (ii) virtually none of those who would employ that reasoning in the Gikuyu case are willing to employ it in the Luo case. So much for this particular causal story. I like to discuss the consequences of distrust between groups though will not get lured into a tribal discussion which tribe is more distrusting than another. What I am looking for is a discussion to discover a way-out of the distrust situation and its negative consequences. Although I am of the opinion that you have other objectives polluting the thread with a considerable amount of data and information to burry opinion and statements by others which you do not like to be exposed, did I find some useful information in your reference to which state on page 20 International experience has shown that when groups distrust each other and are afraid of being victimized, this fear might drive them to resort to violence first in a preemptive move to minimize damage (Bardhan, 1997). This might explain for the animosity towards violence observed after the disputed elections. In December 2007, 12% of respondents answered that it is “Sometimes necessary to use violence in support of a just cause”, this figure went up to 21% by August 2008. In other words today, one in four Kenyans say that violence is justified. The survey also shows that Kenyans are divided on whether amnesty should be given to those who were involved in the outbreaks of violence. Roughly 43% of respondents disagree with granting amnesty, whilst 45% agree amnesty should be given. A curious part of this report is; who won the 2007 presidential elections. However this issue has become redundant as we have to accept the situation we find ourselves in. We have to look forward to the next elections and can ask ourselves have leaders understood the consequences of distrust between groups and stopped to incite animosity and distrust between groups? If this has to be answered negatively then what can we do as JUKWAA discussion group and opinion formers to deflate distrust between groups or how to prepare ourselves for another round of mayhem and destruction? |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 1:29:33 GMT 3
If you are going to talk about percentages in any scientific statistical survey, they MUST add up to 100%! I do not know what they teach you in the US of A or wherever you are based, but in Kenya as well as my current country of residence my professors taught me that any time I mention percentages the final analysis MUST add up to 100 %! This is unambiguously false. Stats papers often have a disclaimer explaining that percentages will not sum to 100, because of rounding errors. I actually looked up examples on stats.sx and stack overflow just to make sure, and what do you know, it's perfectly standard. The stackoverflow example is pretty good since it raises a good point in favour of rounding errors above and below 100. OO, I would like you to follow this discussion with very keen interest. I hope you are not allowing a repeat pathological hate-monger to prowl the board unhindered! It's unclear what you're on about. Merlin said that Gikuyu don't trust anyone, which made them more likely to resort to violence (or whatever). One reasonable reply to that is to ask whether he's looked at the evidence, since that evidence unambiguously shows that even lower levels of trust elsewhere. It's not obvious how you can argue that I'm a hatemonger, while leaving Merlin well alone, but I imagine you'll explain soon enough. |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 1:35:55 GMT 3
I like to discuss the consequences of distrust between groups though will not get lured into a tribal discussion which tribe is more distrusting than another. Is that so? You could have fooled me: your began your discussion by explaining how horrible, nasty and untrusting Gikuyu were. That's a very strange way to begin a sober discussion about inter-ethnic trust. What I am looking for is a discussion to discover a way-out of the distrust situation and its negative consequences. We're among friends and honourable enemies here. It's OK to admit you wanted to get some Gikuyu-bashing done. Unfortunately, you chose ineptly. Iwe funzo kwako na wengine wenye nia kama hizo. Triggers and Characteristics of the 2007 Kenyan Electoral Violence ....International experience has shown that when groups distrust each other and are afraid of being victimized, this fear might drive them to resort to violence first in a preemptive move to minimize damage (Bardhan, 1997). This might explain for the animosity towards violence observed after the disputed elections. In December 2007, 12% of respondents answered that it is “Sometimes necessary to use violence in support of a just cause”, this figure went up to 21% by August 2008. In other words today, one in four Kenyans say that violence is justified. The survey also shows that Kenyans are divided on whether amnesty should be given to those who were involved in the outbreaks of violence. Thanks. I've read the paper; I've quoted it before on Jukwaa; I'm writing a book about ethnic violence in Kenya; and I'm perfectly familiar with the first movers where ethnic violence was concerned in Kenya. In particular, political violence in Kenya was not the result of some spontaneous outbreak of mistrust. Rather, the mistrust was cultivated, encouraged by several years of propaganda. The violence which succeeded it was planned, paid for and eventually perpetrated by structured organisations. There was no clean causal path from mistrust to violence, and the mistrust itself is the product of the planning for violence, not the other way round. In my view, the most useful model here is that developed in The Political Economy of Hatred, especially once you take note of some crucial facts: (i) ODM was an opposition party, (ii) the creation of the requisite hatred had to occur in a society where the groups are relatively mixed, and (iii) the information which was used to create the mistrust and hatred had to be transmitted in a variety of languages (since ODM's coalition consists of speakers of different languages). But you don't, of course, have to take my word for it. The key points are briefly stated: Violence against the Kikuyu was a deliberate strategy electoral strategy of Raila Odinga (Collier 2009: 72) and Raila Odinga ran a campaign that was tantamount to promising ethnic cleansing (Collier 2009: 70) and by far the main culprit [in ethnic mobilization] was the opposition leader, Raila Odinga. Recall that the incumbent has the advantage in respect of bribery and miscounting, so the opposition is indeed more likely to resort to the cheaper strategy of playing on ethnic identity" (Collier 2009: 70). So. Your point, caller? |
|
|
|
Post by mank on Jan 30, 2012 2:15:05 GMT 3
(2) The study does no such thing. It confirms what has been found by a variety of researchers going back many years: that, of the major ethnic groups in Kenya, Luo are the least trusting of those outside their group. And this remained true after a period of several years--2005 to 2008---during which the politics of the country revolved around continuous Gikuyu hatred, and after a period of 15 years in which the government of the day carried out a campaign of rolling ethnic cleansing. Even after all that, it turns out Gikuyu aren't the least trusting of outsiders.
How much do you Trust Kenyans from other Ethnic Groups in %
Kikuyu: 20.8 (Not at all ); 42.0 (Just a little ); 28.8 (Somewhat); 7.5 ( A lot)
Luo: 20.3 (Not at all ); 41.9 (Just a little ); 30.4 (Somewhat); 4.7 ( A lot)
Folks,
I need help. I think my math is letting me down here.
Danielwaweru says, according to the statistics above, Luo are the least trusting of those outside their group. Is this really what the stats above indicate?
Please also note that while the total % for Kikuyu in all categories of response sums up to 99.1%, for Luo it sums up to 97.3%. How does one explain this discrepancy in this survey?
Source: www.econ.uconn.edu/working/2008-35.pdf on page 5The study does not confirm that the Luo are the least trusting of those outside their group. You cannot just read the magnitudes of parameter estimates in statistics and conclude that one is larger than another. You have to be equiped with information as to the statistical significance of the given magnitudes to conclude that those magnitudes are sufficient proof of a difference in the two groups for which the magnitudes speak. In this case the study does not test the statistical significance of the responses across ethnic groups. There is more. If we assume the response "a lot" is equivalent to "perfect trust or no distrust", then we would need to sum the first 3 columns to determine the percentage of interviewees that indicated that they at least somewhat distrust those outside their group. If you do that, you find that in simple magnitudes, the Kamba seem to be the least likely to trust those outside their group (at 94.6%), followed by the Luo (at 92.6%), then the Kikuyu (at 91.6%). We cannot conclude that those percentages are a confirmation of differences of distrust for those outside one's ethnic group, but we can say that the study provides basis for such an opinion ( a hypothesis that Kambas are most distruting, followed by Luos, then Kikuyus). If we want to claim that the study "confirms" such an ethnic relationship, we need a test of significance between those 3 percentages and between each of the pairs that can be compared among them (that is we need to test the hypothesis stated above). From my quick look at the paper, I do not see where the paper addressed such a test of significance. Note: The fact that we have to make an assumption like the one I suggest above points to the incompleteness (or ambiguity) of the response space, and the likely explanation of why the sums of responses are so distant from 100%. I think the percentage response is determined on the basis of the total number of interviewees while some did not answer the question , or at least did not indicate answers inside the "Not at all Just a little Somewhat A lot" space. Rounding errors alone cannot explain the wide deviation of summed %s from 100%. The signficance of the study design problem is clearer when you look at what it tells us about the community that it suggests is most trusting of outsiders - the Mijikenda, for whom only 80% seem to habor some distrust. Now, look at the summed percentage responses across the response space, and you will notice the Mijikenda were also the least responsive to the question. Now, if you look back at the ranking of "apparent likelihood to trust" and "deviation of summed responding %s from 100%", you notice a very good correlation between the two. This suggests that the imperfect design of the response space led to a selective non-response whereby the category of non responders seem to be likely to pick a response that was lost or contorted by the imperfection. |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 2:54:30 GMT 3
The study does not confirm that the Luo are the least trusting of those outside their group. You cannot just read the magnitudes of parameter estimates in statistics and conclude that one is larger than another. You have to be equiped with information as to the statistical significance of the given magnitudes to conclude that those magnitudes are sufficient proof of a difference in the two groups for which the magnitudes speak. In this case the study does not test the statistical significance of the responses across ethnic groups. The data are from their survey, and they ran significance tests on it in a previous publication. Let me see if I can dig it out. If we assume the response "a lot" is equivalent to "perfect trust or no distrust", There is no reason whatsoever to assume this. That is, if you have a threshold (perfect trust) you have no reason to assume that some arbitrarily close point to that threshold ought to be treated similarly. Lotteries are an easy way to make the point, since people are tempted to assume that a very high probability of not winning makes it rational to assume that a given ticket in a fair lottery will not win. Unfortunately, people keep proving results which show that if you make that assumption, then it either trivialises or you'll have to give up basic principles of rationality. |
|
|
|
Post by mank on Jan 30, 2012 3:05:32 GMT 3
The study does not confirm that the Luo are the least trusting of those outside their group. You cannot just read the magnitudes of parameter estimates in statistics and conclude that one is larger than another. You have to be equiped with information as to the statistical significance of the given magnitudes to conclude that those magnitudes are sufficient proof of a difference in the two groups for which the magnitudes speak. In this case the study does not test the statistical significance of the responses across ethnic groups. The data are from their survey, and they ran significance tests on it in a previous publication. Let me see if I can dig it out. If we assume the response "a lot" is equivalent to "perfect trust or no distrust", There is no reason whatsoever to assume this. That is, if you have a threshold (perfect trust) you have no reason to assume that some arbitrarily close point to that threshold ought to be treated similarly. Lotteries are an easy way to make the point, since people are tempted to assume that a very high probability of not winning makes it rational to assume that a given ticket in a fair lottery will not win. Unfortunately, people keep proving results which show that if you make that assumption, then it either trivialises or you'll have to give up basic principles of rationality. I am not sure we are thinking about the same concept here: my suggestion is that the 4 options given do not seem to perfectly cover all possible reaponses to the questions posed. Look at this, when asked "How much do you Trust Kenyans from other Ethnic Groups?", one may say "Not at all", "Just a little", "Somewhat", "same as I trust those in my group", or "more than I trust those in my group". It is not clear what "A lot" means relative to how the respondent trusts people of its own group. That ambiguity seems to be the reason not all interviwees responded to the question. As to your first statement, that test of significance was presented in earlier publications, then I would say that when making a statement that this study confirms the ethic rankings of trust you need to clarify that the confirmation is in other studies. ... otherwise my comment was only to clarify that such a claim cannot be made on the basis of what is reported in this paper. You did not make any comment on my deduction that even if the percentages were statisically significant from each other it would be the Kamba that would be "most distrusting" of outsiders (i.e. on the basis of results reported in this particular study). |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 3:18:52 GMT 3
I am not sure we are thinking about the same concept here: my suggestion is that the 4 options given do not seem to perfectly cover all possible reaponses to the questions posed. Look at this, when asked "How much do you Trust Kenyans from other Ethnic Groups?", one may say "Not at all", "Just a little", "Somewhat", "same as I trust those in my group", or "more than I trust those in my group". It is not clear what "A lot" means relative to how the respondent trust people of its own group. That ambiguity seems to be the reason not all interviwees responded to the question. Still won't do the trick. That is, if your complaint is that the responses available did not cover all possible interviewee responses then it is not a strong criticism of the study because it is not a general requirement that studies of this kind give options that match all possible interview responses. For example: in designing studies of this kind (single-response, multi-option) you begin by throwing out all possible options which are also logically incompatible e.g. you don't give respondents the option to give the same answer to two logically-incompatible options, even though some would have done so if the option had been on the menu. For example, suppose you're asking the question: Will you vote for Bush or Gore in the 2000 election? In general, you don't give people the option to answer Yes to both, even though there are some people who would answer yes. If your criticism is not the threshold point, but rather the claim that there were some possible options that could have been included but were not, then that isn't yet a reason to reject the finding. As to your first statement, that test of significance was presented in earlier publications, then I would say that when making a statement that this study confirms the ethic rankings of trust you need to clarify that the confirmation is in other studies. ... otherwise my comment was only to clarify that such a claim cannot be made the basis of what is reported in this paper. Yes, fair point. (But I did mention Kasara (2011) which has a full-ish discussion of method issues in the appendix.) |
|
|
|
Post by mank on Jan 30, 2012 3:27:58 GMT 3
I am not sure we are thinking about the same concept here: my suggestion is that the 4 options given do not seem to perfectly cover all possible reaponses to the questions posed. Look at this, when asked "How much do you Trust Kenyans from other Ethnic Groups?", one may say "Not at all", "Just a little", "Somewhat", "same as I trust those in my group", or "more than I trust those in my group". It is not clear what "A lot" means relative to how the respondent trust people of its own group. That ambiguity seems to be the reason not all interviwees responded to the question. Still won't do the trick. That is, if your complaint is that the responses available did not cover all possible interviewee responses then it is not a strong criticism of the study because it is not a general requirement that studies of this kind give options that match all possible interview responses. For example: in designing studies of this kind (single-response, multi-option) you begin by throwing out all possible options which are also logically incompatible e.g. you don't give respondents the option to give the same answer to two logically-incompatible options, even though some would have done so if the option had been on the menu. For example, suppose you're asking the question: Will you vote for Bush or Gore in the 2000 election? In general, you don't give people the option to answer Yes to both, even though there are some people who would answer yes. If your criticism is not the threshold point, but rather the claim that there were some possible options that could have been included but were not, then that isn't yet a reason to reject the finding. As to your first statement, that test of significance was presented in earlier publications, then I would say that when making a statement that this study confirms the ethic rankings of trust you need to clarify that the confirmation is in other studies. ... otherwise my comment was only to clarify that such a claim cannot be made the basis of what is reported in this paper. Yes, fair point. (But I did mention Kasara (2011) which has a full-ish discussion of method issues in the appendix.) DW, You clearly do not appreciate the significance of the flaw I am pointing out. It may not be a flaw if it was not the intention of the researchers to interprete the results in the way you interprete them. But if you are going to insist on interpreting the results the way you did, then you have to accept that there is a design issue that makes it impossble to reasonably support the "confirmation" you allege. This is not an issue of phiolosophical debate, but one of study design. I think I have sufficiently shown a design flaw of the study, and its effect - a selective non-response. |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 3:44:45 GMT 3
You did not make any comment on my deduction that even if the percentages were statisically significant from each other it would be the Kamba that would be "most distrusting" of outsiders (i.e. on the basis of results reported in this particular study). It was a fair point. However, it depends on treating the trust a lot as perfect trust. That seems to me very difficult to pull off, because people will just run an arbitrariness challenge at you. To see what I mean, let me run the test in terms of belief. Suppose someone asks you whether you believe something a bit, somewhat or a lot. Now, first of all, intuitively (although the distinction is vague) there is a difference between believing something a lot, and being certain of it. (Let certainty be the threshold here.) Second, there are intuitive (though vague) distinctions between believing something a bit, somewhat, or a lot. Those distinctions don't collapse into each other---at the qualitative level, they remain distinct. If you try to formalise them, or pick some magnitude to represent them, then you get an arbitrariness challenge that goes like this: there is clearly a difference between believing something a lot, and believing it somewhat. If you believe, a lot, that the tricycle is safe, then you can put your child on it. If you believe that the tricycle is only somewhat safe, then you shouldn't put your child on it. Now, the challenge is to find a number which expresses the cutoff point. If you say you believe the tricycle is safe when your credence (your subjective probability) is, say, 0.6, then you will be asked why the move from 0.6 to 0.5999 should make a difference. And if you accept 0.5999, you'll be asked why not 0.5998 and so on. There don't seem to be two numbers such that the magnitude of the interval between them captures the significance which was familiar from the qualitative description. This suggests that there was no problem with the original qualitative descriptions, even though they were vague. That's a long way of saying that I'm happy to stick with the original partition, even though it's not sharp. Having said all that, let me say a bit about why I don't want to dismiss your argument out of hand. I've seen other strange results for Kamba voting. In some surveys, the results will show low levels of trust, in others high levels of trust. And in the Bratton paper, we see a very high incidence of ethnic voting. So even though I don't agree with your particular argument here, I think there's an issue with Kamba data (or maybe we need some finer-grained numbers.) |
|
|
|
Post by mank on Jan 30, 2012 4:01:07 GMT 3
You did not make any comment on my deduction that even if the percentages were statisically significant from each other it would be the Kamba that would be "most distrusting" of outsiders (i.e. on the basis of results reported in this particular study).
It was a fair point. However, it depends on treating the trust a lot as perfect trust. That seems to me very difficult to pull off, because people will just run an arbitrariness challenge at you. ... Mine was not to provide an alternative design, but to interprete the results given. So, I was not suggesting alternative options for the researchers' trust a lot. What I was doing is making an assumption as to what the trust a lot could have meant to respondents. To be able to claim that these results reveal something about ranking of ethnicities by their trust of outsiders, 100% minus the percentage indicating "at least some distrust" must be the indication of "no distrust". In this case it is not the case ... that difference turns out to be correlated to the suggested trust/distrust by the responses, which is an indication of selective non-response - selective in the sense that the missing responses seem to be of a particular quality relative to the question. ... Having said all that, let me say a bit about why I don't want to dismiss your argument out of hand. I've seen other strange results for Kamba voting. In some surveys, the results will show low levels of trust, in others high levels of trust. And you do not think this is an issue of study design? Do you think Kamba preferences are that unstable? |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 4:31:01 GMT 3
Mine was not to provide an alternative design, but to interprete the results given. So, I was not suggesting alternative options for the researchers' trust a lot. What I was doing is making an assumption as to what the trust a lot could have meant to respondents. Quite so. But there is no reason to make an assumption that it meant something other than its plain meaning to respondents, especially when trust a lot is perfectly comprehensible, even though the number assigned to the qualitative option is mushy. To be able to claim that these results reveal something about ranking of ethnicities by their trust of outsiders, 100% minus the percentage indicating "at least some distrust" must be the indication of "no distrust". In this case it is not the case ... that difference turns out to be correlated to the suggested trust/distrust by the responses, which is an indication of selective non-response - selective in the sense that the missing responses seem to be of a particular quality relative to the question. Again, the point can be made in terms of belief. If your argument were sound, 100% minus the percentage indicating at least some belief in a proposition p should yield the total of those who do not believe p. But this is not the case in belief surveys, especially about probability. If you run conjunction fallacy tests, and remove everyone who affirms that they believe that a conjunction can't be more probable than its conjuncts, you'll generally find that among those who are left, some of them also believe sentences which entail that they believe that some conjunctions can be more probable than their conjuncts. People have inconsistent sets of beliefs; more generally, they have inconsistent sets of mental states (preferences, intentions....). Having said all that. I seem to be out of luck. I've found the original paper, and here's their assessment of the ignificance of the differences in trust numbers: The pattern by which interpersonal trust diminishes with social distance is a common feature of the way that all ethnic groups in Kenya view the wider world. In Kenya, we detect a slight tendency for Embu and Meru people to be more trusting of other ethnic groups than the Luo (12 percent versus 5 percent). Otherwise, any differences across the country’s main ethnic groups are minor. The way to rescue the trust point---given these findings re significance---might be to insist either on counting Embu or Meru together with Gikuyu, or to rely instead on Kasara's findings which do seem better supported. |
|
|
|
Post by danieldotwaweru on Jan 30, 2012 4:44:19 GMT 3
... Having said all that, let me say a bit about why I don't want to dismiss your argument out of hand. I've seen other strange results for Kamba voting. In some surveys, the results will show low levels of trust, in others high levels of trust. And you do not think this is an issue of study design? Do you think Kamba preferences are that unstable? I think there are several options. (1) Just accept that preferences are weird. There is some evidence for this option: across African elections, ethnic identity strengthens very sharply at election time. Since lots of preferences are strongly associated with ethnic identity, one might expect preferences also to change very rapidly near election time (also, election violence). (2) Decouple preferences completely from political behaviour. There is also some reason for this. First, Kenya has a very volatile political culture: four years ago, Ruto was very unpopular, to say the least, in Central and Eastern, and yet UDM did reasonably well in Eastern a while ago. If political behaviour is used to predict preferences, then they might seem more volatile than they actually are. (3) Accept that there are study-design issues. This is not separate from (2), because a lot of the studies have relied, at least partly, on political behaviour as guide to preferences. (4) Worry that this is something to do with political structure rather than ethnicity. There's a famous study from a while ago which tracked two groups in Zambia and Malawi. In Zambia, they're political enemies, in Malawi they're best of friends. This is supposed to have something to do with the differing structure of the political space. The issues between Pokot and Kalenjin, and between Kipsigis and Maasai, in Kenya are a reasonable approximation. At the local level, they would prefer not to vote together. At the national level, they form a coalition. So maybe we're looking at multi-level preferences. |
|
|
|
Post by mank on Jan 30, 2012 5:05:05 GMT 3
DW, Again, as I said, this is not a quesiton for philosophical debate rather one of designing a study so it can support the test of hypothesis you wish to test. Here you impose an hypothesis on a study, and you are trying any argument to counter my illustration to you why the study design is a hinderance to your desired statement. There is no more for me to say. In an afterthought I can try to clarify my point: that there is a clear selection bias (which I hypothesize was created by ambiguity of the response space). See below: The interview question was: How much do you Trust Kenyans from other Ethnic Groups? Results: | Ethnic Group | % non-response | At least some Distrust | | Kamba | 0.9 | 94.6 | | Luo | 2.7 | 92.6 | | Kikuyu | 0.9 | 91.6 | | Kalejin | 1.0 | 89.3 | | ALL | 3.4 | 88.8 | | Luhya | 6.4 | 87.8 | | mijikenda | 6.7 | 80 |
[/td][/tr][/table] Note that my category "at least some distrust" is the summation of all responses except those indicating "At Lot" (so a summation of % responders under "Not at all", "Just a little"and "Somewhat"). This might help clarify the assumption which you have been disagreeing with unnecessarily. That assumption allows the analyst (me) to look at a response as a coin with two sides (that if your response did not fall in any of the 3 columns, it must fall in the fourth, unless the survey design is bad for the comparison of ethnicities as you presented it). That is you either have at least some distrust, or you do not have any. The correlation coefficient (between "no-response" and "at least some distrust") is -79.36%. This suggests that some people who had some distrust, but did not feel that their REAL response was reflected in any of the 4 options given, were UNABLE to respond. This is a selection bias that is more serious than multicolinearity even though it presents itself as a multicolinearity hazard. Even the authors should not be making the claims that their study can show differences in trust or distrust between various ethnic groups - you have pasted at least one such a claim by them. You cannot test the hypothesis of difference in the groups with such a selection bias. Or, in regular language, you simply cannot tell if there is a difference between responding ethnic groups when the failure to respond is systematically correlated with the quality you are studying (in those ethnic groups)! |
|
|
|
Post by mank on Jan 30, 2012 7:22:08 GMT 3
To be able to claim that these results reveal something about ranking of ethnicities by their trust of outsiders, 100% minus the percentage indicating "at least some distrust" must be the indication of "no distrust". In this case it is not the case ... that difference turns out to be correlated to the suggested trust/distrust by the responses, which is an indication of selective non-response - selective in the sense that the missing responses seem to be of a particular quality relative to the question. Again, the point can be made in terms of belief. If your argument were sound, 100% minus the percentage indicating at least some belief in a proposition p should yield the total of those who do not believe p. That is as it should be! Else a test of hypothesis that two ethnic groups have different trust/distrust levels would be meaningless. The probability that I have at least some belief is 1 minus the probability that I have no belief at all. But this is not the case in belief surveys, especially about probability. What?? If you run conjunction fallacy tests, and remove everyone who affirms that they believe that a conjunction can't be more probable than its conjuncts, you'll generally find that among those who are left, some of them also believe sentences which entail that they believe that some conjunctions can be more probable than their conjuncts. People have inconsistent sets of beliefs; more generally, they have inconsistent sets of mental states (preferences, intentions....). What I point out is a mathematical concept ... not a concept in psychology! We design studies strategically to benefit from mathematical identities in decoding responses. So the design is what has to be mathematically sound ... in this case it is not! |
|
|
|
Post by merlin on Jan 30, 2012 10:26:40 GMT 3
I like to discuss the consequences of distrust between groups though will not get lured into a tribal discussion which tribe is more distrusting than another. Is that so? You could have fooled me: your began your discussion by explaining how horrible, nasty and untrusting Gikuyu were. That's a very strange way to begin a sober discussion about inter-ethnic trust. We're among friends and honourable enemies here. It's OK to admit you wanted to get some Gikuyu-bashing done. Unfortunately, you chose ineptly. Iwe funzo kwako na wengine wenye nia kama hizo. So. Your point, caller? I do not know your intention to misquote me though you cannot do this. If you go back in the thread (page 4) you will discover that I responded to the comment of njamba who stated; This triggered my curiosity not only why they mistrust though more about the consequences of mistrust related to the PEV. It has unintentional to do with the Kikuyu though I recognised the mistrust and fear for others in my personal relations with Kikuyu friends and colleagues before, during and after the PEV of 2007/8. However the same could apply to any other group or tribe. Would you be so courteously to rectify your comment in which you quote the following; The same request applies to your following comment; I am not aware to have stated how horrible and nasty Kikuyus are. Regarding your following comment; You have read the paper and even writing a book about ethnic violence in Kenya. You could be an excellent contributor to the discussion I seek. However I feel there is lack of impartiality in your approach. You seems to be emotionally bound by mistrust even hatred for specific tribes. This disables you to understand what I write, adapting and distorting my quotes to fit into your perception of reality. |
|
